High soil fertility supports greater soil quality, microbial network complexity, and keystone taxa in arable black soil

Abstract

The depletion of soil organic matter (SOM) represents one of the most conspicuous hallmarks of Mollisol degradation in Northeast China. This study focuses on typical black soil with high, medium, and low fertility levels to investigate the responses of soil properties and microbial communities to varying soil fertility. We observed a significant decline in the Soil Quality Index (SQI) and key enzyme activities from high to low fertility levels, which mirrored the synchronous depletion of soil carbon fractions (SOM, POC, and MBC). Our results indicated that while microbial alpha diversity peaked in soils with medium fertility, overarching network complexity was highest in environments with high fertility. Compositionally, this robust community structure was primarily associated with significant increases in the abundance of Actinobacterota and Mortierellomycota. Furthermore, network analysis identified specific keystone taxa, including Gemmatimonadaceae, Roseiflexaceae, Penicillium, and Trichoderma, which are closely associated with efficient nutrient cycling and enhanced disease resistance. This study elucidates the microbial taxa associated with high-fertility soils. And by elucidating these underlying microbial associations, this study provides a solid theoretical foundation for restoring degraded black soils through targeted microbiome regulation. Ultimately, our findings highlight promising avenues for developing novel soil conservation technologies centered on core functional microorganisms.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12866-026-05412-x.

Introduction

The degradation of soil fertility has increasingly become a significant challenge to sustainable agriculture and food security on a global scale [37]. Since the beginning of the 21^st century, global climate change has intensified, coinciding with a continuous increase in population. Under the dual pressures of climate change and population growth, soil ecosystems are facing unprecedented stress [8]. SOM is a fundamental component of soil fertility and a critical driver of ecosystem functions. The continuous depletion of soil fertility and organic matter is gradually evolving into a critical bottleneck that constrains carbon balance in terrestrial ecosystems and agricultural productivity [32]. Traditionally, SOM is considered to be generated primarily through the continuous accumulation of plant and animal residues [9]. However, the formation and stabilization of SOM do not solely rely on the direct input of plant and animal residues. It is a complex transformation process driven by soil microbial groups [57].

In agricultural ecosystems, microorganisms play a critical role in regulating over 90% of energy flow and nutrient supply through processes such as nitrogen fixation, phosphorus solubilization, and organic matter decomposition [22]. The soil ecosystem is a highly complex ecological system in which microorganisms function as indispensable bioengineers [49]. Through their roles in decomposing organic matter, driving nutrient cycling, and facilitating soil aggregate formation, they sustain soil health and functional integrity [14]. Research indicates that microbial metabolic products and microbial remnants are key sources contributing to the stabilization of SOM pools. Their contribution surpasses that of plant-derived carbon that has not undergone microbial transformation [24]. This mechanism underscores the central role of microbial activity in soil carbon sequestration and in maintaining soil fertility. This highlights the importance of thoroughly investigating the composition and functions of microbial communities as a key aspect of enhancing soil quality [24]. Soil microbial communities regulate carbon and nitrogen cycling enzymes, with structural shifts directly affecting nutrient transformation efficiency [54]. During the ecological restoration process, nitrogen-fixing microorganisms and ammonia-oxidizing microorganisms can regulate the soil nitrogen supply capacity, thereby influencing ecosystem productivity [69]. These microbial activities are essential for maintaining ecosystem functioning and nutrient cycling, contributing significantly to soil fertility and overall agricultural productivity.

Soil microbial communities show a high sensitivity to environmental changes. Alterations in land use practices, adjustments in management strategies, and variations in organic matter inputs can all trigger corresponding shifts in the structure and function of microbial communities [17]. In highly productive agricultural regions such as the black soil areas, the gradual degradation of soil fertility significantly reshapes the soil microbiome, directly impacting its capacity for nutrient cycling and overall soil health [60]. Therefore, deciphering how microbial networks and community compositions respond to varying fertility levels is essential for developing targeted restoration strategies in degraded agricultural soils. Similarly, under different tillage practices, variations in the fractions of soil particulate organic matter (POM) are closely associated with changes in the abundance of specific microbial taxa. These changes further influence the stability of soil aggregates [30]. The sensitivity of these microbial communities to environmental disturbances prompts consideration of how scientific management practices can be employed to maintain their stability and functionality[38]. It has been found that the structural stability and functional resilience of soil microbial communities play a critical role in sustaining the long-term viability of agricultural ecosystems [10]. The structural characteristics of microbial co-occurrence networks are closely associated with soil health. The stability of this microbial network structure reflects the health status of the soil ecosystem [64]. Moreover, it is directly related to the sustainability of agricultural production. Although existing research has recognized the significant role of microbial communities in the formation of soil fertility. The understanding of the composition characteristics, network structure, and ecological functions of key microbial taxa in soils with different fertility levels remains relatively limited.

