The maize–peanut rotation field experiment started in 2015 at the Halahai Agroecosystem Experimental Station in Nong’an County, Jilin Province, China (44°05′N, 124°51′E). There were four agricultural practice treatments: 1) conventional practice (C), where maize was continuously planted without residue retention; 2) continuous maize cropping with straw mulching (CS); 3) maize–peanut rotation without straw retention (R); and 4) maize–peanut rotation with straw mulching (RS). Considering the operability of agricultural management, the experiments were performed using a split–plot design. Continuous and rotational cropping were applied to the main plots, rotation was adopted for 1–year maize and 1–year peanut, and all peanut straw (approximately 2,000 kg/ha) was returned to the field. Maize straw retention (0%, 30%, 60%, and 100%) was employed for the subplots through a random arrangement with triplicate plots (120 m² per plot). Specifically, this study selected 0% and 60% (approximately 6,000 kg/ha) retention amounts. After harvesting the maize in October, the straw was mechanically crushed and mulched on the soil surface. A strip–till machine was used to prepare a straw–free seeding belt for sowing in early May of the following year. The planting density of maize was 64,000 plants/ha. The application rate of the base fertiliser was 145 N kg/ha, 145 P₂O₅ kg/ha, and 145 K₂O kg/ha, and an additional 75 N kg/ha was applied at the 11th leaf stage of maize. The planting density of peanut was 120,000 plants/ha. The base fertiliser application rate was 86 N kg/ha, 122 P₂O₅ kg/ha, and 115 K₂O kg/ha when planting the peanut in R and RS.

Plants were collected at the tassel stage (VT) for each treatment on 25 July 2020. Three maize plants were randomly selected from each plot (three plots in each treatment). The aboveground parts were cut into small pieces and dried at 80°C to constant weight. To obtain the root samples, a soil column with a diameter of 30 cm and a depth of 20 cm was excavated with the base of the maize stalk acting as the centre. After excavation, the roots were washed in water to remove the soil. The root samples were then oven–dried at 80°C to constant weight. At the physiological maturity stage (5 October), the maize grain yield (14% grain moisture content) was measured within an area of two central rows (65 cm wide × 20 m long) in each plot.

Bulk and rhizosphere soil samples for maize were collected at the 10th leaf (V10), tassel (VT), and dough stage (R4) for each treatment on 3 July, 25 July, and 20 August in 2020, respectively. Bulk soil samples were randomly collected from three cores (4.3 cm in diameter) in each plot at 0–20 cm and homogenised into one replicate. Rhizosphere soil samples were defined as the soil attached tightly to the root surface (within 2 mm). Three plants were randomly collected from each plot. After shaking the roots to remove loosely attached soil, the rhizosphere soil samples were brushed and homogenised into one replicate. Soil samples were collected in triplicate for each treatment. Soil samples were sieved using a 2-mm filter to eliminate plant residues, stones, and other impurities, separated into two parts and stored at 4°C and −80°C for chemical analysis and DNA extraction, respectively.

Soil total organic carbon (TOC), active organic carbon (AOC), dissolved organic carbon (DOC), and microbial biomass carbon (MBC) in both rhizosphere and bulk soils were measured to assess the effects of agricultural practices and plant growth stages on soil carbon. Soil TOC was measured using the K₂Cr₂O₇–H₂SO₄ oxidation method (Nelson and Sommers, 1982). Soil AOC was measured using the potassium permanganate (KMnO₄) oxidation method. Exactly 1.00 g of air–dried soil was mixed with 20 ml of 0.02 M KMnO₄ and shaken at 200 rpm at 25°C for 2 min. Subsequently, the samples were centrifuged at 950 g for 5 min, and 1 ml of the supernatant was pipetted and diluted with deionized water to 50 ml. An ultraviolet spectrophotometer (UV-2450, Shimadzu) was used to measure the absorbance of the diluted samples at 550 nm. The amount of MnO₄ ⁻ consumed was calculated from the difference with the blank group (no soil). The reduction of 1 mM MnO₄ ⁻ is equivalent to the oxidation of 0.75 mM or 9 mg of carbon (Blair et al., 1995). Soil DOC was extracted by agitating 10 g of soil with 20 ml deionized water at 25°C for 1 h. The extraction was centrifuged at 10,000 g for 5 min, and the supernatant was filtered through a 0.22-μm syringe filter. For DOC determination, the filtered supernatant was measured using a TOC analyser. Soil MBC was measured using the chloroform (CHCl₃) fumigation–extraction method (Vance et al., 1987). The soil (20.00 g) was fumigated using purified liquid chloroform for 24 h. The soil was then extracted with 0.5 mol/L K₂SO₄ (1:4 soil:extractant) for 30 min. The unfumigated soil samples were also extracted. After filtration, the filtrate was used to establish the concentration of soluble organic carbon in the soil using a Liqui TOCII analyser (Elementar, Germany). The mass fraction of soil MBC was calculated using the formula MBC = E _C /K _EC , where E_C is the difference between the fumigated and unfumigated soil samples (K_EC = 0.45).

