1 Introduction

Agricultural management practices are crucial for maintaining soil quality, stability, and sustainable agricultural production (Sarma et al. 2017). Long-term chemical fertilizer use has caused varying degrees of damage to soil nutrients and physical structure, such as severe fertilizer leaching, loss of soil organic matter (SOM), soil acidification, and compaction (Liang et al. 2019; Zhang et al. 2019), seriously affecting crop growth (Curtin et al. 2013) and producing significant environmental filtering effects on soil microbes, further affecting microbial metabolism (Han et al. 2022). Microorganisms, as performers in numerous biochemical metabolic processes and participants in soil biochemical functions, play critical roles in soil quality through microbial community succession (Eo and Park 2016; Dai et al. 2020, Tang et al. 2022). Exclusive use of chemical fertilizers alters the soil nutrient balance, leading to a decrease in the soil C: N ratio, soil acidification, and conversion of soil C metabolism from biochemistry to respiration under acidic conditions and low soil organic carbon (SOC) (Duan et al. 2024), which further reduces the SOC content and leads to a deleterious cycle of soil ecology, continuously degrading soil quality. Therefore, maintaining soil quality while ensuring sustainable agricultural production is essential for agricultural research.

Green manure is considered an essential means of repairing soil damage caused by chemical fertilizers and improving soil quality (Brookes et al. 1985). Green manure addition can improve soil microbial metabolism (He et al. 2020), which significantly affects nutrient cycling and biotransformation processes in soil (Bowles et al. 2014; Hai-Ming et al. 2014). The heterogeneity of soil C and N nutrients caused by exogenous addition of organic matter (OM) is a prerequisite for changes in microbial ecology (Bender et al. 2016; Banerjee et al. 2019). Green manure affects soil microbial communities directly by enhancing soil nutrients, and indirectly by altering soil and crop traits that affect alkaline solubility (Khan et al. 2018; Zhang et al. 2017). Das et al. (2017) showed that organic amendments primarily drive soil microbial community succession by regulating nutrient availability in the soil, with soil C, N, and pH being the main driving factors. Common green manure species include alfalfa, barley, red clover, and mustard oats (Khan et al. 2020); whereas, ryegrass is more widely used for pasture planting and remediation of soil contaminants (Li et al. 2020; Habibul et al. 2021). Ryegrass green manure (RG) incorporation contributes adequate C and N nutrients to the soil, promoting soil microbial activity and succession, thus significantly altering soil microbial metabolism. However, studies on the effects of RG on soil metabolism and functional genes involved in soil cycling are lacking.

Biochar (BC) amendments can have multiple positive effects on soil quality (Gaunt and Lehmann 2008; Johannes 2012). Increasing soil total C (TC), OM, and humic acid contents is an essential approach for BC to maintain soil quality (Bolan et al. 2012; Farrell et al. 2015), and BC also promotes OM stability, improving soil water holding capacity and nutrient retention (Abujabhah et al. 2016; Kamali et al. 2022; Yang et al. 2023a, b). Changes in soil SOC caused by BC also lead to alterations in the soil microbial community structure (Farrell et al. 2015). In one study, BC application significantly enhanced growth of Proteobacteria, Actinobacteria, and Gemmatimonadota in soil (Azeem et al. 2021), which are essential bacterial phyla that play crucial roles in soil biogeochemical cycling. BC addition to soil alters the soil characteristics and microbial habitats, thereby altering the microbial turnover metabolism (Gorovtsov et al. 2020). Previous research has shown that BC-derived dissolved OM can induce soil metabolism towards mineralisation of simple substances, slowing the SOM decomposition rate and thereby altering the soil quality (Azeem et al. 2023). However, current research on the inherent effects of BC on soil functional cycling and soil metabolic characteristics is unclear.

The stability of soil carbon content is of significant importance for agricultural production. The application of organic additives in agriculture plays a crucial role in the accumulation of soil organic carbon, the enhancement of soil microbial metabolism and function, as well as the improvement of soil quality. However, the differences in soil improvement mechanisms at the microscale among different types of carbon sources used as organic additives remain unclear. Therefore, the two materials we selected, biochar and black ryegrass, represent two types of carbon sources used as organic additives. The comparison between them was aimed at investigating the micro-level regulatory differences in soil improvement resulting from the long-term addition of two organic additives with different types of carbon sources. To achieve this goal, we established a long-term agricultural field trial using two management practices (RG and BC) and investigated their effects on soil physicochemical properties, analyzed the composition and symbiotic patterns of soil microbes, and identified soil metabolic profiles using untargeted metabolomics to characterize potential metabolic pathways. Finally, we explored the relationships among microbial clusters, soil functional genes, soil metabolic products, and soil quality properties to reveal the effects of microbial community assembly mediated soil metabolism on soil quality.

2 Materials and methods

2.1 Experimental design

The experiment was conducted in 2022 at the Science and Education Park of Henan Agricultural University, Zhengzhou City, China. To ensure uniform land fertility, prior to the commencement of the formal experiment, we planted ryegrass three times consecutively from 2017 to 2018 on the experimental site. Once the ryegrass matured, it was subsequently removed.

Following two years of soil preparation, the official cultivation of ryegrass commenced at the end of 2018. Tobacco was planted in late March of the following year, and the first season of flue-cured tobacco cultivation began in 2019. In the plot where ryegrass was originally planted, the entire crop was incorporated back into the soil.

The day before plowing, we selected two 1 m2 plots from each replicate of the ryegrass treatment (a total of 2 × 3 × 3 plots) to sample and analyze the nutrient content of the ryegrass (Walker et al. 2022). The ryegrass was excavated with its roots, after which soil and impurities were removed, and the quantity of ryegrass in each plot was calculated. The amounts of ryegrass ploughed in the years 2019, 2020, 2021, and 2022 were 6297.58, 7807.68, 7496.53, and 7583.82 kg hm−2 (dry weight), respectively.