Microbial community assembly is generally governed by two fundamental ecological processes: deterministic processes and stochastic processes [50]. Soil fertility, primarily driven by variations in SOM and nutrient availability, is expected to strongly influence the balance of these processes. Contrasting fertility levels can act as potent environmental filters, exerting deterministic selection pressures on microbial taxa adapted to specific resource niches [9]. However, how the relative importance of stochastic versus deterministic assembly shifts across this soil fertility gradient remains largely unexplored [40]. The key objective of this study was to evaluate the underlying assembly mechanisms of bacterial and fungal communities across different fertility levels, hypothesizing that variations in organic matter content would act as the primary deterministic driver shaping microbial assembly.

In degraded black soil regions, the mechanisms by which high-fertility soils achieve organic matter accumulation and efficient nutrient cycling through the formation of specific microbial community structures remain unclear. Analyzing the composition patterns, network characteristics, and functional potential of microbial communities in soils with varying fertility levels is of critical scientific significance for elucidating the microbial mechanisms underlying soil fertility enhancement. This study focuses on soils with varying fertility levels in black soil regions as the subjects of investigation. This study aims to elucidate, through the integration of soil physicochemical analyses and metabarcoding technologies: (1) the differential characteristics of organic matter components and nutrient content in soils with varying fertility levels; (2) the composition of key functional microbial taxa in high-fertility soils; and (3) the response patterns of microbial community structure, diversity, and co-occurrence networks to gradients of soil fertility. The research findings will provide a theoretical basis for restoring soil fertility through the regulation of microbial communities and establish a scientific foundation for the development of conservation agriculture techniques based on key functional microorganisms.

Materials and methods

Experimental plot

The experimental plot was located at Youyi Farm in Heilongjiang Province (46°38'2" N, 131°30'54" E). The total area of the farm was approximately 1.8 × 10³ hm², with an elevation ranging between 50 and 100 m. The zonal soil in the experimental site was predominantly meadow soil, characterized by a temperate continental monsoon climate, with an average annual temperature of 3.1℃ and an average annual precipitation of 530 mm.

Experimental design

A stable growth area with minimal disturbance and uniform topography, ensured via Digital Elevation Model analysis, was selected within Youyi Farm. To identify the three sampling zones characterized by high, medium, and low soil fertility, we utilized multitemporal optical remote sensing data. Specifically, we extracted the visible and near-infrared bands, which are sensitive to the dark-colored SOM, and combined them with environmental covariates such as vegetation indices and moisture indicators. Based on the spatial variation in spectral reflectance and vegetation growth characteristics, the selected area was stratified into three distinct subzones representing high, medium, and low SOM accumulation potential, respectively [66]. The macroscopic distance between these three distinct fertility subzones was approximately 2–3 km. Within each specific fertility subzone, nine independent sampling plots were randomly established, resulting in a total of 27 plots across the study area (i.e., nine replicates for High, nine for Medium, and nine for Low fertility). To avoid spatial autocorrelation and capture representative within-zone variability, the distance between any two adjacent plots within the same subzone was strictly maintained at a minimum of 10 m. Given the high spatial heterogeneity of soil microenvironments and microbial communities, these plots separated by 10 meters were treated as true independent biological replicates for all subsequent analyses. Detailed experimental layouts can be found in the supplementary material 1 (Fig. S1).

Soil sample preparation and collection

Soil samples were collected from each plot in September 2023. The sampling tool used was a soil auger with a diameter of 5 cm, and the sampling method employed was the “S” sampling technique. Specifically, ten sampling points were selected in each plot following an “S” pattern. First, undecomposed and partially decomposed litter was removed, and surface soil samples were collected from a depth of 0–20 cm, predominantly characterized by black color. High, medium, and low represent the treatments for high-fertility soil, medium-fertility soil, and low-fertility soil, respectively. After thoroughly mixing every ten samples, they were combined to form a single composite sample. Stones, animal matter, and roots larger than 1 mm were removed from the surface soil samples. The samples were then passed through a 2 mm soil sieve, yielding 27 samples comprising three distinct fertility groups, with nine biological replicates per group. The soil samples were initially divided into three parts. Some of them were immediately frozen in liquid nitrogen and stored at −80℃ for strict preservation, to facilitate DNA extraction and high-throughput sequencing. Fresh soil samples were used to determine soil available nutrients, enzyme activity, and general microbial activity indicators. The air-dried soil samples were used to assess certain soil physicochemical properties.

