Effects of rhizoma peanut cultivars (Arachis glabrata Benth.) on the soil bacterial diversity and predicted function in nitrogen fixation

Abstract There is a growing awareness of the importance of soil microorganisms in agricultural management practices. Currently, much less is known about whether different crop cultivar has an effect on the taxonomic structure and diversity, and specific functions of soil bacterial communities. Here, we examined the changes of the diversity and composition and enzyme‐encoding nitrogenase genes in a long‐term field experiment with seven different rhizoma peanut cultivars in southeastern USA, coupling high‐throughput 16S rRNA gene sequencing and the sequence‐based function prediction with Tax4Fun. Of the 32 phyla detected (Proteobacteria class), 13 were dominant: Acidobacteria, Alphaproteobacteria, Actinobacteria, Betaproteobacteria, Bacteroidetes, Verrucomicrobia, Gammaproteobacteria, Deltaproteobacteria, Gemmatimonadetes, Firmicutes, Nitrospirae, Chloroflexi, and Planctomycetes (relative abundance >1%). We found no evidence that the diversity and composition of bacterial communities were significantly different among different cultivars, but the abundance of some dominant bacterial groups that have N‐fixation potentials (at broad or fine taxonomic level) and predicted abundances of some enzyme‐encoding nitrogenase genes showed significant across‐cultivar differences. The nitrogenase genes were notably abundant in Florigraze and Latitude soils while remarkably lower in Arbook and UF_TITO soils when compared with other cultivars, indicating different nitrogen fixation potentials among different cultivars. The findings also suggest that the abundance of certain bacterial taxa and the specific function bacteria perform in ecosystems can have an inherent association. Our study is helpful to understand how microbiological responses and feedback to different plant genotypes through the variation in structure and function of their communities in the rhizosphere.


| INTRODUC TI ON
Rhizosphere microorganisms in cropland systems are critical members of the plant microbiome regulating key aspects of plant growth, nutrient cycling, and soil health. Plant roots can promote or impede rhizosphere microorganisms' recruitment through the secretion of root exudates and volatiles (Berendsen, Pieterse, & Bakker, 2012;Broeckling, Broz, Bergelson, Manter, & Vivanco, 2008;Schmidt, Ulanova, Wick, Bode, & Garbeva, 2019). For example, recent researches have reported that rhizosphere microbial community assembly and microbial taxa colonization can be largely determined by root secondary metabolites of host plants (Veach et al., 2019), including the chemical composition and compounds quality (Gransee & Wittenmayer, 2000;Lesuffleur, Paynel, Bataille, Deunff, & Cliquet, 2007). Root exudates will thus affect the composition and diversity of rhizosphere microbial communities (Aira, Gomez-Brandon, Lazcano, Baath, & Dominguez, 2010;Badri & Vivanco, 2009), but such an effect may depend greatly upon the plant species and genotype (Rengel & Marschner, 2005). Different plant genotypes are likely to develop distinct microbial communities via the interactions between plant roots and microorganisms, which may also result in the changes of microbial-driven function in the soil (Philippot, Raaijmakers, Lemanceau, & Putten, 2013;Turner, James, & Poole, 2013). Leveraging on next generation amplicon sequencing protocols, we have highlighted the importance of various biotic, abiotic, and spatial factors in structuring microbial assemblage (Berg & Smalla, 2009;Wang et al., 2017). However, more direct evidence is needed to understand the involvement of genomic-basis factors in selecting for soil microbiomes. Particularly, the knowledge of cultivar effects (different plant genotypes) on soil microbial communities remains very limited. In agricultural ecosystems, growing evidence has emerged that the composition and diversity of the soil microbial communities differed among cultivars of crops, including wheat (Germida & Siciliano, 2001;Siciliano, Theoret, Freitas, Hucl, & Germida, 1998), potato (Manter, Delgado, Holm, & Stong, 2010;Weinert et al., 2011), sorghum (Schlemper, Veen, & Kuramae, 2018), and rice (Briones et al., 2002;Feng et al., 2015). To date, however, few studies were conducted to evaluate the effects of cultivars in legumes, such as rhizoma peanut (RP, Arachis glabrata Benth.) on the soil microbial communities.
Rhizoma peanut is a warm-season, perennial legume crop that is well-adapted to the southern United States (Dubeux et al., 2017;Ortega, Sollenberger, Quesenberry, Cornell, & Jones, 1992), which makes it a good option for pasture integration (Garay, Sollenberger, Staples, & Pedreira, 2004;Williams, Hammond, Kunkle, & Spreen, 1991). Rhizoma peanut has a high potential for increasing soil N supply and improving litter quality in C4 grass pastures (Cathey, Sinclair, & Mackowiak, 2013). It can be grown in monoculture for hay or in combination with perennial grasses for grazing systems (Castillo et al., 2015). Similar to many other legume species, RP largely relies on its rhizobium symbionts for biological nitrogen fixation (BNF), which is a primary source of nitrogen (N) in agricultural systems  Tg N annually, Herridge, Peoples, Boddey, 2008). A recent study has shown that the potentials of BNF in RP are genotype dependent, which is probably attributed to the differences in rhizosphere microbial communities (Dubeux et al., 2017).
Nitrogenase-mediated BNF process has been reported to benefit legume growth by the improvement of the availability of N that can be directly used by plants (Mus et al., 2016;Welbaum, Sturz, Dong, & Nowak, 2004). BNF, a key component of the nitrogen cycle, is particularly important in agricultural ecosystems because it helps reduce N fertilizer utilization and resulting environmental issues related to N fertilizer inputs such as ammonia volatilization and nitrate leaching (Ju et al., 2009). Bacteria are central performer in BNF process and reveal considerable biodiversity among diazotrophic organisms. It is estimated that 15,000 legume species form symbiotic associations with N-fixing rhizobia bacteria such as Rhizobia belonging to Alphaproteobacteria (Sprent, 2001). In addition to Rhizobia, N-fixation ability has been found in other bacterial phylogenetic groups, including green sulfur bacteria, Firmibacteria, Actinobacteria, Cyanobacteria, and other subdivisions of the Proteobacteria (Dixon & Kahn, 2004;Welbaum et al., 2004). Biological nitrogen fixation is an enzymatic reduction process, which is generally controlled by the nitrogenase enzyme that conventionally consists of the iron (Fe) protein and the molybdenum-iron (MoFe) protein (Zehr & Turner, 2001). The conventional nitrogenase is encoded by a core of nif genes including nifH, nifK, nifB, nifZ, nifE, nifN, nifZ, and nifX, which are all required for nitrogenase synthesis and catalysis (Dos Santos, Fang, Mason, Setubal, & Dixon, 2012;Mus et al., 2016). To our knowledge, little has been known for assessing the changes of nitrogenase genes in symbiotic bacteria associated with different RP cultivars.
The objective of our study is to identity the influences of different RP cultivars on the community structure and predicted function in N fixation of soil bacterial communities. We describe the diversity (16S rRNA genes) for bacteria using next generation amplicon sequencing. Tax4Fun coupled with metabolic references (Kyoto Encyclopedia of Genes and Genomes [KEGG]) was performed to link the phylogenetic and functional diversity of detected bacterial communities (Asshauer, Wemheuer, Daniel, & Meinicke, 2015).