During ryegrass cultivation, only irrigation and tillage were performed, and no fertilizer was applied. Therefore, the nutrients contained in the ryegrass originated from the soil, with the exception of a portion of the C that resulted from photosynthesis. Tobacco stalk BC, which has been continuously applied since 2019, was used in this study. The nutrient contents of the BC were as follows: total nitrogen (TN), 1.12%; total phosphorus (TP), 1.37%; total potassium (TK), 2.31%; and organic carbon, 45.81%.

The soil type under investigation was sandy loam soil, with the following basic physical and chemical properties: pH 7.57, OM 16.76 g kg−1, TN 0.69 g kg−1, alkali hydrolyzed nitrogen (AN) 32.40 mg kg−1, available P (AP) 29.19 mg kg−1, and available K (AK) 135.47 mg kg−1.

The experiment was designed with three treatments (each treatment was replicated three times), and a randomized block design was used for the experiment (Table 1). Each treatment’s pure N quantity was 52.5 kg hm−2, and a N:P2O5:K2O ratio of 1:2:3 was selected. Tobacco-specific compound fertilizer (N:P2O5:K2O = 8:12:20), potassium sulfate (K2O 52%), and calcium magnesium phosphate fertilizer (P2O5 16%) were used as base fertilizers and were applied to strips before transplantation. Potassium nitrate (N:P2O5:K2O = 13.5:0:44.5) served as a topdressing fertilizer, dissolved in water and applied to the planting holes. Each treatment was repeated thrice. The plot area was 66 m2, with 110-cm row spacing and 50-cm plant spacing. Ryegrass was incorporated into the soil prior to plowing, and biochar was applied before plowing.

Table 1 Experimental treatment design
Full size table

2.2 Soil sampling, sample preparation, and treatment

On June 14 and July 14, 2022, the entire crop root system was excavated using a shovel. Excess soil was removed by shaking, the soil within 2 mm of the rhizosphere was carefully cleared with a soft brush and then a fresh soil sample was collected. This sample was subsequently transferred into a 50 mL centrifuge tube, transported on dry ice, and stored at −80 ℃ within 1 h. Topsoil (0‒20 cm) was air-dried and stored in self-sealing bags concurrently with the collection of fresh soil samples.

2.3 Soil environmental factor detection

Soil pH was determined following extraction with deionized water (soil: deionized water = 1:2.5). TC and TN were detected by dry combustion using a CHN-2000 analysis (Leco Corp., St. Joseph, MI, USA). Organic matter (OM) was determined using the potassium dichromate volumetric method, available potassium (AK) was analyzed using a flame photometer, and available phosphorus (AP) was estimated using the bicarbonate leaching method.

2.4 Soil DNA extraction and sequencing of bacteria and fungi

DNA was extracted from soil subsamples using the Mag Pure Soil DNA LQ Kit (Magan) following the manufacturer's instructions. DNA concentration and integrity were measured with NanoDrop 2000 (Thermo Fisher Scientific, USA) and agarose gel electrophoresis. The extracted DNA was used as a template for PCR amplification of bacterial 16S rRNA genes and fungal ITS genes with the barcoded primers and Takara Ex Taq (Takara). For bacterial diversity analysis, V3-V4 variable regions of 16S rRNA genes was amplified with universal primers 343F (5'-TACGGRAGGCAGCAG-3') and 798R (5'-AGGGTATCTAATCCT-3') (Carlos W Nossa and rn A Aas, 2010). For fungal diversity analysis, the ITS1 variable regions of ITS genes was amplified with universal primers ITS1F (5'-CTTGGTCATTTAGAGGAAGTAA-3') and ITS2 (5'-GCTGCGTTCTTCATCGATGC-3') (Mukherjee et al. 2014). The sequencing of polymerase chain reaction products was performed using the Illumina NovaSeq 6000 platform. Raw sequencing data were in FASTQ format. The original sequencing data were analyzed using DADA2 and QIIME2 (Callahan et al. 2016) (Bolyen et al. 2019). The same operational taxonomic units (ASVs) were clustered with 100% similarity. For detailed steps of microbial information analysis, please refer to the supplementary information.

2.5 Shotgun metagenomic sequencing

We had a total of nine samples for metagenomic analysis (3 treatments, each treatment with 3 repetitions). Sequencing of the metagenomic DNA shotgun was performed using the Illumina NovaSeq platform (Illumina, Inc., San Diego, CA, USA). After contaminant and quality filterin, the DNA sample was fragmented by sonication to a size of 350 bp, then DNA fragments were end-polished, A-tailed, and ligated with the full-length adaptor for Illumina sequencing. The cleaned sequences were de novo assembled into contigs using the MEGAHIT software (v1.0.4-beta) assembler. DIAMOND software (V0.9.9) was adopted to blast Unigenes to functional database with the parameter setting of blastp, -e 1e-5(Li et al., 2014). Functional database excluded KEGG (Kanehisa et al. 2014, 2006) database (Version2018-01–01), egg NOG(Powell et al. 2014) database (Version 4.5), and CAZy (Cantarel et al. 2009) database (Version 201801). In addition, DIAMOND(Buchfink et al. 2015) software (V0.9.9) was used to blast the Unigenes to the sequences of Bacteria, Fungi, Archaea and Viruses which are all extracted from the NR database (Version: 2018–01-02) of NCBI and the parameter setting are blastp, -e 1e-5. Using the LCA algorithm (applied in the taxonomic classification of MEGAN software(Huson et al. 2011)) to determine species annotation information for the sequence, abundance information and zero-gene tables for each sample at various taxonomic levels (kingdom, phylum, class, order, family, genus, species) were obtained from the LCA annotation results and gene abundance tables. For detailed steps of shotgun metagenomic, please refer to the supplementary information.