Determination of soil physicochemical properties

Soil physical and chemical properties were systematically determined using standard methodologies. Regarding soil physical properties the determination of soil field capacity (FC) and bulk density (BD) was performed using the core sampling method [6]. Regarding soil chemical properties and nutrient contents the soil pH was measured with a calibrated pH meter (Sartorius PB-10). The total nitrogen (TN) content in the soil was determined using the concentrated sulfuric acid digestion method whereas the total phosphorus (TP) content was evaluated via the digestion molybdenum antimony colorimetric method. Additionally the available nitrogen (AN) content was determined through Kjeldahl digestion and the available phosphorus (AP) content was measured using the ammonium molybdate method [3]. Potassium in the soil extract was quantified using a flame photometer (Shanghai Yidian FP6410) [36]. Regarding SOM characterizations the quantitative analysis of SOM was conducted using the potassium dichromate oxidation method. Under controlled temperature conditions an excess of the potassium dichromate sulfuric acid mixture reacted with the sample to achieve complete oxidation of organic carbon and the residual oxidant was then back titrated with ferrous sulfate solution. Regarding SOM characterizations the quantitative analysis of SOM was conducted using the potassium dichromate oxidation method. Under controlled temperature conditions an excess of the potassium dichromate sulfuric acid mixture reacted with the sample to achieve complete oxidation of organic carbon and the residual oxidant was then back titrated with ferrous sulfate solution. Furthermore the determination of particulate organic matter (POM) and mineral associated organic matter (MAOM) in soil was conducted using a particle size fractionation method [27]. The subsequent determination of soil particulate organic carbon (POC) and mineral associated organic carbon (MAOC) was performed using a combination of wet sieving and high temperature potassium dichromate oxidation [25]. Soil microbial biomass carbon (MBC) was determined using the chloroform fumigation extraction method, wherein organic carbon in both fumigated and unfumigated soil samples was extracted with 0.5 M K₂SO₄ and subsequently quantified using a total organic carbon analyzer,, and an extraction efficiency factor (k_EC) of 0.45 was applied to calculate the total MBC content [53].

Soil enzyme activity assays

Soil enzyme activities were determined using microplate-based fluorometric and spectrophotometric assays [51]. Briefly, 1.0 g of fresh soil was homogenized in 125 mL of 50 mM sodium acetate buffer (pH = 5.0) to create a soil suspension.

The activities of alkaline phosphatase (S-ALP), cellobioside hydrolase (S-CBH), and leucine aminopeptidase (S-LAP) were measured fluorometrically using 4-methylumbelliferyl phosphate, 4-methylumbelliferyl-β-D-cellobioside, and L-leucine-7-amino-4-methylcoumarin as substrates, respectively [13]. The soil homogenate and substrates were incubated in 96-well black microplates at 25 °C in the dark for 4 h. Reactions were terminated by adding 10 µL of 1 M NaOH. Fluorescence was quantified using a Synergy H1 hybrid multi-mode microplate reader (BioTek Instruments, Winooski, VT, USA) with excitation at 365 nm and emission at 450 nm. The activity of soil chitinase (S-CS) was assayed spectrophotometrically [68]. The reaction mixture consisted of the soil suspension, acetyl-CoA, 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB), and oxaloacetate. The enzyme activity was determined by monitoring the increase in absorbance at 412 nm utilizing the same microplate reader.

Calculation of Soil Quality Index (SQI)

The soil quality index (SQI) was calculated using indicators such as nutrient content [63]. Initially, all indicator data were standardized to values ranging from 0 to 1. Following this standardization process, key parameters were selected, and principal component analysis was conducted utilizing the R statistical software environment. Key parameters related to soil health were selected according to the minimum data set principle to ensure accuracy. The principal components with eigenvalues > 1 were extracted, yielding two principal components. PC1 and PC2 together contributed 89.89% of the explained variance. The weights for each indicator were calculated using the factor analysis method (Tables S1 - S3), and the soil quality index was calculated using Equation (1).

1

In the formula, PCi represents the score of the i-th principal component, wi represents the weight of the i-th principal component (usually corresponding to its contribution rate), and k represents the number of selected principal components.

Soil DNA extraction, sequencing and bioinformatic analysis

Genomic DNA was isolated from soil samples using the FastPure Soil DNA Isolation Kit (Mingke Biotechnology, Hangzhou, China). The DNA quality was assessed using 1% agarose gel electrophoresis. Concentration, purity, and integrity were measured with a NanoDrop 2000 (Thermo Scientific, USA). The extracted DNA was used as a template. The bacterial 16S rRNA V3-V4 region was amplified utilizing the specific forward primer 338 F (5′-ACTCCTACGGGAGGCAGC-3′) and the reverse primer 806R (5′-GGACTACHVGGGTWTCTAAT-3′) [15]. The fungal ITS one region was amplified utilizing the specific forward primer ITS1F (5′-CTTGGTCATTTAGAGAGGAAGTAA-3′) and the reverse primer ITS2R (5′-GCTGGTCTTCATCGATGC-3′) [31]. Subsequently, 16S rRNA and ITS rRNA sequencing were performed on the Illumina NextSeq 2000 platform.

Following high-throughput sequencing, the raw sequence data were processed using the QIIME 2 pipeline. Sequence quality filtering, merging of paired-end reads, denoising, and chimera removal were conducted using the DADA 2 algorithm to generate exact amplicon sequence variants (ASVs). For taxonomic classification, representative sequences were systematically annotated utilizing a naive Bayes classifier implemented via the q2-feature-classifier plugin within the QIIME 2 platform. This classifier was specifically trained on our target primer regions to optimize taxonomic assignment accuracy. Bacterial operational taxonomic units (OTUs) were classified against the SILVA database (release 138), while fungal OTUs were aligned and classified against the UNITE database (version 8.0) to ensure reliable taxonomic identification. To ensure comparability across all samples and to account for uneven sequencing depths, the OTU tables were rarefied to a uniform depth of 44,000 reads per sample prior to downstream alpha and beta diversity analyses. Finally, the raw sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) database under the BioProject accession number PRJNA1438304.

