Analysis of the community composition and bacterial diversity of the rhizosphere microbiome across different plant taxa

Abstract Rhizobacteria play an important role in bridging the soil and plant microbiomes and improving the health and growth of plants. In this study, the bacterial community structures and compositions of rhizosphere microbiomes associated with six plant species, representing two orders and three families of wild plants grown in the same field, were evaluated. The six plant species examined harbored a core and similar bacterial communities of the rhizosphere microbiome, which was dominated by members of Rhizobiales, Sphingomonadales, Burkholderiales, and Xanthomonadales of Proteobacteria, Subgroup 4 of Acidobacteria, and Sphingobacteriales of Bacteroidetes. Plant species had a significant effect on the microbial composition and Operational Taxonomic Unit (OTU) abundance of the rhizosphere microbiome. Statistical analysis indicated a significant differential OTU richness (Chao1, p < 0.05) and bacterial diversity (Shannon index, p < 0.0001) of the rhizosphere microbiome at the plant species, genus, or families levels. The paralleled samples from the same plant species in the PCoA and hierarchical cluster analysis demonstrated a clear tendency to group together, although the samples were not strictly separated according to their taxonomic divergence at the family or order level. The CAP analysis revealed a great proportion (44.85%) of the variations on bacterial communities could be attributed to the plant species. The results demonstrated that largely conserved and taxonomically narrow bacterial communities of the rhizosphere microbiome existed around the plant root. The bacterial communities and diversity of the rhizosphere microbiome were significantly related to the plant taxa, at least at the species levels.


| INTRODUC TI ON
The microbes in the rhizosphere are a diverse mixture of microorganisms that can actively interact with the host plant in different ways. Since the rhizosphere represents the interface between the soil and plants, the rhizosphere microbiome is thought to have substantial importance in bridging the soil and plant microbiomes and improving plant host health and soil fertility (Berg et al., 2005;Bulgarelli, Schlaeppi, Spaepen, Themaat, & Schulze-Lefert, 2013;Turner, James, & Poole, 2013). The rhizobacterial microbiota improves plant health by protecting the plant hosts from phytopathogens, providing them with relevant nutrients by biologically fixing nitrogen, and producing phytohormones to promote plant growth or enhance plant fitness (Berendsen, Pieterse, & Bakker, 2012;Bulgarelli et al., 2013;Mendes, Garbeva, & Raaijmakers, 2013;Pii et al., 2015;Spaink, 2000;Tian, Yang, & Zhang, 2007). To fully understand the functions and activities of the rhizosphere microbiome for the beneficial management of plant health, it is necessary to explore the composition, assembly, and variation of the microbial communities that are present in the rhizosphere and the underlying mechanisms that drive microbiome assembly.
Recently, the communities, composition, and variation of the plant root-associated microbiome from several plant species, such as the model plant species Arabidopsis (Bulgarelli et al., 2012;Lundberg et al., 2012;Schlaeppi et al., 2014), and economically important crop plants such as maize (Peiffer et al., 2013), rice (Edwards et al., 2015;Knief et al., 2011), potato (Rasch et al., 2006), tomato (Tian, Cao, & Zhang, 2015), tobacco (Robin et al., 2006), and soybean (Mendes, Kuramae, Navarrete, Veen, & Tsai, 2014;Xu et al., 2009), have been revealed using culture-independent 16S rRNA gene-based sequencing techniques. These studies have given us a glance about the bacterial community, composition, and diversity of the rhizosphere microbiome and its relationship with the soil microbiome. The community structure and composition of the plant-associated microbiome depends on several factors, such as the soil properties, plant nutritional status, climate, plant genotype, and even the developmental stage of the host plant (Bulgarelli et al., 2015;Pii et al., 2016;Trognitz, Hackl, Widhalm, & Sessitsch, 2016;Turner, James, et al., 2013). Plants recruit their own microorganisms from the surrounding soil and provide entry into the root. Soils provide the bacterial inoculum and serve as a pool of bacterial species present in each soil type (Bulgarelli et al., 2013;Dombrowski et al., 2017;Pii et al., 2016). Therefore, it is clear that both soil type and plant species affect the microbial community and composition of the rhizosphere microbiome (Berg & Smalla, 2009;Bulgarelli et al., 2012;Inceoglu, Abu Al-Soud, Salles, Semenov, & Elsas, 2011;Lundberg et al., 2012). However, the effects of the factors on the community compositions in the rhizosphere and endosphere microbiomes were significantly different.
Compared with the bulk soil and endosphere environment, the biomass and activity of microorganisms in the rhizosphere are enhanced as a result of the exudation of compounds by the roots (Chaparro et al., 2013;Raaijmakers, Paulitz, Steinberg, Alabouvette, & Moënne-Loccoz, 2009;Stringlis et al., 2018).
Therefore, plants can also influence the structure and function of the bacterial communities in the rhizosphere soil around their roots. The studies on the microbial rhizosphere communities have shown the significant influence of plant species and cultivars in shaping microbial communities in the rhizosphere, including studies on different cultivars of potato (Inceoglu et al., 2011;Weinert et al., 2011), maize (Peiffer et al., 2013), Arabidopsis (Bulgarelli et al., 2012;Lundberg et al., 2012), rice (Edwards et al., 2015;Knief et al., 2011), and soybean (Mendes et al., 2014;Xu et al., 2009), or intraspecies comparison (Bouffaud, Poirier, Mulle, & Moënne-Loccoz, 2014;Bulgarelli et al., 2015;Ofek, Voronov-Goldman, Hadar, & Minz, 2014;Pongsilp, Nimnoi, & Lumyong, 2012;Schlaeppi et al., 2014;Turner, Ramakrishnan, et al., 2013;Wieland, Neumann, & Backhaus, 2001). Even different genotypes of the same plant species have also an effect on the bacterial community structure and composition of their rhizosphere microbiome (Marques et al., 2014;Rasch et al., 2006;Robin et al., 2006). These studies have demonstrated that the diversification in the community structure of the rhizosphere microbiome can be partially explained by the phylogenetic distance of the plant hosts. For example, Bulgarelli et al. (2015) found that the host genotype accounts for approximately 5.7% of the variance in the rhizosphere microbiome composition. However, the degree to which the plants contribute to the rhizobacterial communities and the underlying mechanisms by which the plants drive the rhizosphere microbiome are not well understood. The effects on the rhizosphere microbiome of plant species, cultivars, or even genotypes should be subjected to more research, but a study from the higher phylogenetic distance of plant hosts is lacking.
To examine the degree to which the plant taxa drive the assembly of bacterial communities in specific soil environments, we evaluated the bacterial community structures and compositions of rhizosphere microbiomes associated with six plant species representing two orders, three families, and six genera of wild plants grown in the same field using high-throughput DNA sequencing techniques.
Understanding the mechanisms that shape and drive the microbiome assembly in the rhizosphere will provide a basis on which to construct a healthy plant rhizosphere microbiome to benefit plant breeding, improve soil management strategies, and introduce universal biological control agents and fertilizers to develop more sustainable agricultural practices. Plants in the late vegetative stage (6 or 7 months old) were harvested separately, and the roots were shaken to remove the large soil particles. The soil that attached tightly to the roots was carefully collected with a sterile filter paper strip and used as the source of rhizosphere soil (Bulgarelli et al., 2012;Tian et al., 2015).