2.6 Non-targeted metabolomics detection of soil GC–MS conditions

Soil samples (1 g) were placed in an EP tube, and 1 mL of methanol: water = 1:3 methanol solution and 1 mL of ethyl acetate were added, with adonitol (10 µL; 0.5 mg mL−1) as an internal standard. After 5 min of ultrasonication in an ice bath with a 45 HZ ball mill and 15 min of centrifugation (12,000 rpm min−1; 4 °C), the supernatant was extracted, and 1 mL of methanol solution and 1 mL of ethyl acetate were added again. The ball milling and centrifugation steps were repeated, and all supernatants were mixed and dried without heating in a vacuum concentrator. Samples were dissolved with 30 µL of methoxyamine hydrochloride (20 mg mL−1 in pyridine) and incubated at 80 °C for 30 min. Metabolites were detected using a combined gas chromatograph (Agilent 7890, Santa Clara, CA) and a Pegasus HT time-of-flight mass spectrometer (GC-TOF–MS, LECO, St. Joseph, MI). The specific process was described by Song et al. (2020).

2.7 Network analysis

Microbial networks were constructed to elucidate complex interactions within the microbial community, with each treatment comprising six replicates. We selected the top 200 genera of bacteria and fungi at two time points (45 and 75 days after transplanting, respectively) to mitigate potential false correlations. Spearman’s correlations among ASVs were computed using the “psych” package. ASVs exhibiting strong correlations (|r|> 0.6, p < 0.01) were then visualized in Gephi (v.0.9.2) using the Fruchterman‒Reingold layout algorithm (1000 permutations). Node-level topological attributes (degree and closeness centrality) were computed, and a modularity analysis was performed using Gephi software. The "igraph" package in R (version 4.1.0) was utilized to compute within-module connectivity (Zi) and among-module connectivity (Pi) for each node in the network. Nodes were categorized as peripherals (Zi < 2.5 and Pi < 0.62), connectors (Zi < 2.5 and Pi > 0.62), module hubs (Zi > 2.5 and Pi < 0.62), and network hubs (Zi > 2.5 and Pi > 0.62).

2.8 Data processing

In R (version 4.1.0), the “pheatmap” package was used to generate the heatmaps of soil function genes, differential metabolites, and biomarkers correlated with differential metabolites. The “picante” package in R (version 4.1.0) was utilized to compute the nearest taxon index (NTI) and β-NTI. SPSS software was employed to conduct analysis of variance (ANOVA) for soil nutrients. The “minpack.lm” package was utilized for neutral community modeling of bacteria and fungi. The “randomForest” package was employed for performing random forest analysis on function genes and environmental factors, microbial biomarkers and differential metabolites, as well as microbial biomarkers and environmental factors. Differential metabolites were determined using the OPLS-DA model, differential microbes were identified through the linear discriminant analysis effect size (LEfSe), and the calculation method was consistent with the microbial interaction network.

3 Results

3.1 Soil physicochemical properties

All soil nutrient indicators exhibited increasing trends in both the RG and BC treatments. The RG treatment significantly increased the contents of quick-acting soil nutrients, i.e., AK, AP, and AN, which were 48.09%, 13.72%, and 15.80% higher than those in CK, respectively (Supplementary Table S1). BC significantly enhanced the contents of TC, TN, SOM, and FA, including a 1.1-fold increase in soil FA content and a 1.21-fold increase in soil MA content. Evidently, the RG treatment readily increased the contents of soil available nutrients, and BC increased the content of soil C components.

3.2 Soil microbial community structure

The Shannon and Chao1 indices reflect the response of microbial α diversity to long-term fertilization. The Chao1 index reflects species richness and changes in rare species, whereas the Shannon index reflects species evenness and abundance (Tang et al. 2023). Compared to CK, the α-diversity of bacteria in the RG treatment showed a decrease trend, and the Shannon index of fungi was significantly reduced in two periods by 15.53‒20.29% (p < 0.05) (Fig. 1A‒D). The α-diversity of fungi in the BC treatment significantly decreased by 9.93‒20.20% (p < 0.05) (Fig. 1C and D). The bacterial microbial assembly process is determined more by random processes (Rsqr = 0.791), whereas the fungi microbial assembly process is determined more by deterministic processes (Rsqr = 0.337) (Sloan et al. 2006). The neutral community model analysis revealed (Fig. 1E, F) that the number of sequences was 46,828 for bacterial samples and 73,288 for fungal samples, resulting in estimated m values of 0.023 for bacteria and 0.003 for fungi, respectively. These results suggest that after the addition of organic material to the soil, the species distribution of bacteria was greater than that of fungi.

Fig. 1
figure 1

Long-term use of organic additives for (A and B) bacterial and (C and D) fungal α diversity. Neutral models for (E) bacteria and (F) fungi. In (A‒D), “45d” represents microbial α diversity 45 d after transplanting, and “75d” represents microbial α diversity 75 d after transplanting

Full size image

Principal coordinate analysis (PCA) based on the microbial classification level (genus) showed that bacterial communities in different treatments clustered together, indicating that both RG and BC treatments altered the soil bacterial community composition (Supplementary Fig.S1 A,B). Long-term addition of organic additives had less impact on the structure of bacterial communities than on that of fungal communities. Changes in the relative abundance of Gemmatimonadota and Actinobacteriota (i.e., two bacterial phyla) were noticeable in the RG treatment (Fig. 2A and B; Supplementary Fig.S2); whereas, long-term BC addition changed the bacterial community structure. Ascomycota and Mortierellomycota in the two materials showed contrasting trends. Notably, the RG treatment resulted in a significant increase in Basidiomycota and a noticeable change in the relative abundance of Waitea at the genus level (Fig. 2A and B). Based on the results of the null model analysis, it was found that the assembly processes of fungal and bacterial microbial communities in each treatment were mainly dominated by stochastic processes (|β-NTI|< 2). Classifying the assembly processes of individual treatments by combining RC bray values, the BC treatment increased variable selection (deterministic processes) in bacterial communities, while both RG and BC treatments enhanced the stochastic processes of soil fungal succession (Fig. 2C, D).