Co-occurrence network and structural equation modeling analysis

To explore the interrelationships among microbial taxa, a co-occurrence network analysis was performed. For the co-occurrence network analysis, species were filtered to retain those with a relative abundance greater than 0.05% and occurring in more than 80% of the samples to reduce the impact of rare taxa and the compositional nature of the data. The network was constructed based on robust Spearman’s rank correlation coefficients (|ρ| > 0.7). To rigorously control for multiple-comparison false positives, statistical significance was evaluated using P-values adjusted by the False Discovery Rate (FDR), and only edges with an adjusted P < 0.01 were included in the final networks. Species were filtered based on Spearman correlation values, and the correlation network was visualized using Gephi (v1.0) [1]. The complex path relationship network was subjected to hypothesis testing using the dplyr and lavaan packages for structural equation modeling (SEM) [41, 58]. Initially, an a priori conceptual model was constructed based on established ecological knowledge, hypothesizing that soil physicochemical properties determine microbial community diversity and composition, which subsequently drive soil enzyme activities and overall soil quality. These components were incorporated into the model as observed variables. To facilitate interpretation and ensure model convergence, all input variables were z-score standardized prior to the analysis. The model was fitted using the maximum likelihood (ML) estimation method. For model optimization, a stepwise sequential simplification approach was employed: non-significant pathways (P>0.05) were removed iteratively, and the akaike information criterion was evaluated at each step to ensure that model simplification did not compromise its explanatory power. The adequacy of the final model was rigorously evaluated using a comprehensive suite of fit indices: a non-significant Chi-square test (x²>0.05), degree of freedom (df), the Comparative Fit Index (CFI >0.95), root mean square residual (RMR), and the Standardized Root Mean Square Residual (SRMR <0.08).

Statistical analysis

Statistical analyses and visualizations were performed utilizing SPSS version 25.0 alongside Origin 2021 Pro. One-way analysis of variance (ANOVA) was conducted to evaluate the significance of differences among all treatment groups. For each response variable, normality of model residuals was assessed using the Shapiro-Wilk test, and homogeneity of variance was evaluated using Levene’s test. The α-diversity indices of microbial communities were evaluated using the Wilcoxon rank-sum test in Mothur software version 1.30.2 [42]. All statistical analyses were performed using R [39] version 4.4.3). The Wilcoxon rank-sum test and the Kruskal-Wallis test were performed using the native stats package to evaluate significant differences among groups. To visualize the overall variation in soil physicochemical properties across different fertility groups, non-metric multidimensional scaling (NMDS) analysis was performed based on a Euclidean distance matrix calculated from normalized environmental variables. Microbial β-diversity was evaluated through principal component analysis (PCA) and permutational multivariate analysis of variance (PERMANOVA) based on Bray-Curtis index, utilizing the adonis2 function in the vegan package [34]. Stacked bar plots at the phylum and genus levels were generated in R. LEfSe was used to quantify biomarkers across different groups [43]. The Mantel test was performed using the mantel function in the vegan package to examine relationships between environmental factors, microbial community structure, and soil physicochemical properties. To evaluate the relative importance of stochastic processes in microbial community assembly, the Sloan neutral community model (NCM) was constructed to fit the relationship between the detection frequency of OTUs and their mean relative abundance. The model estimated the immigration rate (m) and the the overall goodness of fit (R²), with calculation executed using the minpack.lm package in R [48].

Results

Fertility degradation is characterized by significant declines in soil quality index and active carbon fractions

In agreement with the a priori remote sensing classification, the comprehensive assessment of soil health revealed a significant gradient corresponding to the defined fertility levels. The SQI exhibited a stepwise decline from high to low fertility, with the high fertility group recording the highest score, significantly surpassing the medium and low groups (P < 0.05; Fig. 1a). This stratification was corroborated by the NMDS analysis, which showed distinct clustering of samples, indicating that the overall physicochemical profile varies significantly across different soil fertility statuses (Fig. 1b). Soil fertility was significantly positively correlated with the activities of key enzymes (S-LAP, S-CBH, and S-ALP), which exhibited a consistent decline from high to low fertility levels, with S-ALP demonstrating the most distinct stepwise stratification (P < 0.05; Table S4).

Fig. 1.