| Soil collection and metagenomic DNA preparation
For each plant, five replicates were randomly collected. Therefore, a total of 30 rhizosphere soil samples were obtained (Supporting Information Table S1).
The total genomic DNA was separately extracted using a Power Soil ® DNA Isolation Kit (Mo Bio Laboratories, Carlsbad, CA, USA) according to the manufacturer's instructions. The extracted genomic DNA was dissolved in 50 μL of elution buffer and stored at −20°C for subsequent sequencing.

| PCR amplification and highthroughput sequencing
The concentration and purity of the metagenomic DNA extracted were measured using a spectrophotometer (NanoDrop 2000, Thermo Scientific, Waltham, MA, USA). Approximately, 400 bp DNA fragments of the bacterial 16S rRNA gene targeting the hypervariable region V3-V4 were amplified using the primer pair 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) fused with the Illumina MiSeq adaptors and a 6 bp barcode sequence unique to each sample (Tian et al., 2015). The PCR amplification products were subsequently purified, combined in equimolar ratios, and subjected to high-throughput sequencing on an Illumina MiSeq sequencing platform to produce paired 250-nucleotide reads at Sangon Biotech (Shanghai, China).

| Data processing and bacterial community analysis
The raw sequence was spliced using FLASH (version 1.2.3), which can generate much longer reads by overlapping and merging read pairs (2 × 250 bp) before assembling a gene segment (Magoč & Salzberg, 2011), and adaptors, barcodes, and primers were removed using Cutadapt (version 1.9.1). Sequences with ambiguous bases, average quality scores <25, or lengths shorter than 200 bp were removed to control sequence quality. Chimeric sequences were identified and removed with a de novo method using USEARCH (version 8.1.1861) (Edgar, 2010). After the removal of the chimera, high-quality bacterial sequences were collected for subsequent analysis. A summary of data processing steps is provided in Supporting Information Table   S2.
To correct for the differences in sequencing depth, bacterial read numbers per sample were rarefied to the smallest number of reads.
Effective bacterial sequences were separately subsampled for each sample for the subsequent statistical analysis. After subsampling, the data were processed using a modified SOP pipeline based on USEARCH and the software package QIIME (Caporaso et al., 2010;Tian et al., 2015). Briefly, the selected sequences were clustered to Operational Taxonomic Units (OTUs) using a two-stage clustering algorithm with USEARCH (version 8.1.1861) at 97% sequence identity (Edgar, 2010). Representative sequences in each OTU were aligned to the SILVA reference alignment (Database release 128 updated September 2016) (Yilmaz et al., 2014). Taxonomy was subsequently assigned to each representative sequence using RDP with a minimum confidence of 85%.