Fig. 2
figure 2

(A) and (B) represent the relative abundance of bacterial and fungal communities at phyla and genus levels, respectively. (C) Assembly processes and (D) βNTI changes for bacteria and fungi under different treatments. In (A) and (B), “45d” represents the microbial structure 45 d after transplanting, and “75d” represents the micrAobial structure 75 d after transplanting

Full size image

Microbial network analysis of each treatment showed that the network nodes were mainly composed of Proteobacteria (26‒36%), Gemmatimonadetes (18‒23%), Actinobacteria (11‒15%), and Ascomycota (7–11%) (Fig. 3A‒C). Network nodes and connections in the RG treatment showed a clear decrease compared to those in CK (Supplementary Table S2), and the interactions between bacteria and fungi in this treatment showed the lowest percentage of positive correlations (64%). Compared to CK, the interaction networks of the RG and BC treatments showed a clear simplification trend. The average degree (8.12, 8.12) and graph density (0.04, 0.03) of the two treatments were the lowest. The average variation degree (AVD) of fungi showed a significant decreasing trend between 45d–75d (Supplementary Fig.S3 A,B), indicating that BC treatment significantly enhanced the stability of the microbial community. In the microbial network, key nodes were determined based on the within modular degree (Zi > 2.5) or between modular degree (Pi > 0.62). Among the 41 ASVs identified as key nodes, the most abundant taxa were Proteobacteria (n = 14), Gemmatimonadetes (n = 9), Actinobacteria (n = 7), and Ascomycota (n = 4). LEfSe analysis indicated that the biomarker species in the RG treatment were Arthrobacter, Piloderma, and Blastococcus, and those in the BC treatment were Winogradskyella, Jiulongibacter, and Finegoldia (Supplementary Fig. S4).

Fig. 3
figure 3

Microbial interaction networks under (A) conventional fertilisation, (B) ryegrass straw incorporation, and (C) biochar application. Node classification of organisms under (D) conventional fertilisation, (E) ryegrass straw incorporation, and (F) biochar application to identify key species within the fertilisation network

Full size image

3.3 Soil functional genes and their metabolic pathways

Organic amendment addition generally led to enrichment of metabolic functional genes in soil. Long-term RG and BC treatments resulted in significantly different abundance and types of nutrient transport genes involved in C, N, P, and S metabolism in the soil compared to conventional fertilization (Fig. 4). The RG treatment resulted in more significant changes in the soil C, N, P and P cycles. Based on the number of genes (KO) level C, N, P, S cycle heatmap, the RG treatment had a greater impact on the soil C and P cycles than the BC treatment (Supplementary Fig. S5). Moreover, the RG treatment enhanced the abundance of C degradation genes in the soil; whereas, only the BC treatment significantly increased the abundance of N-fixing-related genes nifH and Hao (Fig. 4; Supplementary Fig. S5). This suggested that BC treatment can promote soil N fixation and P transport but suppresses most soil metabolic cycles, which is consistent with the soil nutrient data.

Fig. 4
figure 4

Impact of long-term organic amendments on functional gene abundance

Full size image

Kyoto Encyclopedia of Genes and Genomes (KEGG) database difference statistics at the module level showed 14 significantly different modules between treatments, comprising 12 metabolic pathways (Fig. 5A; Supplementary Table S3), including methane metabolism (M00567, methanogenesis, CO2 =  > methane; M00344, xylulose monophosphate pathway), nitrate assimilation (M00615), amino sugar and nucleotide sugar metabolism (M00549, nucleotide sugar biosynthesis), N metabolism (M00530, dissimilatory nitrate reduction), pentose phosphate pathway (M00345, ribulose monophosphate pathway), citrate cycle (TCA cycle), (M00307, pyruvate oxidation), oxidative phosphorylation (M00153, cytochrome bd ubiquinol oxidase; M00151, cytochrome bc1 complex respiratory unit), tetrahydrofolate biosynthesis (M00126, folate biosynthesis), heme biosynthesis (M00121, porphyrin metabolism), ubiquinone and other terpenoid-quinone biosynthesis (M00117, ubiquinone biosynthesis), pentose and glucuronate interconversions (M00061, D-glucuronate degradation), and histidine metabolism (M00045, histidine =  > N-formiminoglutamate =  > glutamate). A total of 167 genes (KO numbers) were involved in differential module pathways.

Fig. 5
figure 5

(A) Number of genes (KO) involved in the metabolic pathways in the different treatment modules. (B) Linear discriminant analysis effect size (LEfSe) showing the biomarker pathways in each soil sample, and explaining their effects on intergroup differences. The random forest model explains the changes in cycle genes associated with the contribution of environmental factors to the participation of the (C) C, (D) N, (E) P, and (F) S cycle genes. TN total nitrogen, TC total carbon, AN alkali-hydrolysable nitrogen, SOM organic matter, FA fulvic acid, HA humic acid, MA menadione, pH pH, AP available phosphorus, AK available potassium

Full size image

Further analysis of the C and N cycle functional genes in the soil showed that, compared to conventional fertilization, the RG treatment mainly enhanced the citrate cycle (TCA cycle) (M00307, pyruvate oxidation) and N metabolism (M00530, dissimilatory N reduction) and decreased ubiquinone and other terpenoid-quinone biosynthesis (M00117, ubiquinone biosynthesis) (Supplementary Table S4). Overall, the RG treatment mainly changed the energy metabolism related cycles in the soil; whereas, the changes in the soil cycle functional gene modules due to the BC treatment were relatively small. LEfSe analysis also showed that peroxisome (LDA = 2.06; p < 0.05), starch and sucrose metabolism (LDA = 2.47; p < 0.05), and transport and catabolism (LDA = 2.08; p < 0.05) were significantly enriched in the RG treatment. Infectious diseases viral (LDA = 2.04; p < 0.05) was significantly enriched in CK, and its relative abundance was 3.96% and 6.26% higher than that in the RG and BC treatments, respectively (Fig. 5B).