Total soil DNA was extracted using a Fast^® DNA SPIN Kit (MP Biomedicals, Santa Ana, CA, United States) following the manufacturer’s protocol. The primer pairs 338F/806R (5′–ACTCCTACGGGAGGCAGCA–3′; 5′–GGACTACHVGGGTWTCTA AT–3′) (Mori et al., 2014) and ITS5F/ITS1R (5′–GGA AGT​AAA​AGT​CGT​AAC​AAG​G–3′; 5′–GCT​GCG​TTC​TTC​ATC​GAT​GC–3′) (White et al., 1990) were used to amplify the V3–V4 region of the bacterial 16S rRNA gene and region 1 of the fungal ITS gene. Sequencing was performed using the Illumina NovaSeq-PE250 platform (Illumina, San Diego, CA, United States) at Shanghai Personal Biotechnology Co., Ltd., and the sequencing data were analysed using QIIME2-2019.4 (Bolyen et al., 2019). Raw sequence data were demultiplexed via q2-demux. The sequences were then quality–filtered, deionised, and merged, and the chimaeras were removed using DADA2 (via q2-dada2) (Callahan et al., 2016). Non-singleton amplicon sequence variants (ASVs) were aligned with MAFFT (via q2-alignment) (Katoh et al., 2002). Overall, 45,601–113,681 clean 16S rRNA gene sequences per sample and 65,619–146,894 clean ITS sequences per sample were obtained. For downstream analysis, all samples were rarefied to 43,320 16S rRNA gene sequences and 54,546 ITS gene sequences. The sequencing depth was adequate because the good coverage sequences in all samples were above 96%. Taxonomy was assigned to ASVs using the classify–sklearn naive Bayes taxonomy classifier (via q2-feature–classifier) (Bokulich et al., 2018). After rarefying the samples, the alpha diversity metric (observed species) and beta diversity metric (Bray‒Curtis distance) were estimated using the q2-diversity plugin. SILVA (release 132) and UNITE (release 8.0) were used as reference databases for 16S rRNA and ITS genes, respectively. Microbial community compositions were illustrated at the family and genus levels. The ecological guilds of the fungal operational taxonomic units (OTUs) were predicted using FUNGuild (Nguyen et al., 2016). Sequencing data were deposited in the NCBI under the accession number PRJNA776676.

Statistical analyses were performed in IBM SPSS Statistics (Windows 24.0). One–way ANOVA followed by Duncan’s multiple range analysis (p < 0.05) was performed to test the significance of the effect of agricultural practices on root weight and grain yield. Two–way ANOVA followed by Duncan’s multiple range analysis (p < 0.05) was performed to test the significance of the effects of agricultural practices and plant growth stages on soil organic carbon, as well as bacterial and fungal observed species. Principal coordinate analysis (PCoA) and permutational multivariate analysis of variance (PERMANOVA) for microbial community structures were performed based on the Bray‒Curtis distance (ASV level). The Mantel test was performed using the Spearman correlation method in the vegan library in R (Oksanen et al., 2013). The soil TOC, AOC, DOC and MBC were calculated based on the Euclidean distance, and microbial community structures (ASV level) were calculated based on the Bray‒Curtis distance. Spearman correlation analysis between microbial groups and organic carbon fractions was performed using the psych package in R software (Revelle, 2018). At the family level, to clearly visualise the variations in relative abundance among different treatments, the normalised relative abundance of a soil sample (NRA_i) was calculated using the formula NRA_i = l g (RA_i/RA_bulk-V10-C), where RA_i is the relative abundance of a soil sample and RA_bulk-V10-C is the relative abundance in the C treatment at the V10 stage for bulk soil.