Assessment of soil quality, physicochemical properties, and key organic carbon fractions. a soil quality index (SQI); b non-metric multidimensional scaling (NMDS); c soil fertility and physical-chemical heat map; d soil organic matter (SOM); e soil particulate organic carbon (POC); f microbial biomass carbon (MBC). The results were presented as means ± standard deviation from 9 replicates. Different letters indicate significant differences among treatments (P < 0.05)

Heatmap analysis further visualized these disparities (Fig. 1c), revealing that high-fertility soils were characterized by elevated abundances of key nutrients (e.g., SOM, TN, AN) and improved structural indices (e.g., Porosity and macro-WSA), whereas low-fertility soils were positively associated with higher BD. Moreover, soil carbon fractions, which were pivotal for nutrient cycling, followed a consistent trend. SOM, POC, and MBC were all significantly higher in the high-fertility soils, followed by the medium and low groups (Fig. 1d–f). Collectively, these results demonstrate that the fertility gradient effectively captures the synchronous deterioration of soil quality and the depletion of the carbon pool from high to low levels.

Soil fertility gradient alters microbial community composition with maximum bacterial diversity in medium-fertility soils

Analysis of microbial data revealed that, in terms of bacterial diversity (Fig. 2), the Shannon index was significantly higher in medium-fertility soils than in low-fertility soils (P < 0.05). Additionally, the Simpson index in both medium- and high-fertility soils was significantly higher than that in low-fertility soils (P < 0.05). In terms of fungal diversity, no significant differences were observed in either the Shannon or Simpson indices across the different soil fertility levels (P > 0.05).

Fig. 2.

Diversity of microbial communities in soils with varying fertility levels. a Bacterial α-diversity (Shannon/Simpson); b Fungal α-diversity (Shannon/Simpson); c–d Bacterial and fungal community β-diversity (PCA). Results were presented as means ± standard deviation from 9 replicates. * indicates differences between treatment groups. *P < 0.05; **P < 0.01; ***P < 0.001

In terms of β-diversity, PCA results indicated a significant difference between bacterial communities in medium-fertility soils and those in low-fertility soils (P < 0.05). There was an overlap between high-fertility soils and low-fertility soils. However, for fungal communities, a significant difference was observed between high-fertility soils and low-fertility soils (P < 0.05). These results indicate that both bacteria and fungi altered their community composition and stability to adapt to soil fertility.

Soil fertility is closely associated with microbial community restructuring and the enrichment of keystone taxa

The relative abundance of bacterial and fungal phyla varied significantly across the soil fertility gradient (Fig. 3). In high-fertility soils, the abundance of the bacterial phylum Actinomycetota and the fungal phylum Mortierellomycota showed a significant increase (P < 0.05). In low-fertility soils, the abundance of the bacterial phylum Nitrospirota and the fungal phylum Basidiomycota showed a significant increase (P < 0.05). Notably, several bacterial phyla, including Armatimonadota and Elusimicrobiota, along with the fungal phylum Mucoromycota, exhibited the highest relative abundance under medium-fertility conditions. These indicate that these taxa have a stronger competitive ability under medium-fertility conditions.

Fig. 3.

Microbial composition in soils of varying fertility. a Bacterial composition; b Fungal composition; c Random forest analysis of bacterial genera; d Random forest analysis of fungal genera, which identifies the most important genera affected by soil fertility using a random forest classification model. Results were presented as means ± standard deviation from nine replicates. * indicates differences between treatment groups. *P < 0.05; **P < 0.01; ***P < 0.001

Variations in organic matter content were strongly linked to the emergence and disappearance of specific soil microbial groups, a trend most pronounced in fungal communities. LEfSe analysis revealed that the abundances of fungi such as Phoma sp., Leptodontidium elatius, Keissleriella sp., Pseudogymnoascus roseus, and Ophiocordyceps cochlidiicola were significantly higher in high-fertility soils compared to other groups (P < 0.05) (Fig. S2). Random forest analysis identified the bacterial and fungal phyla that were most important for predicting soil fertility. Among bacteria, Actinomycetota, Firmicutota, Latescibacterota, WS2, BRC1, and Tectomicrobiota ranked among the most influential phyla. Among fungi, Mortierellomycota, Blastocladiomycota, Calcarisporiellomycota, and Basidiobolomycota were also identified as important predictors. Of these, Tectomicrobiota and Mortierellomycota showed the strongest positive association with high soil fertility.

High-fertility soils harbor greater network complexity and dominant key taxa

Microbial network analysis revealed that microbial network complexity in bacterial communities was highest in soils with high fertility (Fig. 4). The key microbial nodes include families such as Gemmatimonadaceae, Roseiflexaceae, and Gaiellaceae. The complexity of microbial networks in low-fertility soils was lower. The core microbial communities include families such as Acetobacteraceae, Bacillaceae, and Gemmatimonadaceae. The complexity of fungal community networks significantly decreases with declining soil fertility. In high-fertility soils, the core fungal communities include genera such as Penicillium, Trichoderma, and Pseudogymnoascus. In medium-fertility soils, core microbial communities that exhibit community differences include genera such as Cladosporium and Collarina. In low-fertility soils, the core microbial taxa exhibiting differences include genera such as Coniochaeta and Acremoniopsis. Collectively, these findings suggest that taxa widely recognized as potentially beneficial microorganisms and biocontrol agents occupy prominent topological positions in high-fertility soils. This network structure implies a theoretically enhanced potential for nutrient cycling and disease suppressiveness in these environments.