| Experimental site and sampling strategy
The study is located at the North Florida Research and Education  (Dubeux et al., 2017). Planting material was obtained from the UF/IFAS, NFREC RP germplasm collection. The soil was disked and harrowed before planting. Treatments consisting of F I G U R E 1 Bacterial alpha diversity estimated by OTU richness, Shannon index, Inverse Simpson, and Chao 1 in soil cropped with seven rhizome peanut cultivars. Points represent the samples in each cultivar. The boxplot shows quartile values for each taxon colored by seven rhizome peanut cultivars. No significant differences were found among different cultivars (One-way ANOVA, p > .05) TA B L E 1 The number of reads and bacterial alpha diversity estimated by OTU richness, Shannon index, Inverse Simpson index, and Chao 1 (mean ± SE) using taxon resolution of 97% sequence similarity, respectively in seven different peanut cultivars in Marianna, Florida of USA  Sampling work was carried out in April 2017. To decrease the random effect of spatial variations in soil microbial communities, we selected three soil cores (3 cm diameter × 10 cm depth) in each plot (Table S1). Thus, a total of 84 samples were collected in the experimental site. Each soil sample was placed into a sterile plastic bag and put in the icebox immediately. All samples were then transported to the laboratory within 2 hr where they were sieved to 2 mm, thoroughly homogenized, and stored at 80°C for DNA extraction. An amplicon survey of the 16S ribosomal RNA (rRNA) genes (primers 515F/806R targeting the V4 region) was performed to provide a higher resolution and more in-depth analysis of the taxonomic composition and diversity of the bacterial communities (Caporaso et al., 2012). Amplicon libraries were prepared utilizing a frame-shift tagging system in the primer sequence (Lundberg, Yourstone, Mieczkowski, Jones, & Dangl, 2013) to generate high diversity libraries necessary for MiSeq sequencing. The frame-shift method has been reported to be able to eliminate the need for Phi X spiking before sequencing (Lundberg et al., 2013). Three PCR amplification steps were performed with a S1000 Thermal Cycler (Bio-Rad). Reagents and reaction conditions for three PCR amplification steps were shown in Table S2.