The root biomass, shoot biomass, and root–shoot ratio at the VT stage were shown in Table 1. The root biomass in the rotation systems (23.5 g/plant and 23.2 g/plant in R and RS, respectively) was significantly greater than that in the conventional practice (C, 19.3 g/plant). The root biomass in CS was 16.5 g/plant, which was the smallest. The shoot biomass in the straw mulching treatments (182 g/plant and 178 g/plant in CS and RS, respectively) was significantly greater than that in the no–straw retention treatments (164 g/plant and 170 g/plant in C and R, respectively). The root–shoot ratio in CS was the smallest compared with that in the other treatments. For the grain yield (Table 1), compared with the conventional practice (C, 11.7 × 10³ kg/ha), the yield was the smallest in CS (10.9 × 10³ kg/ha) and was the greatest in the rotation systems (12.9 × 10³ kg/ha in the R and RS). The root biomass (r = 0.75) and root–shoot ratio (r = 0.75) had significantly positive correlations with maize yield.

Soil organic carbon fractions were analysed according to agricultural practice and maize growth stage (Figure 1; Supplementary Table S1 in the Supplementary Material S1). Additionally, the levels of these soil carbon fractions under four agricultural practices at three growth stages are listed in Supplementary Figure S1. Straw mulching significantly increased the soil TOC. The soil TOC level was 11.8–14.1 g/kg, 11.6–13.6 g/kg, 11.1–12.7 g/kg, and 10.5–12.3 g/kg in CS, RS, R, and C, respectively. The soil AOC and MBC were significantly affected by agricultural practices and plant growth stages. They were significantly increased by straw mulching. The soil AOC at the V10 and VT stages was significantly greater than that at the R4 stage, whereas the soil MBC at the VT stage was significantly greater than that at the V10 and R4 stages. The level of DOC in the rhizosphere soil samples was significantly influenced by agricultural practices and plant growth stages. The DOC level at the VT stage (222–556 mg/kg) was significantly greater than that at the V10 and R4 stages (87–178 mg/kg and 34–116 mg/kg, respectively). At the VT stage, the mean DOC in the rotation systems (364 mg/kg and 523 mg/kg in R and RS, respectively) was significantly greater than that in the continuous systems (251 mg/kg and 248 mg/kg in C and CS, respectively) (Supplementary Figure S1C in the Supplementary Material S1).

Based on the Bray‒Curtis dissimilarity matrix, PERMANOVA (Table 2) was performed to assess the effects of agricultural practices and plant growth stages on microbial community structures. When PERMANOVA was performed using total soil samples, the bacterial community structure was significantly different between rhizosphere soil and bulk soil (17.4% explained variance), while the fungal community structure was not significantly different. Agricultural practices had greater effects on both bacterial and fungal community structures (10.7% and 26.9%, respectively) than plant growth stages (7.1% and 8.1%, respectively). When PERMANOVA was executed using rhizosphere or bulk soil samples, only the plant growth stage had no significant effect on the fungal community in bulk soil. The PCoA plots also showed the variations in microbial communities in the different treatments (Figures 2A,B). For bacteria, the plots were separated by rhizosphere and bulk soil along axis 1 (19.7%) and agricultural practice C was separated from the other agricultural practices (CS, R, and RS) along axis 2 (6.4%). For fungi, the plots were separated by different agricultural practices along axis 1 (20.8%) for fungi. In addition, the PCoA plots representing different treatments (agricultural practices and maize growth stages) in rhizosphere soil and bulk soil are shown in Supplementary Figure S2 (Supplementary Material S1). The observed species in the samples were calculated to evaluate the effects of agricultural practices and plant growth stages on microbial diversity (Figures 2C,D). The results showed that agricultural practices had a greater effect on fungal diversity, while plant growth stages had a greater effect on bacterial diversity (Supplementary Table S1 in the Supplementary Material S1).