Fig. 4.

Microbial co-occurrence networks in soils with different fertility levels. The panels display the microbial co-occurrence networks under different treatments, specifically showing (a) bacterial networks and (b) fungal networks across high, medium, and low-fertility soils. Within the networks, red lines represent positive correlations, blue lines indicate negative correlations, and grey lines denote no significant correlation

SOM acts as the primary driver shaping microbial assembly and regulating overarching soil quality

Analysis of the NCM results indicates that bacterial and fungal communities in soils across different fertility levels were predominantly governed by deterministic processes (0 < R² < 1). The fitting performance of the bacterial microbiome (Fig. S3) was superior to that of the fungal microbiome (bacterial R² = 0.697, fungal R² = 0.647). This difference indicates that stochastic processes played a relatively greater role in the assembly of soil bacterial communities than in that of fungi (Fig. 5). Bacterial and fungal communities exhibit similar compositional dispersion patterns in response to varying organic matter concentrations. Communities in medium-fertility soils exhibited a higher degree of dispersion,, followed by those in high-fertility soils. Microbial communities in low-fertility soils were the least favorable.

Fig. 5.

Assembly processes of microbial communities in soils with different fertility levels. The panels display the results of the Sloan neutral community models (NCM) for (a) bacterial and (b) fungal communities across high, medium, and low fertility soils. To determine how these analyses were performed, the NCM was constructed by fitting the detection frequency of operational taxonomic units (OTUs) against their mean relative abundance, thereby evaluating the relative importance of stochastic processes in community assembly. The calculated immigration rate (m) and the overall goodness of fit (R²) are presented for each fertility level to quantify the fit to the neutral model

The Mantel test analysis revealed that SOM, TN, TK, AP, and FC were significantly correlated with the bacterial community structure. SOM, pH, TK, AP, FC, and BD showed significant correlations with the fungal community structure (Fig. 6). The results indicate that SOM, TK, AP, and FC were the primary environmental variables associated with shifts in soil bacterial and fungal communities. The results from the structural equation modeling indicate that SOM has significant direct pathways connecting to soil nutrients and physical properties. These factors, in turn, are strongly linked to the variations in bacterial and fungal communities, as well as associated enzyme activities, which collectively covary with the soil degradation index. The model exhibited a highly reliable fit to the observational data (CFI = 0.973, SRMR = 0.030, RMR = 0.029, x² = 10.380, df = 6.000, P = 0.110). Full model parameters, including all standardized path coefficients and their significance levels, are detailed in Table S5.

Fig. 6.

Correlation analysis between rhizosphere microbial community structure and environmental factors, along with the structural equation model. a Mantel test diagrams showing the relationships between bacterial and fungal communities and soil physicochemical properties. The rectangles represent correlation heatmaps for soil physicochemical properties. The lines in different colors indicate the significance levels of the correlations between experimental conditions and physiological parameters. Pink lines indicate highly significant correlations (P < 0.01), blue lines indicate significant correlations (P < 0.05), and gray lines indicate no significant correlations (P > 0.05). b Structural equation model diagram, with numbers above the lines representing path coefficients. Green lines indicate positive path coefficients, while pink lines indicate negative path coefficients. The asterisk indicates the significance of the date, with *P < 0.05; **P < 0.01; ***P < 0.001

Discussion

Dynamics of soil quality and bio-physicochemical feedback mechanisms across fertility gradients

The stepwise decline in the SQI across fertility gradients, combined with the distinct clustering observed in the NMDS analysis, reveals a fundamental directional differentiation in the overall functioning of the soil system associated with different fertility levels. This divergence was primarily driven by the natural accumulation of organic matter and nutrients via rainfall-induced surface runoff in high-fertility soils. This substantial organic input strongly regulates soil physical structure. Specifically, the resulting enhancement in macro-aggregate stability and porosity is significantly associated with reduced soil BD [61]. Acting as a binding agent for soil structure, organic matter improves aeration and water retention by promoting aggregate formation. This mechanism aligns with the strong correlation between high porosity and high fertility observed in the heatmap analysis. This finding corroborates the theory proposed by [47] regarding soil organic carbon promoting aggregate turnover, suggesting that a primary physical feature associated with higher soil fertility lies in alleviating soil compaction and reshaping the pore network. Conversely, the high BD characteristic of low-fertility soils reflects physical degradation that potentially restricts root proliferation. Although some studies indicate that excessive fertilization may lead to secondary salinization and quality decline [21], the high-fertility soils in this study predominantly exhibited a positive synergy between structure and nutrients. Consequently, the SQI constructed based on physicochemical properties effectively captures the variations of soil environmental quality across the gradient, demonstrating that the improvement of physical structure is the foundation for maintaining high-level soil health.

