*2.2. Vegetation Samples*

In each plot, we investigated plant species, abundance, height, coverage and other indicators, and calculate important values. Species diversity is measured by species richness index and Shannon diversity index. All the above-ground parts of the sample were collected to determine the above-ground biomass.

#### *2.3. Soil Sampling*

July is the growing season for plateau vegetation. The middle of July 2019 was selected as the time of soil sample collection. Use a thermometer to measure the soil temperature before sampling. The soil density was determined by a soil compaction meter (TJSD-7500II). Soil samples were collected using a soil auger with a diameter of 5 cm and a sampling depth of 0–20 cm. Ten soil cores were evenly distributed in each quadrat. Five plots were

set for each site. Four quadrats were set for each plot for repetition, and 5 times of sampling were set in each quadrat for mixing. Thus, a total of 15 treatments were set and 60 soil samples were collected.

After removing the rocks and grass roots, the sample is thoroughly mixed and passed through a 2 mm sieve. Approximately 5 g of samples were immediately put in a 2 mL centrifuge tube and brought back at low temperature, and kept in a refrigerator at −80 ◦C for DNA analysis. About 300 g samples were sealed and brought back into the refrigerator at 4 ◦C for soil physical and chemical analysis. About 1000 g of Soil samples are dried for soil nutrient and heavy metal analysis. During the entire collection and processing of soil samples, bacteria and heavy metal contamination should be avoided.

#### *2.4. Soil Laboratory Analysis*

#### Chemical Analysis

Soil pH was determined by mixing the soil sample and water in a ratio of 1:5 using a pH meter. Soil organic C content was determined by a potassium dichromate external heating method [33]. Soil total N was determined by Kjeldahl digestion and automatic azotometer [34]. Soil alkali-hydrolytic nitrogen, total nitrogen, available P, and available K were determined by the method of previous studies [35,36].

The contents of Cu and Zn in the soil were determined by flame atomic absorption spectrophotometry. The main instrument is a TAS-990F atomic absorption spectrophotometer. Soil Pb and Cd contents were determined by graphite furnace atomic absorption spectrophotometry using a 240ZAA atomic absorption spectrophotometer [37].

### *2.5. DNA Extraction, PCR Amplification*

According to the manufacturer's instructions, use the Fast DNA SPIN Kit for Soil (DNeasy PowerSoil Kit, QIAGEN, Hilden, Germany) to extract total soil DNA. Finally, the DNA was eluted with 100 μL DNA eluent in the kit. Dilute the successfully extracted DNA to a concentration of 1 ng/μL and store at −20 ◦C until further processing.

The barcoded primers and Takara Ex Taq (Takara) were used to amplify the *16S rRNA* genes of bacteria using the diluted DNA as a template, and V3-V4 variable regions of *16S rRNA* genes were amplified with universal primers 343F and 798R for bacterial diversity analysis [38]. To verify the size and quality of the PCR products, all of them were electrophoresed in 1.5% (wt/vol) agarose [39,40].

#### *2.6. Cloning, Sequencing and Phylogenetic Analysis*

The quality of the amplicons was visualized using gel electrophoresis, purified with AMPure XP beads (Agencourt), and subjected to another round of PCR amplification. After purification again using AM-Pure XP magnetic beads, the final amplicons were quantified using the Qubit dsDNA Detection Kit. Equal amounts of purified amplicons were pooled for subsequent sequencing [41,42].

The raw sequencing data were in FASTQ format [43]. Pre-processing of double-ended reads, including detection and cleavage of ambiguous bases (N), was performed using Trimmomatic software (Bolger AM: Golm, Brandenburg, Germany) [44]. Low-quality sequences with an average quality score below 20 were cut off using the sliding window pruning method [40]. Parameters of assembly were: 10 bp of minimal overlapping, 200 bp of maximum overlapping, and 20% of maximum mismatch rate. Assembly parameters were: minimum overlap of 10 bp, maximum overlap of 200 bp, and maximum mismatch of 20%. Reads with 75% of bases above Q20 were retained. Reads with chimeras were then detected and deleted. QIIME software (version 1.8.0, Caporaso JG: Boulder, CO, USA) was used to implement the above two steps [45].

Using the Vsearch software (Edgar RC: Cambridge, MD, USA) with 97% similarity threshold, the clean readings were generated by primer sequencing and clustering to generate surgical taxons (OTUs) [46]. We used the QIIME package to choose a representative reading for each OTU; the RDP classifier (confidence threshold 70%) to annotate all representative readings and annotate the Silva database version 123 (*16s rDNA*) [42]; and the blast to annotate all representative reads and blast the Unite database (ITSs rDNA) [43].

#### *2.7. Statistical Analyses*

Statistical analysis software such as Excel 2010 (Microsoft: Seattle, WA, USA) and SPSS 22 (International Business Machines Corporation: Armonk, NY, USA) were used to arrange and plot the measured data. The data were analyzed by Microsoft Excel 2010, and the differences of vegetation data, soil physical, and chemical properties and soil bacterial diversity index were analyzed by One-way ANOVA in SPSS 22. R 3.5.2 was used to perform analyses of species composition and diversity, non-metric multidimensional scaling (NMDS), Adonis and Mantel tests. The OmicShare tool was used to perform the FAPROTAX analysis (https://www.omicshare.com/tools, accessed on 10 September 2021).