The random forest results showed that the functional genes involved in the C, N, P, and S cycles were significantly correlated with soil humus (p < 0.05) (Fig. 5C‒F). Pearson’s correlation analysis further showed that the functional genes involved in the C cycle were more affected by soil physicochemical properties than those involved in the N, P, and S cycles (Supplementary Fig. S6), and the relative abundance of most genes involved in C cycling increased with an increasing soil humus content.

3.4 Soil metabolites and metabolic pathways

Soil metabolomics provides a new way to understand the metabolic changes in soil microecology caused by different C sources. A total of 147 soil metabolites were detected in all soil samples, including organic acids (AC), lipids (LP), alcohols (AH), and sugars (CH). PCA showed that RG and BC treatments significantly changed the distribution of soil metabolites. Metabolites in each soil treatment were separated along the second axis (Fig. 6A). Compared to CK, 16 differentially expressed metabolites (DEMs) were observed in the RG treatment and 15 DEMs were observed in the BC treatment, of which 7 DEMs were regulated in both treatments (p < 0.05, VIP > 1, FC ≤ 0.5 or FC ≥ 2) (Fig. 6B). Among the soil metabolites, AC accounted for 21.23% of the total number of soil metabolite types, followed by LP (19.18%), AH (11.64%), and CH (9.52%) (Fig. 6C). A total of 27 DEMs were identified using OPLS-DA, most of which belonged to CH (25%), AC (22.23%), and AM (14.81%) (Fig. 6D; Table S4). The types of DEMs in soil with long-term addition of organic additives were fewer but higher in content and showed obvious homogenisation in terms of type.

Fig. 6
figure 6

A Principal component analysis (PCA) of soil metabolites after long-term addition of different organic additives. B Differences and similarities between comparison groups. C Classification of soil metabolites in soil samples. D Comparative analysis of the differential metabolite abundance in each treatment based on the abundance of differential metabolites in all samples. A positive sign represents an increase in the differential metabolite in the comparison group, a negative sign represents a decrease in the comparison group, and a blank space indicates its absence. M1‒M27 represent the differential metabolites listed in Supplementary Table S5

Full size image

The differential metabolites were analyzed for metabolic pathways on KEGG, and the metabolites were placed into their corresponding pathways. The metabolites mainly existed in the galactose metabolism pathway (Fig. 7). The RG treatment significantly increased the soil D-galactose, D-fructose, and sucrose contents in galactose metabolism. In the pentose and glucuronate interconversion pathways, the RG treatment significantly increased the D-arabitol content. The BC treatment mainly increased the soil D-mannitol, D-allose, L-pipecolate, and decanoic acid contents in fructose and mannose metabolism, pyrimidine metabolism, lysine degradation, and fatty acid biosynthesis. Notably, acyl sugars were the metabolites with the highest content in the RG treatment soil (Supplementary Table S5). The (Z)-13-docosenamide and (Z)-9-octadecenamide contents were highest in the BC treatment (Supplementary Table S5). In addition, the RG treatment significantly reduced the abundance of 4,4'-(1-methylethylidene) bis-phenol, 3-methylpentanoicacid, nonanoic acid, and decanoic acid, and significantly reduced the levels of soil pollution metabolites such as BAPE (4,4'-[1-methylethylidene] bis-phenol) and bisphenol A (4,4'-[1-methylethylidene]bis-phenol). Furthermore, the RG treatment significantly increased sucrose, 4-(2-methylbutyryl) sucrose, 1-hexadecanol, myo-inositol, other sugars, AM, and AH, and reduced (Z)-13-docosenamide. The BC treatment significantly increased the (Z)-9-octadecenamide, octadecenamide, galactitol, and D-mannitol levels (Fig. 6D; Supplementary Table S5).

Fig. 7
figure 7

Differential metabolite metabolic network. Metabolic pathways were queried in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database for visualization. Bold metabolites indicate that the substance was detected in the results, and different colored markers represent elevation or reduction of differential metabolites in different comparison groups. CK: no organic additives; RG: ryegrass incorporation; BC biochar addition. Red EC numbers indicate those detected in this study

Full size image

3.5 Soil microbial and metabolite correlation analysis

By screening the differential metabolites (p < 0.05, VIP > 1, FC ≥ 2 or ≤ 0.5) in the samples from both treatments, we obtained a total of 27 DEMs. To explore the relationship between soil microbial changes and metabolites, a co-analysis was performed on 27 biomarkers and DEMs according to the treatments, generating a correlation heatmap matrix (Fig. 8). The results showed that all biomarkers in the samples were correlated with differential metabolites, with the strongest correlation in the BC treatment, followed by that in the RG treatment; whereas, CK was mostly significantly negatively correlated with DEMs. The biomarkers of BC treatment were mostly positively correlated with DEMs, with 22 differential metabolites related to BC biomarkers, among which the Winogradskyella and Solirubrum genera were significantly positively correlated with most DEMs. The biomarkers of BC treatment were strongly correlated with AC (6) and AM (3). In the RG treatment, the biomarkers were highly correlated with CH and AH; sucrose, 4-(2-methylbutanoyl) sucrose, and D-galactose were all significantly and positively correlated with microbial biomarkers of RG treatment. This suggested that the RG treatment primarily altered the microbial community structure through carbohydrates and alcohols, thereby affecting the soil C metabolic cycle. In the control group, the biomarkers of CK treatment were mostly negatively correlated with DEM. In summary, ryegrass and biochar can promote the microbial dominated soil metabolic cycle, while single fertilizer can inhibit the soil microbial cycle.