As central components of the core material cycle, the synchronous deterioration of SOM and its active fractions (POC and MBC) with decreasing fertility reflects that active carbon pools were more sensitive to fertility variations than total carbon pools [35]. High-fertility soils exhibited significantly higher contents of POC and MBC, primarily because high-fertility environments are typically associated with higher crop biomass and root exudate inputs, providing the soil with abundant exogenous carbon [62]. As a labile active carbon fraction, POC serves as a direct substrate for microbial activity, and its variation often precedes that of total organic carbon [59]. This result identified POC as an early warning indicator of shifts in soil environmental quality. Notably, the highly consistent decline in MBC and SOM in this study indicates that microbial biomass accumulation is strictly regulated by substrate availability [55]. This cascading decline effect of carbon pools suggests that the maintenance of soil fertility depends not only on the accumulation of total carbon but, more crucially, on the rapid turnover of labile active fractions, as the abundance of the active carbon pool is a key factor supporting the energy metabolism level of the soil system.

The significant decline in soil enzyme activities, particularly S-ALP, along the fertility gradient highlights the stoichiometric coupling relationship between microbial metabolic activity and nutrient resource availability. As metabolic products synthesized by microorganisms to acquire limiting nutrients (C, N, and P), the level of enzyme activity directly reflects the intensity of soil biochemical reactions [56]. In this study, the higher activities of S-LAP, S-CBH, and S-ALP in high-fertility soils may be linked to the substrate induction effect, in which abundant organic substrates stimulate microorganisms to secrete more extracellular enzymes to participate in mineralization [71]. The distinct stepwise stratification exhibited by S-ALP may imply that phosphorus cycling was the most responsive process to fertility changes within this ecosystem [19]. Sinsabaugh’s theory of ecoenzymatic stoichiometry posits that enzyme activity is largely influenced by the supply-demand balance of microbial resources [45]; our results further support the concept that a favorable fertility environment is associated with the alleviation of resource limitations on microbial metabolism, thereby sustaining high enzyme activity levels. In summary, enhanced biological activity is a core feature closely linked to efficient nutrient cycling in high-fertility soils, and future soil management strategies should focus more on increasing nutrient bioavailability by regulating enzyme activities.

Differences in bacterial and fungal diversity and composition across soils of varying fertility

In this study, the analysis of soil microbial communities with varying fertility and organic matter content revealed that soil fertility is strongly associated with shifts in the diversity and composition of bacteria and fungi. The Shannon and Simpson indices of the bacterial community were significantly higher in soils with medium fertility compared to other soil groups. This may be linked to the nutrient conditions that support a greater coexistence of species while preventing excessive competition among dominant groups in high-fertility soils [29]. This was consistent with previous research that has found a positive effect of soil fertility on bacterial community diversity [65]. And there was no significant difference in the α-diversity of soil fungi among different fertility levels. However, the response of the fungal community was comparatively slower than that of the bacterial community. This moderate disturbance effect was less pronounced in the fungal community, suggesting that fungi are less sensitive to changes in fertility [44]. Their community structure may be more influenced by other environmental factors.

The soil fertility gradient also corresponds to changes in the relative abundance of specific microbial groups. Under high-fertility conditions, the abundance of Actinobacterota and Mortierellomycota increases, which aligns with previous studies indicating that higher organic matter content is often associated with the enrichment of bacterial Actinobacterota and fungi. This demonstrates their ability to utilize complex organic compounds [33]. In low-fertility soils, the abundance of Nitrospirota and Basidiomycota increases, showing that these microorganisms may play significant ecological roles in nutrient-poor environments, such as nitrogen cycling and decomposition [4]. In this study, the abundance of taxa such as Armatimonadota and Elusimicrobota was higher under medium-fertility conditions compared to other soil types. This may be due to their stronger competitive ability in medium-fertility environments [5, 52]. This aligns with the theoretical hypothesis regarding the adaptability of these microorganisms to specific environments. These findings align with the conclusions in agricultural soil microbial ecology research, which suggest that bacteria dominate nutrient cycling. This study also highlights the niche preferences of specific microbial groups under different fertility conditions. This provides a new perspective for understanding microbial community assembly. Future research should integrate metagenomics and metabolomics techniques to focus on elucidating the metabolic functions of key microbial groups and their pathways in nutrient transformation. This can provide a more precise theoretical basis for the targeted regulation of soil microbial communities.

Differences in core microorganisms and community assembly in soils with varying fertility levels

