Statistical analysis

After obtaining the raw data, data filtering was first required: using Mothur (1) (Schloss et al., 2009) to remove sequences with an average quality score of ≤ 20, to remove sequences containing N, and to remove sequences with excessively long homopolymers (>10 bp); (2) to remove sequences with excessive primer mismatches (≥4 bp) and to remove primer sequences (Schloss et al., 2009); (3) to remove sequences with excessively short (≤200 bp) and too long (≥500 bp) sequences (Schloss et al., 2009). In addition, (4) bacterial sequences were removed from chimera using UCHIME (Edgar et al., 2011) with the Gold dataset (Haas et al., 2011) as a reference, and fungal sequences were removed from chimera using UCHIME (Edgar et al., 2011) with the UNITE database (Koljalg et al., 2013) as a reference.

The bioinformatics analysis was based on the R.¹ The dataset was analyzed using Mothur (Schloss et al., 2009). Use USEARCH software to cluster valid sequences into operational taxes based on the 97% similarity threshold. Mothur (v.1.30.1) calculated the relative abundance of rhizosphere microbial taxa, Chao1, ACE, Shannon, Simpson, and other alpha diversity indices. Bacterial sequences were systematically classified using the RDP classifier using the SILVA (SILVA 132) database as a reference for OTU sequences. Fungalsequences were systematically classified using the RDP classifier using the UNITE database (Koljalg et al., 2013) as a reference for OTU sequences. The OTU sequences were classified into phylotypes and matched to the SILVA database using PyNAST. The intercommunity distance matrix was generated using UniFrac, and the UniFrac matrix was subjected to principal coordinate analysis (PCoA). Data were analyzed using Statistical Analysis Software (SPSS software, 22.0, SPSS Institute Inc., United States), and treatment effects were determined using Duncan’s multiple range test (p < 0.05). Soil physicochemical properties, soil enzyme activity, and microbial characteristics of different fertilization treatments were analyzed by one-way analysis of variance (ANOVA) for data (p < 0.05). Correlations among the soil microbial compositions and soil properties activities were determined using redundancy analysis (RDA). The RDA was performed using the CANOCO 5 software package. Plotting was performed using Origin 2021. Sequences were uploaded to the National Center for Biotechnology Information (NCBI) Sequence Read Archive under BioProject PRJNA852613 and PRJNA852643.