| Statistical analyses
The diversity index and species richness estimator (α-diversity) for each sample, including OTU richness, Chao-1 diversity, ACE diversity, and the Shannon index, with respect to a sequence depth of 3%, were calculated using QIIME script function alpha_diversity.py (version 1.8.0; Supporting Information Table S3). Rarefaction and rank-abundance curves were calculated at a level of 97% similarity of the OTUs. Statistical analysis was performed using an analysis of variance (ANOVA) with p values to determine whether the diversity indices or species richness estimators were statistically significantly different among the plant rhizosphere soil samples (Cúcio, Engelen, Costa, & Muyzer, 2016). In addition, the statistically significant differential OTUs (p < 0.05) in the different sample groups were identified on a normalized OTU table by comparing OTU frequencies of the within-group to the between-group using the QIIME script function group_significance.py (Caporaso et al., 2010). Relative abundances of the 100 most differentially abundant OTUs in each sample were visualized by drawing a heatmap.
To estimate the beta diversity, weighted UniFrac distances were used to calculate the similarities of the memberships and structures found in the various plant species at the OTU levels (QIIME script function beta_diversity.py). PCoA plots were used to visualize the difference in bacterial community and compositions of the plant-associated microbiome. Canonical Analysis of Principle Coordinates (CAP) was computed using the function capscale from the R Package Vegan (Anderson & Willis, 2003;Oksanen et al., 2015). Variance partitioning and significances on bacterial communities for experimental factors, including the taxonomy, life_cycle, and root_system, were determined by running a permutation-based ANOVA test using 999 permutations (Supporting Information Table S2).

| Sequencing quality control and summary
A series of processes were used to control sequence quality: screening, filtering, preclustering processes, and chimera removal, resulting in 822,483 reads of high-quality bacterial 16S rRNA V3-V4 gene sequences; an average of 27,416 ± 2,228 reads per sample (min = 20,850, max = 32,016) was obtained (Supporting Information Table S3). The bacterial read numbers per sample were rarefied to the smallest number of reads. In this case, 20,850 effective bacterial sequences were randomly extracted for subsequent statistical analysis (Supporting Information Table S4).
Core microbiome analysis using QIIME software covering all six plant species revealed a total of 1,109 core OTUs belonging to 113 bacterial genera of 25 classes, accounting for 73.46% of the total sequencing data. The predominant genera (above 1% of the total reads belonging to core OTUs) included Blastocatella, Ferruginibacter, Bradyrhizobium, Variibacter, Sphingomonas, Variovorax, Acidibacter, and some of unclas-  Information Table S5).
Variations in microbial community compositions and OTU abundance were observed in six different plant species For example, F I G U R E 2 Heatmap depicting the relative abundance of the most 100 differentially abundant Operational Taxonomic Units (OTUs) in six plant rhizosphere soil sample. Dendrogram linkages and distances of OTUs are not phylogenetic, but based upon reads number (log transformed) of OTUs within the samples. Legend and scale shown in the upper right corner of the figure represent colors in heatmap associated with the relative abundance of OTUs (cluster of variables in Y-axis) within each plant and soil sample (X-axis clustering). Aco, Ageratum conyzoides; Ean, Erigeron annuus; Bbi, Bidens biternata; Aar, Artemisia argyi; Vja, Viola japonica; Ehi, Euphorbia hirta; Con, Bulk soils. The corresponding taxonomic profiles for each OTU were presented in Supporting Information Table S6 Proteobacteria was the predominant bacterial group of the rhizosphere microbiome, being the least represented (29.33% ± 2.50%) in E. hirta and the most represented (41.55% ± 2.59%) in V. japonica (Figure 1a). At a more detailed level, Rhizobiales (7.78% ± 2.58%) of Proteobacteria was highly enriched in V. japonica and A. argyi, and a higher proportion of Myxococcales (4.53% ± 0.91%) in V. japonica, similar to that of Nitrosomonadales (4.28% ± 1.24%) were observed in E. hirta, Sphingomonadales (3.23% ± 1.29%) and Burkholderiales (2.50% ± 0.54%) in E. annuus, Xanthomonadales (3.15% ± 0.99%) in V. japonica, demonstrating a highly varied community composition at the order level (Figure 1 and Supporting Information Figure S1).
Similar results were also observed in other bacterial groups from the bacterial phyla Bacteroidetes and Acidobacteria. The OTUs belonging to Sphingobacteriales of Bacteroidetes showed a higher relative abundance in the samples of A. conyzoides and E. annuus.
In contrast, the OTUs belonging to Acidobacteria Subgroup 4, Sphingobacteriales and Sphingomonadales were absent in V. japonica ( Figure 1 and Supporting Information Figure S1).
The analysis of variance (ANOVA) with p values (p < 0.05) was used to identify statistically significant differential OTUs among the rhizosphere soil samples from six plant species. The reads from the identified differential OTUs, accounting for 74.51% of the total rarefied reads, primarily belonged to Sphingobacteriales of  Table S6).