| Bioinformatics analysis of sequencing data
Raw sequences with an average quality score >30 were obtained in MiSeq sequencing machine in FASTQ format. The forward and reverse directions for each sample were generated into separated files. First, the files corresponding to the forward and reverse directions of each sample were merged into a single file and added tags for each sequence. The primers of the forward and reverse directions were then trimmed with up to one mismatch allowed and starting position ≤20. Secondly, forward and reverse reads of same sequence with at least 30 bp overlap and <0.25 mismatches were combined as single sequence by using FLASH v1.2.5 (Magoc & Salzberg, 2011). The sequences were then quality trimmed using Btrim (Kong, 2011)  F I G U R E 4 Relative abundance of the dominant groups at the phylum levels (>1% average relative abundance across all soils) of bacterial communities in soil cropped with seven rhizome peanut cultivars provides a high correlation between functional profiles predicted by 16S sequencing data and whole metagenome sequence data in the soil system (Asshauer et al., 2015;Kaiser et al., 2016). We here mainly focused on the differences of predicted abundances of nitrogenase genes across cultivars. In order to access the functional diversity of BNF across RP cultivars, 13 genes encoding nitrogenases were selected based on Tax4Fun analysis, including Nif-specific regulatory protein, nitrogenase (EC:1.18.6.1), NifB, nitrogenase molybdenum-iron protein alpha chain (EC:1.18.6.1), NifE, NifH, nitrogenase molybdenum-iron protein beta chain (EC:1.18.6.1), NifN, NifT, NifV, nitrogenase-stabilizing/protective protein, NifX, and NifZ.

| Statistical analysis
All statistical analyses were performed in R version 3.5.0 (R Core  (ANOSIM, 9,999 permutations;Clarke, 1993). Both NMDS and ANOSIM were performed with the vegan and ape R package (Oksanen et al., 2013). Hierarchical cluster analysis of the bacterial communities based on the Bray-Curtis dissimilarity matrix was conducted with the hclust function in package vegan, and classification boxes were added using the ordicluster function.
To investigate the differences of the bacterial taxonomic composition across different cultivars, we first chose 13 dominant phyla (class) based on the taxonomic abundance data (relative abundance >1%) to display the differences among cultivars at the phylum levels. We then chose 20 dominant genera based on the taxonomic abundance data (average abundance >10 across all soil samples) and conducted multivariate data analysis with FactoMineR R package. The methods implemented in the package are conceptually similar with classical multivariate data analysis like principal component analysis (PCA) or correspondence analysis, but the graphical outputs including variables factor map and individuals factor map can display clearer distribution of the dominant bacterial taxa across cultivars than using the traditional PCA method.

| Distribution of bacterial taxa and alpha diversity
Across all soil samples, we obtained a total of 446,107 high-qual- F I G U R E 6 Predicted abundances of enzyme-encoding genes involved in nitrogen fixation. The color code refers to gene abundance, with high predicted abundances (red) and low predicted abundances (gray) The estimated bacterial diversity (i.e., OTU richness, Shannon index, Inverse Simpsons index, and Chao 1) did not vary considerably across seven different cultivars (Figure 1). No statistically significant differences (ANOVA, p > .05) in alpha diversity were found for OTU richness, Shannon index, Inverse Simpson index, and Chao 1 under different cultivars (Table 1). Soils from Arblick and Florigraze plots had relatively high diversity while soils from Arbrook plots were relatively low.

| Bacterial beta diversity
We further investigated the variation in bacterial beta diversity across different RP cultivars. Community composition similarity across all soil samples was calculated using the Bray-Curtis abundance-based distance index, which was highly correlated with the incidence-based Jaccard index (Mantel test: R = 0.99 and p = .001).
The variation in community composition across cultivars is deline-

| The taxonomic abundance of bacterial dominant groups
The relative abundances of dominant bacterial phyla (>1% average relative abundance across all samples) showed some differences across seven cultivars (Figure 4). Significant among-cultivar differences were found in Alphaproteobacteria, Actinobacteria, Verrucomicrobia, and Nitrospirae (ANOVA, p < .05, Table 2). Relative abundance of Alphaproteobacteria in Florigraze soils was significantly higher than that in other RP cultivar soils (LSD, p < .05). At a finer taxonomic level, we also found significant differences among RP cultivars in some bacterial genera including Nitrospira, Pseudolabrys, Streptophyta, Bacillus, and Subdivision3_genera_incertae_sedis (ANOVA, p < .05, Table S3).