At the family taxonomic level, families with the relative abundance greater than 0.5% were selected (37 bacterial and 31 fungal families, Supplementary Materials S2, S3). The normalised relative abundance [NRA_i = lg (RA_i/RA_bulk-V10-C)] was calculated to visualise the variations in different samples (Figure 3). The bacterial community composition showed remarkable differences between the rhizosphere and bulk soils. The relative abundances of Intrasporangiaceae, Micrococcaceae, Nocardioidaceae, Streptomycetaceae, Sphingobacteriaceae, Bacillaceae, Planococcaceae, Devosiaceae, Rhizobiaceae, Burkholderiaceae, Pseudomonadaceae, and Rubritaleaceae were significantly higher in rhizosphere soil than in bulk soil. In the rhizosphere soil, the relative abundances of Sphingobacteriaceae, Bacillaceae, and Planococcaceae were significantly higher at the maize V10 and VT stages than at the R4 stage, whereas the relative abundances of Intrasporangiaceae, Microbacteriaceae, Micrococcaceae, Nocardioidaceae, Pseudonocardiaceae, and Streptomycetaceae were significantly greater at the VT stage. The fungal community composition showed apparent differences among the four agricultural practices. The relative abundances of Didymellaceae, Sporormiaceae, and Apiosporaceae were significantly greater in agricultural practices with straw mulching (CS and RS) than in those without straw retention (C and R). The relative abundances of Didymellaceae, Sporormiaceae, Torulaceae, Plectosphaerellaceae, and Nectriaceae were significantly higher in the rotation systems (R and RS) than in the continuous systems (C and CS).

At the genus level, the classified bacterial genera whose relative abundances (>0.5%) were higher in rhizosphere soil than in bulk soil were shown in Figure 4A. In the rhizosphere soil, the relative abundances of Arcticibacter and Bacillus were significantly greater at the V10 stage, while the relative abundances of Kribbella, Nocardioides, Lechevalieria, and Streptomyces were significantly greater at the VT stage. The sum of these classified genera was significantly higher at the VT stage than at the other stages. The classified fungal genera (relative abundance >0.5%) are shown in Figure 4B, and their trophic modes (pathotroph, saprotroph, and symbiotroph) were predicted by FUNGuild (Supplementary Table S2 in Supplementary Material S1). Correlations between bacterial genera and fungal genera were assessed using the total samples (Figure 4C). Cylindrocarpon, Mycosphaerella, Penicillium, Paraphoma, Torula, Chaetomidium, Neosetophoma, Helotiaceae, and Tetracladium were negatively correlated with many bacterial genera. Most of these fungal genera were predicted to be plant pathogens.

The Mantel test was used to analyse the correlations between soil carbon fractions and microbial communities (at the ASV level) in different agricultural practices (Table 3). In the C treatment, all carbon fractions had significant correlations with the bacterial community. The soil DOC also had a significant correlation with the fungal community. Straw mulching (CS) increased the correlation between the soil TOC and the microbial community (r = 0.29 and r = 0.19, respectively, in the bacterial and fungal communities, p < 0.05). In the R treatment, the soil DOC and MBC were significantly correlated with the bacterial community. In the RS treatment, the soil TOC and DOC were significantly correlated with the bacterial and fungal communities.

Spearman correlation analysis identified the number of correlations (p < 0.05) between soil carbon fractions and specific microbial groups at the family level (Figure 5). In the C treatment, soil MBC and DOC were correlated (including positively and negatively) with 20 families and 14 families, respectively. In the CS treatment, the soil TOC was correlated (including positively and negatively) with 19 families. In the R and RS treatments, the soil DOC was correlated (including positively and negatively) with 16 families and 19 families, respectively. Bacillaceae, Burkholderiaceae, Chitinophagaceae, Devosiaceae, Intrasporangiaceae, Microbacteriaceae, Micrococcaceae, Sphingobacteriaceae, and Streptomycetaceae were a group of families with significantly positive correlations with SOC fractions, and they also had complex interactions with other microbial families.