Fig. 8
figure 8

Heatmap of correlations between differential metabolites and differential microorganisms. M1‒M27 represent the differential metabolites shown in Supplementary Table S5

Full size image

Random forest analysis showed that CH, AH, aromatics (AR), and esters (ES) were the main categories of metabolites affected by different soil microorganisms, among which CH accounted for 79.24% of the variation, AH accounted for 76.39%, AR accounted for 77.92%, and ES accounted for 82.72% (Fig. 9A). Fretibacterium and Haloplanus jointly served as key driving factors for AC, AM, ketone (KT), LP, heterocyclic compounds, and N-containing compounds; whereas, Finegoldia acted as a universal key driving factor for all categories of metabolites except AR (Fig. 9A). Notably, the key driving factors for AR showed clear specificity, being mostly positively related only to AR (Fig. 9A), such as g__ Candidatus, Saccharimonas, g__Senegalimassilia, and g__Nitratiruptor. The explanatory power of the differential microorganisms for humic acid was the highest among the environmental factors, followed by OM (Fig. 9B). Notably, the microorganisms driving soil nutrients were mostly positively related to most soil nutrients; however, the microorganisms correlated with AP had almost no positive correlation with other soil nutrients (Fig. 9B).

Fig. 9
figure 9

Random forest analysis of (A) differential microorganisms and soil differential metabolites and (B) environmental factors and differential microorganisms. The total explanatory power of the differential microorganisms on differential metabolites or environmental factors is displayed as a histogram above the heatmap. CH carbohydrate, AC acid, AM amine, AH alcohol, KT ketone, LP lipid, AR aromatic compound, HC heterocyclic compound, NC nitrogen-containing compound, ES ester, TN total nitrogen, TC total carbon, AN alkaline nitrogen, SOM organic matter, FA fulvic acid, HA humic acid, MA humin acid, pH acidity/alkalinity, AP available phosphorus, AK available potassium

Full size image

4 Discussion

Soil quality is a comprehensive evaluation method consisting of multiple indices such as soil nutrients, soil physical structure, and soil microorganisms. As key participants in soil nutrient cycling, soil microorganisms have significant effects on soil quality changes, and the effects of RG and BC treatments on soil metabolism and cycling functions are direct reflections of soil quality improvement. Therefore, observing changes in soil microorganisms, soil metabolism, and cycling functions under long-term RG incorporation and BC addition can further elucidate the mechanisms of soil quality improvement.

4.1 Microbial community succession under different carbon source organic material additions

The impacts of RG and BC treatments on the soil microbial community structure differed significantly (Wang et al. 2020, Xie et al., 2021a, Sun et al. 2022) (Supplementary Fig. S1). In our study, the RG treatment and BC treatment significantly reduced fungal α diversity, possibly due to the addition of organic fertilizers in high pH soil, which decreased fungal access to soil nutrients, thereby restricting microbial growth (Deshoux et al. 2023). Relative to CK, the RG treatment notably enhanced the abundance of Actinobacteria, likely due to their crucial roles in decomposing SOM and solubilizing phosphorus (Xie et al. 2021). The RG treatment introduced substantial cellulose into the soil, creating a food source and habitat for Chaetomium, thereby significantly increasing the genus' relative abundance (Zhou et al. 2023) (Supplementary Fig.S2). The result is consistent with the findings of Zhang et al. (2017). In addition, the BC treatment significantly increased the relative abundance of Nitrospirota, and Gemmatimonadetes (Supplementary Fig.S2). According to the random forest results, the genus g__Candidatus Saccharimonas plays a crucial role in BC-promoted bisphenol A degradation. Integrating these findings suggests that organic amendments selectively influence soil microbial taxa, possibly due to environmental stress changes induced by these amendments (Chen et al. 2018).

In our studies, long-term organic amendments are primarily influenced by soil nutrient factors that affect soil microbial community succession (Xu et al. 2014; Tang et al. 2023), rather than by soil pH (Fig. 9B). However, several other studies have highlighted that soil pH is among the most crucial factors influencing soil microorganisms (Geisseler and Scow 2014; Glassman et al. 2017). This may have been due to the high initial soil pH in this study, which was within the suitable range for microbial growth (7.1‒7.5); on the other hand, the effect of organic amendments on soil pH is limited and changes occur slowly (Chavarría et al. 2016; Khan et al. 2020). Among the different types of organic amendments, there were also noticeable differences in the factors influencing microbial community succession (Fig. 9B). Humic substances play critical roles in various environmental processes and terrestrial ecosystems as they regulate global C and N cycling and plant and microorganism growth (Lipczynska-Kochany 2018). Humic substances have the highest explanatory power for microbial community variation, with humic substances explaining 38.49% of bacteria and 52.19% of fungi variation (Supplementary Fig.S7). This indicates that humic substances are an important factor influencing microbial succession. And humic substances have a greater impact on fungal succession than bacterial succession, possibly due to the biodegradation microbial composition of humic substances being primarily fungi (Collado et al. 2018).

However, the debate over the relative importance of stochastic and deterministic processes in microbial assemblies persists. Deterministic processes are thought to dominate microbial succession at large spatial scales, whereas stochastic processes are more influential at smaller scales (Wang et al. 2013). In our study, biochar (BC) treatment enhanced the variable selection process in bacterial communities, promoting bacterial adaptation to specific environments and the evolution of specific functions. Our findings align with previous studies, indicating that biochar increased the proportion of deterministic processes in bacterial assembly by enhancing variable selection (Feng et al. 2018; Shi et al. 2020; Wang et al. 2021). The biochar treatment increased the variable selection process in bacterial succession, which could be due to the narrow pH tolerance range of bacteria, so the addition of biochar led to a greater influence of pH on bacterial community succession, resulting in an increase in the proportion of variable selection (Rousk et al. 2010; Zhao et al. 2022). In contrast, the microbial assembly of RG-treatment soil is always dominated by random processes, increasing diffusion-limited processes in fungi, which are essential for increasing biological response efficiency, soil nutrient availability, and soil quality improvement. Green manure (e.g., RG) is an essential soil improvement method. Planting green manure can increase the SOM content and soil aggregate structure, leading to changes in the soil pore structure and microbial metabolic activity, which in turn promote dispersal limitation in soil.