| Estimating the bacterial diversity and species richness of the rhizosphere microbiomes in six plant species
Bacterial diversities in the samples of each plant (alpha diversity) were evaluated using an OTU-based analysis method. Alpha diversities for all the samples are summarized in Supporting Information Table S4. The rarefaction curves showed that all the samples reached the saturation phase with a satisfactory level of confidence and a Good's coverage index of at least 94% (Supporting Information   Table S4; Supporting Information Figure S2). Significant differential OTU richness estimated by Chao1 (p = 0.022 < 0.05) and bacterial diversity estimated by the Shannon index (p < 0.0001) were observed among the rhizosphere microbiome of six plant species.
Among these, rhizobacteria of the plant A. argyi showed higher bacterial diversity (Shannon index: 10.069 ± 0.098) and OTU richness (Chao1: 4309.9 ± 144) compared with the rhizobacteria of the other five plant species. In contrast, E. hirta showed a lower bacterial diversity (Shannon index: 9.3542 ± 0.180) and OTU richness (Chao1: 3571.8 ± 120) than those of the other five plant species.
In addition, at the higher taxonomic levels, significant differential OTU richness estimated by Chao1 (p = 0.039 < 0.05) and bacterial diversity estimated by the Shannon index (p < 0.0001) were observed among the rhizosphere microbiome of the plant families of Euphorbiaceae, Asteraceae, and Violaceae. No significant differential OTU richness (Chao1: p = 0.229) and bacterial diversity (Shannon index: p = 0.144) were identified in the plant orders of Asterales and Malpighiales.
The rank-abundance curve visually depicts both species richness and evenness in the six plant species. Erigeron annuus, A. conyzoides, and A. argyi exhibited higher species richness and evenness. In contrast, the E. hirta, B. biternata, and V. japonica samples showed lower species richness and evenness, suggesting that the bacterial species compositions of E. annuus, A. conyzoides, and A. argyi were more abundant and better distributed (Supporting Information Figure S3).
Overall, the results demonstrated that plants grown in the same soil field had a significant effect on the bacterial diversity and species abundance of the microbial communities in the rhizosphere.

| D ISCUSS I ON
The rhizosphere microbiome, which is thought to have great importance to improve plant host health and productivity, has attracted more attention during the past several decades (Berg et al., 2005;Bulgarelli et al., 2013;Turner, James, et al., 2013). Previous studies have demonstrated the effect of plant species, cultivars, or genotypes on the rhizosphere microbiome (Bulgarelli et al., 2015(Bulgarelli et al., , 2012 ;  Figure S4). The results supported the hypothesis that the more phylogenetically distant the plant hosts, the more distinct their associated bacterial communities should be (Lambais, Lucheta, & Crowley, 2014;Pérez-Jaramillo, Mendes, & Raaijmakers, 2016). In this case, the bacterial diversity and OTU abundance of the rhizosphere microbiome were significantly related to the plant taxa, at least at the species and family levels.

ACK N OWLED G EM ENTS
This work was supported by grants from the National Natural Science

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

AUTH O R S CO NTR I B UTI O N
BT and SL designed the experiments. SL, XX, BT, ZC, JX, RM, XY, and YZ carried out the metagenomic analyses. SL, XX, ZC, LZ, and BZ carried out the biochemical analyses. BT, SL XX, and ZC wrote the manuscript. All authors read the final manuscript.

E TH I C S S TATEM ENT
None required.

DATA ACCE SS I B I LIT Y
The raw sequencing reads dataset was deposited at GenBank and the NCBI Short Read Archive under the project accession number PRJNA316593 and accession number SRR7012890, respectively.

R E FE R E N C E S
Aleklett, K., Leff, J. W., Fierer, N., & Hart, M. (2015). Wild plant species growing closely connected in a subalpine meadow host distinct