Species diversity and drivers of arbuscular mycorrhizal fungal communities in a semi-arid mountain in China

Study area

The research site was located in the south of Taihang Mountain (112°28′–112°30′E, 35°01′–35°03′N), a site which belongs to the semi-arid area of China. The climate in the test area is temperate continental monsoon, with an annual average temperature of 14.3 °C and an average annual sunshine rate of 54%; the elevation gradient of our study sites ranged from 231 to 432 m above sea level. Soil in the study area is cinnamon (main part is similar to ustalf USDA), and the parent rock was composed mainly of sandstone and shale. The habitat types in this study were bush/wood land, forest land, grassland, and arable land. The bush/wood land included mainly Vitex negundo L, Lespedeza bicolor Turcz and Ziziphus jujuba Mill. var. spinosa (Bunge) Hu ex H.F. Chow, Forest land included mainly Quercus variabilis Bl., Platycladus orientalis (L.) Franco, and Robinia pseudoacacia L. Dominant herbaceous plants in the grassland were Setaria viridis (L.) Beauv., Artemisia princeps H. Lév. and Vaniot, Pennisetum alopecuroides (L.) Spreng., Arthraxon hispidus (Thunb.) Makino, and Rehmannia glutinosa (Gaetn.) Libosch. ex Fisch. et Mey. Finally, the prevalent herbaceous plants in the arable land were Zea mays L., Triticum aestivum L., Ipomoea batatas L., Brassica campestris L., and Lycopersicon esculentum Mill.

Sample collection

In October 2016, soil samples were collected in triplicate at four sites (W1, BW, WL, and F). The sample collection occurred at the root zone of the plant at a soil depth of 5–10 cm (Table 1). Site W1 represented the forest land soil type; site BW had bush/wood soil type; site WL was characterized by grassland soil type; and arable land soil type was represented in site F. These 12 soil samples collected were placed in sterile plastic bags and transported in freezing boxes to the laboratory, and they were stored at −70 °C until further analysis.

Soil geochemical analyses

We analyzed eight different soil factors, including soil pH, water content, available nitrogen (NH${}_4^{+}$-N), available potassium (K⁺-K) and phosphate phosphorus (PO${}_4^{3-}$-P). Soil pH was examined by a pH meter (PX-KS06; Guangzhou Puxi Instrument, Guangzhou, China). Water content was measured by drying soil method, and the content of soil clay, silt, and sand was performed by using a Malvern Mastersizer (Mastersizer2000; Malvern Instruments, Malvern, UK). The available nitrogen and available potassium were analyzed by an Autoanalyzer (SEAL-AA3; SEAL Analytical, Milwaukee, WI, USA); phosphate phosphorus analyzed by NaHCO₃ Mo-Sb colorimetric method.

Molecular analyses DNA extraction

A total of 50 mg soil was used for metagenomic DNA extraction in each sample, using the Fast DNA Isolation Kit (Q-BIOgene; Heidelberg, Germany). The extracts were stored at −20 °C for PCR. 1.0% agarose gels for checking DNA concentration and purity.

Miseq sequencing step

Using the 18S rRNA gene and primer sets of AMV4.5N Forward 5′-AAGCTCGTAGTT- GAATTTCG-3′ and AMDG R 5′-CCCAACTATCCCTATTAATCAT-3′ to amplify the sequences (from soil DNA extracts), the primer had been reported to be acceptable in several previous studies (Sato et al., 2005). The initial PCR reactions were similar to the existing studies of (Xiao et al., 2016), including :25 µL total volumes, 1–2 µL DNA template, 250 mM dNTPs, 0.25 mM of primer, 1× reaction buffer and 0.5U Phusion DNA Polymerase.

The reactions used a 2720 model Thermal Cycler, and initial PCR amplification was conducted under the steps below: 94 °C for 2-min, then 25 cycles of 30-s denaturation at 94 °C, 30-s annealing at 56 °C, 30-s extension at 72 °C, 5-min extension at 72 °C.The second step PCR used a template, which come from the first 5uL product (without dilution). The second step PCR include: one cycle of 3-min at 94 °C, then 8 cycles of 30-s at 94 °C, 56 °C for 30-s and 72 °C for 30-s, and a 5-min extension at 72 °C. The PCR products were separated by electrophoresis (1.5% agarose gel in 0.5 × TBE) and purified using a gel xxtraction kit (Axygen Biosciences, Corning, NY, USA), then the libraries were sequenced by PE300 sequencing on MiSeq v3 Reagent Kit (Illumina) platform (at Tiny Gene Company, Shanghai).

Bioinformatics methods

The sequence reads were analyzed by the combination of software Mothur version 1.33.3, UPARSE (USEARCH version v8.1.1756) and R 3.2.2 (R Development Core Team, 2015), the original FASTQ files were demultiplexed through the barcode (Schloss et al., 2009). The PE reads for all samples were merged based on mothur. The low quality contigs were removed based on screen.seqs command by the settings filter (maxambig = 0, minlength = 200, maxlength = 580, the higher threshold can protect some longer sequences, which may be the correct fragment, maxhomop = 8). The decoded data information was aggregated (97% homology) to operational taxonomic units (OTUs) (Edgar, 2013).