In previous studies, symbiotic networks with application of organic amendments have been found to be more complex than those with chemical fertilizer application (Ling et al. 2016). In the present study, long-term RG and BC treatments reduced network complexity (i.e., they resulted in lower average degree and graph density) (Supplementary Table S2). The reduced positive correlation ratio in the RG treatment group suggested intensified microbial community competition. We speculate that after long-term addition of ryegrass, the organic carbon source generated by its decay triggers the competitive elimination of limited carbon resources by soil microorganisms, thus simplifying the microbial community (with the fewest nodes). This results in the simplification of the microbial network (with the lowest positive correlation ratio) (Cao et al. 2024; Goldford et al. 2018). The phenomenon of reduced network complexity in BC treatment may be attributed to the prolonged addition of organic materials, which enhances soil nutrient levels but concurrently inhibits microbial conversion and decomposition of recalcitrant nutrients (Ji et al. 2020), thereby diminishing nutrient acquisition by soil microbes (Tang et al. 2023). The decrease in description length and lower ACD values of the RG treatment indicate its simplification of the microbial network and increase in microbial community stability (Supplementary Fig. S3; Supplementary Table S2). In contrast, the CK treatment led to enhanced microbial interactions under relatively low soil nutrient contents (Zhao et al. 2019; Yuan et al. 2021). Notably, the BC treatment reduced network complexity but displayed complex microbial relationships (total links = 2186), which may have been related to the fact that BC mainly affected the soil microbial community structure through its physical characteristics rather than nutrients, with less influence on soil microbial environmental filtering. In the interaction network, nodes with high within-module connectivity (Zi > 2.5) indicate the importance of these nodes in energy flow and substance exchange within the module, while nodes with high between-module connectivity (Pi > 2.5) indicate their significance in energy flow and substance exchange between modules (Deng et al. 2016). Based on Zi-Pi values, the key fungal taxa belonged to the phylum Ascomycota, and the identified key bacterial taxa were from the phyla Proteobacteria, Actinobacteria, and Firmicutes (Fig. 3D‒F). This study showed that the key fungal taxa belonged to the phylum Ascomycota, and the identified key bacterial taxa were from the phyla Proteobacteria, Actinobacteria, and Firmicutes (Fig. 3A‒C). In summary, both RG and BC treatments simplified the microbial network and screened the key taxa of symbiotic networks, stabilizing the soil microbial community and improving soil quality.

4.2 Response of soil functional genes to RG and BC treatments

The long-term application of organic materials exhibits varying effects on soil C, N, P, and S cycling functional genes (Biederman and Harpole 2013; He et al. 2020). The type of organic material determines the changes in C, N, P, and S cycling functional genes (Fig. 4). The RG treatment significantly increased the abundance of soil C, N, P, and S cycling functional genes (Fig. 4), which is similar to the results of other studies on green manure (Wang et al. 2022; Yang et al. 2023a, b). However, the effect of increasing soil C nutrients was not evident, which may be related to the high nutrient content of the experimental soil. LEfSe analysis showed that the RG treatment promoted the citrate cycle and N metabolism in the soil microbial system (Fig. 5B), with a significant increase in the abundance of citrate cycle M00307 (Supplementary Table S3). Previous studies have indicated that oxidative phosphorylation is an important process in the energy metabolism cycles involved during glycolysis, the TCA cycle, and the electron transport chain. M00153 and M00151 are parts of the proton respiratory chain involved in oxidative phosphorylation (Shen et al. 2022), and RG treatment significantly increased the abundance of both module genes (Supplementary Table S3). This may be due to the addition of a large amount of microbial substrates to the ryegrass, which stimulates the enhancement of soil respiration (Lee et al. 2021) and regulates the proton transfer rate in the respiratory chain. Additionally, the RG treatment significantly increased the M00061 abundance. The M00061 (pentose and glucuronate interconversion) module can degrade glucuronate esters into pyruvate and D-glyceraldehyde 3P, both of which are important intermediates in the energy metabolism cycles (i.e., pyruvate is an intermediate in the TCA cycle and D-glyceraldehyde 3P is an intermediate in glycolysis). This pathway increases the intermediates of the two major cycles, indirectly promoting energy production in soil biochemical cycles. Interestingly, the RG treatment significantly boosted the abundance of energy metabolism-related genes such as M00151, M00307, and M00153 (Supplementary Table S4). This stimulates soil energy production; however, further investigation into energy flow and utilization is warranted.

When C, N, P, and S cycling functional genes were further classified, the RG treatment significantly increased the abundance of C degradation and denitrification genes in the soil (Schleuss et al. 2021) (Supplementary Fig. S5). Accordingly, we hypothesized that the changes of soil cycling induced by ryegrass overturning were induced by the decay of ryegrass. As a nutrient, ryegrass stimulated the increase in the abundance of some circulatory functional genes related to ryegrass decay, which was similar to the way that green fertilizer changed soil cycling genes (LeBlanc 2023), and the increase in the content of alkali-hydrolyzed nitrogen also confirmed this result. It is worth noting that the relative abundance of nifH was significantly increased only in the N-cycle gene of biochar (supplement Fig. 3). nifH, a gene encoding the nitrogen enzyme ferritin in the nitrogen fixation cycle, plays an important role in the soil fixation cycle (Jia et al. 2020), indicating that the addition of biochar can improve the nitrogen fixation ability of soil microorganisms. The total nitrogen content of soil treated with biochar also supported this conclusion. This indicated that BC treatment can enhance soil microbial N fixation ability, and the TN content in the BC-treated soil supported this conclusion. Biochar promoted the increase of some nitrogen-fixing genes, which may be due to the fact that its improvement of soil structure and nutrients promotes the growth of nitrogen-fixing microorganisms in soil, and then promotes the increase of some nitrogen-fixing genes (Fan et al. 2019; Schleuss et al. 2021). Soil P and S cycling genes were mainly affected by soil C components (Fig. 5C and D); whereas, C and N cycling genes were mainly influenced by AK and AP in the soil (Fig. 5E and F). Compared to the relationship between N, P, and S cycling-related genes and soil nutrients, C cycling genes showed a higher correlation with soil nutrients (Supplementary Fig. S6).