| Functional profiles of bacterial communities
All nigrogenases genes we examined have been shown functionally associated with bacteria-driven BNF (Dixon & Kahn, 2004).
Generally, the nitrogenase genes were more abundant in Florigraze and Latitude soils while relatively lower in Arbook and UF_TITO soils when compared with other cultivars (Figure 6 and Table 3). The relative abundances of genes for nitrogenase Nif-specific regulatory protein, nitrogene (EC:1.18.6.1), and nitrogen fixation protein NifT were significantly different among seven rhizoma peanut cultivars (ANOVA, p < .05). The relative abundances of genes for Nif-specific regulatory protein and nitrogenase (EC:1.18.6.1) were extremely high and low, respectively, when associated with cultivars compared to other genes examined. The predicted abundances of genes encoding nitrogenase iron protein (NifH), nitrogenase molybdenumcofactor synthesis protein (NifE), nitrogenase molybdenum-iron protein alpha chain (EC:1.18.6.1), and nitrogen fixation protein (NifB) were relatively lower in Arbrook soils compared with other genes.
The predicted abundances of genes encoding Nif-specific regulatory protein were significantly lower in Florigraze soils compared with the soils grown with Arblick, Arbrook, Ecotorf, and UF Peace (p < .05).

| D ISCUSS I ON
In this study, we investigated the changes of soil bacterial communities with seven different RP cultivars in an experimental field in Florida. We found no significant differences in the taxonomic diversity and composition of bacterial communities among seven RP cultivars (Figures 1, 2 (Marschner, Yang, Lieberei, & Crowley, 2001;Weinert et al., 2010). This is probably because bacterial diversity in soils with rhizoma peanut in our experiment may be primarily determined by the soil properties (Girvan, Bullimore, Pretty, Osborn, & Ball, 2003;M et al., 2001). Although different peanut cultivars were cropped in our field experiment, the soil type, pH, and the conditions of soil moisture and nutrient were similar in the fine-scale study region. Previous studies have demonstrated that at small or local scales (centimeters to meters), the diversity of soil microbial communities are often determined by variation in the soil physicochemical parameters (Wang et al., 2015;Wardle, 2002). An alternative explanation is that the samples we collected were the mix of rhizosphere and bulk soils. In general, only the root-associated bacterial communities are relatively sensitive to different cultivars, as host plants and their root exudates have been reported greatly affecting soil bacterial community structures in the previous studies (Marcel, Bardgett, & van Straalen, 2008;Stephan, Meyer, & Schmid, 2000). Furthermore, selection of primers used for amplification of bacterial rRNA gene will affect the outcome of studies examining bacterial community structure. In this study, we only concentrated on bacterial rRNA gene for the bacterial community analysis. However, the relative conservation of the 16S rDNA may not represent all the variation in bacterial community diversity such as certain species-level variation (Ranjard et al., 2001).
Although no significant differences in diversity and composition of bacterial communities were detected among RP cultivars, taxonomic abundance in some bacterial dominant groups, examined by different taxonomic levels showed significant among-cultivar differences (Figures 4, 5, Table 2; Table S3). This finding suggests that plant genotypes may not result in a different selection for the specific taxa in a community, but could cause the changes of dominance of certain taxa in that community. Some microorganisms indeed have a particular affinity for certain plant genotypes, which is in accordance with previous reports showing the effects of plant genotypes on microorganisms in the rhizosphere (Bressan et al., 2009;Meyer et al., 2010).
Interestingly, significant differences in abundances of dominant bac- The availability of N is frequently a major factor affecting crop productivity in agro-ecosystems. Therefore, understanding BNF potential bacteria is of tremendous importance to agricultural management practices (Vitousek et al., 2002). In our study, we investigated the differences of bacterial functioning in BNF among different RP cultivars, based on predicted abundances of enzyme-encoding nitrogenase genes using a new bioinformatic tool Tax4Fun. As expected, the predicted abundances of nitrogenase genes showed some differences across RP cultivars ( Figure 6, Table 3). This finding is consistent with our predictions, indicating the significant differences of abundance in some BNF-related taxa. This suggests that plant genome may regulate the metabolic programs that microorganisms use for BNF, or alternatively select the microorganisms that have different BNF efficiency (Berendsen et al., 2012). Higher abundance in most examined nitrogenase genes in Florigraze soils than other cultivars also indicates that Florigraze may have high potentials in BNF.
In conclusion, this study represents an attempt to assess the cultivar effect of RP on soil bacterial communities in a field experiment of southeastern USA. We provided direct evidence that different RP cultivars have no effects on the bacterial taxonomic diversity and composition, whereas the abundance of some dominant bacterial groups that have N-fixation potentials and the predicted abundances of some nitrogenase genes showed significant among-cultivar differences. Such a result suggests that bacterial specific function such as nitrogen fixation in ecosystems could be determined by the taxonomic abundance of some dominant groups in a community despite of the characteristics of their community composition and diversity.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Sequencing raw data has been deposited into the NCBI Sequence