In this study, we found that soil fertility is strongly linked to the topological structure. Bacterial communities in high-fertility soils exhibit greater network complexity. This indicates that microbial interactions in such soils are more intense and diverse. In particular, the presence of key microbial taxa such as Bacillus, Gemmatimonadaceae, Roseiflexaceae, and Gaiellaceae suggests that these microorganisms may play important roles in nutrient cycling and ecological balance [28]. This may be associated with the rich resource conditions in fertile soils, which support microbial community diversity and functionality [2]. Bacterial community network complexity is lower in low-fertility soils. Furthermore, the core microorganisms include taxa such as Acetobacteraceae, Bacillaceae, and Gemmatimonadaceae. This may be related to the scarcity of resources in the soil and the limited nutritional support available [16]. In contrast, fungal communities are more complex in high-fertility soils. The core presence of microorganisms such as Penicillium, Trichoderma, and Pseudogymnoascus further indicates that fungal functionality and species diversity are enhanced in soils with higher fertility [18]. In low-fertility soils, the complexity of fungal communities significantly decreases, with core microorganisms primarily comprising Coniochaeta and Acremoniopsis. This change may be attributed to limitations in SOM and moisture, which restrict fungal growth and metabolic activity [26]. These topological variations are consistent with existing literature and underscore the significant association between fertility gradients and fungal community structures. Functional predictions further suggested that high-fertility soils are associated with enhanced potentials for both carbon degradation and nitrogen cycling (Fig S4). This suggests that the core microbiome secures ecosystem productivity by synergistically driving organic matter mineralization and nitrogen retention [11]. Interestingly, low-fertility soils showed distinctly high levels of cellulolysis and chitinolysis [46]. This indicates that microbes in nutrient-depleted environments face starvation, compelling them to mine the most recalcitrant carbon pools [23]. Conversely, the high-fertility microbiome capitalizes on readily available labile carbon, enabling it to drive a highly efficient and comprehensive nitrogen cycle. However, because these metabolic pathways are inferred from taxonomic affiliations, targeted metagenomic and functional assays are required to validate these biogeochemical capacities definitively.

At the same time, NCM analysis indicated that both stochastic and deterministic processes contributed to the assembly of soil bacterial and fungal communities. The relatively higher fitting performance (R²) observed for bacterial communities suggests that stochastic processes play a relatively more prominent role in shaping bacterial assemblages compared to fungi [67]. Previous studies have also found that the microbial community assembly processes in sloping fields and garden plots are more random, primarily influenced by homogenizing diffusion [20]. In contrast, the slightly lower fitting performance of fungal communities implies that they might be relatively more constrained by deterministic environmental selection. This may be attributed to the complex adaptability of fungi to specific microenvironmental conditions, as well as their hyphal growth being modulated by various nonlinear deterministic factors [7]. Additionally, environmental variables such as SOM, TK, AP, and FC are significantly correlated with the structural changes of soil bacterial and fungal communities. This is consistent with findings from many studies regarding the regulatory effects of SOM and nutrients on microbial communities [12, 70]. Structural equation modeling analysis further highlighted the topological centrality of SOM as a key hub linking soil microbial communities and revealed potential mechanisms underlying soil degradation. This study provides new evidence regarding the impact of fertility changes on microbial communities and their ecological functions. Future research should further explore the functional characteristics of microbial communities across different soil types and their impact on soil ecological health.

It is important to note that our functional interpretations are primarily based on taxonomic affiliations derived from 16S and ITS amplicon sequencing and co-occurrence network inferences. While these methods provide robust ecological associations, they cannot directly confirm true metabolic functions or ecological interactions. Future studies incorporating metagenomics, metatranscriptomics, or culture-dependent experimental validations are required to mechanistically verify the specific roles of these keystone taxa in nutrient cycling and disease suppression in black soils.

Conclusion

This study thoroughly explored the associations among soil fertility, physicochemical properties, and microbial community structure, and how they relate to soil ecosystem functions. The results showed that higher soil fertility levels are closely associated with the accumulation of organic matter and total nitrogen, and improve the soil aggregate structure and water retention capacity. Microbial analysis indicates that medium-fertility soil has the highest bacterial diversity, while the fungal community responds relatively slowly to fertility changes. In highly fertile soils, the complexity of bacterial and fungal networks was greater, with beneficial microorganisms such as the family Gemmatimonadaceae and the genus Trichoderma occupying central positions within the microbial network. It also confirmed the presence of a stable microbial network constructed around key groups such as Actinobacterota, Mortierellomycota, and Gemmatimonadota. The presence of these taxa is strongly correlated with organic matter accumulation and soil structure optimization, potentially reflecting nutrient cycling and pathogen-suppression functions. The assembly process of soil microbial communities was primarily governed by deterministic factors. SOM, acting as a central hub linking nutrient availability, physical structure, and microbial communities, serves as a critical indicator influencing soil degradation indices. These findings highlight the ecological mechanisms underlying soil fertility enhancement across multiple dimensions. They also provide a theoretical basis and practical approaches for achieving soil health management through microbial regulation.

Authors’ contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Feng Shi, Kang-kang Wang, Chong Luo, Pin-jiao Jin, Huan-jun Liu, and Shuang Wang. The first draft of the manuscript was written by Feng Shi and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This work is supported by the National Key R&D Program of China (2021YFD1500300), the Natural Science Foundation of Heilongjiang Province of China (JQ2024D005), the Heilongjiang Postdoctoral Financial Assistance (LBH-Z25279), and the central government guides the special project of local scientific and technological development - scientific and technological innovation base project (ZY04JD05).

Data availability

The datasets generated and analysed during the current study are available in the National Center for Biotechnology Information repository, https://www.ncbi.nlm.nih.gov/sra/PRJNA1438304.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.