BLAST analysis was conducted using the “Nucleotide collection (nr/nt)” database (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE_TYPE=BlastSearch). No threshold was set for E values, alignment length and identity settings. For each OTU representative sequence, a list of top BLAST hits was compiled. Uncultured clones were deleted from the list of top hits. The BLAST hit getting the highest score was identifed as the match’s species.

Statistical analyses

For the alpha-diversity analysis, Mothur version 1.33.3 software (Schloss et al., 2009) was used to analyze the OTU richness, Coverage, Chao, and Shannon’s indices as reported earlier by Schloss et al. (2009). The values of soil properties and diversity parameters were statistically analysed by SPSS V. 19 software (one-way ANOVA) (SPSS; IBM Corp., Armonk, NY, USA).

The clustering method was used with R v. 3.1.1 software to identify the AMF relationship (based on OTU abundance-based). Further, the indicator species analysis was utilized to identify the AMF communities associated with various habitat types (Dufrene & Legendre, 1997).

Using the Canoco software (Canoco for Windows 4.5 package) (Braak & Smilauer, 2002), we utilized Monte Carlo permutation and distance-based redundancy (db-RDA) tests to explain the correlation between soil AMF and geochemical factors. In addition, the heatmap results of the abundance percentages of AMF genera were obtained by Mothur version 1.33.3 software. The raw sequence information has been deposited into the NCBI database (accession number SRP116770).

Soil properties

For the eight geochemical factors measured, the arable land obtained the maximum values of water content, available phosphorus and available potassium (site F). Meanwhile, the minimum values of water content and available phosphorus were established in the grassland (site WL). In the bush/wood land (site BW), the maximum values of sand content (average 28.9%), but minimum silt content (35.5%) were established (Table 1).

AMF diversity data and community composition

In the current study, we have identified a total of 532,841 sequences and 803 OTUs from the total dataset; there were 320,899 sequences belonged to phylum Glomeromycotina (accounting for 60.2%). The number of sequences in each of the samples ranged from 15,095 to 35,206, and the number of AMF OTUs ranged from 52 to 83 (genetic distances of 3%). The OTUs’ coverage in all soil types reached 99% (Table 2). On the basis of the OTU richness calculated by Chao’s index, the grassland observed the greatest AMF value (site WL: 81). Through the analysis of Shannon’s index, we discovered that the largest AMF diversity was also present in the grassland (site WL: 3.49–3.52 with an average value of 3.51), followed by the arable land (site F: 3.38–3.46 with an average value of 3.43), bush/wood land (site BW: 3.38–3.46 with average 3.42), and the forest land soils (site W1: 2.53–3.15 with an average value of 2.87) (Table 2).

Some variations in AMF community composition at the genus level were also detected among all soil samples. The 119 OTUs that could be classified were affiliated with ten AMF genera, whereas those that could not be identified were assigned as unclassified. The Glomus were the most abundant genera in all samples: 60%–75% in grassland, 70%–75% in arable land, 75%–80% in bush/wood, and 50%–70% in forest land. Meanwhile, their levels varied in the different soil types. Ambispora were found in all samples, but a greater abundance was detected in the grassland and arable land samples than in those of the bush/wood and forest land soils (Fig. 1).

Correlation among the three factors (AMF communities, soil types and environmental condition)

To determine the differences in soil AMF community, the OTU cluster analysis showed that the 12 soil samples were divided into four Soil Types (Fig. 2), and the indicator species analysis revealed that there were 60 AMF indicators (indicator value > 0.25, p < 0.05) in this four groups types, it mainly included bush/wood (Glomus and Diversispora taxa), arable land (Glomus, Septoglomus and Rhizophagus taxa), grassland (Glomus and Septoglomus taxa), forest land (Glomus and Paraglomus taxa) (Table S1).

The top 50 OTUs of all samples were selected and their abundances were compared by heatmap software. It revealed the relative distributions and abundances of the top 50 OTUs in all samples (Fig. 3). More detailed information about the top 50 OTUs was presented in Table S2. There is also a listing of all AMF OTUs and their closest matches in Table S3.

The distance-based redundancy analysis (db-RDA) showed that there was a significant correlation between the combination of eight environmental factors and soil AMF community structure, that 81.9% of the soil community variation was attributed to all environmental factors (Fig. 4 and Table 3). However, using the Monte Carlo permutation test, we found that water content (r² = 0.7332, p < 0.01), available phosphorus (r² = 0.7576, p < 0.01), available potassium (r² = 0.7973, p < 0.01), silt (r² = 0.6461, p < 0.05), and sand (r² = 0.6293, p < 0.05) were important properties (Table 3).