Several studies have indicated significant correlations between soil function and microbial diversity (Jiao et al. 2021). However, in the present study, the fungal microbial diversity of the C, N, P, and S cycling functional genes in the soil did not present a significant correlation; whereas, a negative correlation was observed with bacterial microbial diversity (Supplemental Fig. S9), which contradicts some existing research results (Chen et al. 2020). The nutrients released by organic fertilizers stimulated the increase in the abundance of functional genes in soil (Tang et al. 2023), but the nutrients released were in a narrow spectrum. Due to the difference in carbon utilization capacity of different bacteria, it was only conducive to the growth of some bacteria, so bacterial diversity was negatively correlated with functional genes (Hanson et al. 2008; Rinnan and Baath, 2009). In this study, RG and BC treatments increased the abundance of ABC transport-related genes in soil. The ABC transporters primarily facilitate the transport of substances inside and outside microbial cells (Supplementary Table.S6). The enhancement of this cycling process can promote nutrient absorption and waste excretion by soil microbiota, thereby indirectly accelerating the overall soil cycling (Chen et al. 2021; Trivedi et al. 2013). Therefore, the enhancement of ABC transport indirectly demonstrated the promotion of metabolic cycling in the soil. This also indirectly indicates that RG and BC treatments can promote soil cycle metabolism.

4.3 Response of soil metabolism to RG and BC treatments

Microbial metabolism is collectively determined by various factors, among which the microbial community structure and environmental factors are key components (Fierer 2017). Organic amendments can induce changes in soil metabolism through their effects on the soil microbial community structure and the substances they release (Zhou and Ning 2017). The metabolic profiles of BC and RG treatments differ significantly from those of conventional fertilizer treatments. Compared to CK, nine categories of differential metabolites, including CH, acids, AM, AH, LP, and aromatic hydrocarbons, were identified in the soil treated with organic amendments (Fig. 6C). CH are important sources of energy for plants and microorganisms, and the stimulatory effects of CH on SOM have been demonstrated in multiple studies (Gaudel et al. 2022). Sucrose, the most common disaccharide in the soil environment, is a direct energy source for soil microorganisms that originates in higher plant tissues (Reid and Abratt 2005). In the RG treatment, there is a significant presence of higher plant residues. The decomposition of these residues provides abundant sources of sucrose, which explains the increase in sucrose content in the rhizosphere soil of the RG treatment. It is noteworthy that acyl sugars are the metabolites with the highest content in the differential metabolites of ryegrass treatment (Supplementary Table.S4). It has been reported that acyl sugars exist in multilayered defense interactions, attracting predators by causing insects to consume acyl sugars to produce volatile fatty acids to achieve an insect resistance mechanism (Weinhold and Baldwin 2011). Previous studies have shown that (Z)-13-docosenamide, isolated from the halotolerant Bacillus subtilis BS3, has powerful antiviral capabilities (Donio et al. 2013), and (Z)-9-octadecenamide has antifungal and antibacterial effects (dos Reis et al. 2019). While (Z)-13-docosenamide and (Z)-9-octadecenamide are significantly enriched in the BC treatment, this indicates that the type of organic additive changed the disease resistance metabolism of the soil based on its own characteristics. In addition, the BC treatment markedly elevated the levels of octadecanamide, D-mannitol, and galactitol within the soil matrix (Fig. 6D; Supplementary Table S5). It has been reported that octadecanamide, D-mannitol, and galactitol can be used as plant protectors to enhance crop resistance (Wen et al. 2022; Yue et al. 2023; Yu et al. 2023); therefore, an increase in crop resistance by BC treatment may be achieved by increasing the number of plant protectors.

Organic additives play a crucial role in enhancing soil quality and mitigating soil pollutants. Previous studies have demonstrated that bisphenol A, primarily utilized in plastics and other industrial applications, serves as a significant soil pollutant, entering the soil through agricultural practices such as film coverings (Fu et al. 2023). In this study, the soil bisphenol A content was notably lower in both RG and BC treatments compared to the CK, with the BC treatment showing particularly significant reduction (Fig. 6D; Supplementary Table S5). Biochar (BC treatment) can effectively adsorb bisphenol A in soil through its porous structure (Li et al. 2023); on the other hand, biochar can generate sulfate radical anions by activating persulfate and efficiently degrading bisphenol A (Liu et al. 2020). Therefore, BC treatment can reduce the bisphenol A content in soil. It has been shown that the exudates from the OM of ryegrass can effectively degrade bisphenol A in soil (Gattullo et al. 2015), and the absorption and transformation of bisphenol A during the ryegrass growth process is another important reason for the decrease in the soil bisphenol A content.

5 Conclusion

Long-term addition of organic amendments (i.e., RG and BC) significantly affected the soil microbial community structure, functional genes, and metabolic products, thereby improving soil quality. The RG and BC treatments reduced soil microbial diversity and network complexity, altered microbial assembly processes, and increased the structural stability of the microbial community. The RG treatment had a stronger promotional effect on C, N, P, S, and other nutrient cycling and metabolic capabilities in the soil, and a greater regulatory effect on CH in the soil. However, the BC treatment had a greater effect on degradation of organic pollutants and metabolic products of stress resistance. Therefore, RG and BC treatments can drive soil microbial community changes, improve soil metabolic pathways, and enhance the quality of tobacco field soils.