Cross study reveals ubiquitous bacterial taxa across legume rhizospheres

1 Introduction

Plant-microbe interactions play key roles in plant growth and health. These interactions occur at many plant sites, both within plant tissues and on the surface. One of the most critical environments where these interactions occur is the rhizosphere, the narrow layer of soil near the root and influenced by it (D. 1). The rhizosphere is affected by root activities, including nutrient uptake, metabolite exudation, rhizodeposition, and mucilage biosynthesis. Thus, plants create an environment different from bulk soil that can attract specific microorganisms. Signaling between plants and microorganisms in the rhizosphere is a bidirectional phenomenon (2). In this way, plants actively select their rhizospheric microorganisms, while microorganisms influence plant-being. Recent advances in omics technologies, particularly in next-generation sequencing platforms, have improved our understanding of the microorganisms present in this niche (3, 4).

The primary source of rhizospheric microorganisms is the nearby bulk soil (5). Differences in species, developmental stage, nutritional or health status, and zone, among others, determine the nature and magnitude of root contributions to the rhizosphere (6). Typically, the rhizospheric microorganisms most extensively studied were those with obvious interactions with the plants, i.e., pathogens or symbionts. However, it has become clear that many seemingly neutral microorganisms interact with the plant in ways that constitute an active relationship (2).

The family Fabaceae is commonly known as legumes. Many species of this family represent important crops, crucial to food production worldwide. Most legumes establish a symbiosis with a diverse polyphyletic group of bacteria belonging to the classes Alphaproteobacteria and Betaproteobacteria, collectively known as rhizobia (7; 8). These bacteria inhabit root nodules, where they perform nitrogen fixation, thereby promoting plant growth. While most legumes can associate with a minimal number of rhizobial genera (9), some tend to promiscuity and can be nodulated by up to five different bacterial genera (10). In addition to rhizobia, various non-rhizobial bacteria have been found inhabiting legume nodules, including Bacillus spp., Pseudomonas spp., and Mycobacterium spp. (11).

The hypothesis of this work was that, despite differences in the community structures of the rhizospheres of different legume species, there is a collection of taxa present in all rhizospheres. Because rhizospheric microorganisms are selected by the environment generated in part by the plant, mainly from the surrounding soil, the possible existence of such a core should be considered biologically significant, as the presence of these taxa across numerous samples from different legume species and milieus is unlikely to be the consequence of stochastic effects.

In this paper, we obtained amplicon sequencing results from different legume rhizosphere samples from a public database. In total, eight legume genera were studied, all of them of agricultural importance. The set of studies included in this analysis comprised not only different legume species but also different experimental settings (greenhouse, open field, etc.), different 16S rRNA regions analyzed, and different countries of origin, which would inherently determine different parental soils. Thus, the objectives of our work were to analyze the composition and diversity of the rhizospheric communities of different legumes and to identify ubiquitous bacterial taxa. The approach we used identified ubiquitous bacterial genera, which should encourage new studies on the presence of these taxa across different plant rhizospheres and their ecological roles.

2 Materials and methods

2.1 Data acquisition

All sequences were acquired from publicly available data deposited at the Sequence Read Archive (SRA) of the National Center for Biotechnology Information (https://www.ncbi.nlm.nih.gov/sra) in August 2024. The search for the different BioProjects to include in our analysis was done using the terms “rhizosphere” and the Latin name of each of the plants intended to be included (“Arachis hypogaea”, “Desmodium”, “Glycine max”, “Lotus”, “Lupinus”, “Medicago”, “Mimosa”, “Trifolium”). We conducted a search for each plant, selecting “all fields” for both terms, and retained only results containing the term “amplicon” for further analysis.

The resulting hits from the search were analyzed in the “SRA Run Selector” tool, and its associated metadata was downloaded as.txt files. The corresponding raw sequences from BioProjects that met our criteria (clearly annotated as rhizosphere samples for each plant) were downloaded as.fastq files using the fasterq-dump (v.2.9.1) tool, with the –split3 option to recover forward and reverse reads in separate file. A total of 48 BioProjects, adding up to 1388 samples were finally kept for analysis. The sequencing platforms and region sequenced for each BioProject are described in Table 1.

Table 1. List of datasets included in the analysis.

^*Identified at least at the phylum level.

Each dataset comprised a single SRA BioProject for each legume analyzed. Code names for the datasets (Study_id) consisted of the initial letters of the plant Latin name, and a third identifier letter ordered alphabetically. n.a. = information not available.

2.2 Sequences processing and statistical analysis

All datasets (one for each BioProject) were processed separately. Amplicon sequence variant (ASV) tables were generated from the BioProject.fastq files using the DADA2 package version 1.26 (12) in RStudio 2026.01.0 Build 392. The main steps performed included: quality check, filter and trim (primers were removed in this step), error-rate learning, merging (for datasets with paired sequences only), dereplication of sequences, and chimera removal. The filterAndTrim parameters truncLen and trimLeft were adjusted for each dataset to ensure that most sequences were retained after processing; this depended on the primers’ presence, quality profile, and the resulting number of sequences; quality score threshold and the maximum expected error were set at 2.

Taxonomic assignment of the sequences was performed using the method IDTAXA in DECIPHER version 2.28.0 and the SILVA r138 SSU database (13). A phyloseq object was built for each dataset using the phyloseq package version 1.54.0 (14). Unassigned ASVs, as well as those annotated as family=mitochondria or order=chloroplast, were discarded. For every taxonomic level ranging from phylum to genus, a new phyloseq object was built, and all unassigned ASVs at the corresponding taxonomic level were removed. Samples with fewer than 3000 reads were removed from the analysis. Thus, five phyloseq objects were obtained for each dataset, one for each taxonomic level (phylum, class, order, family, and genus).

All datasets belonging to rhizosphere samples from the same legume genus were merged using the merge_phyloseq command of the phyloseq package. One set of five phyloseq objects per legume genus was thus created by merging all phyloseq objects at each taxonomic level. All phyloseqs were rarefied at this point. These phyloseq objects were used to calculate alpha- and beta-diversity parameters, as well as core microbiomes, for each legume genus and for all legumes included. The methodology employed is schematically described in Figure 1. The microbiome package version 1.32.0 (15) was used for this goal. Statistical significance of differences between datasets for alpha-diversity indices was assessed using ANOVA, and Tukey HSD test was used for post hoc analysis (p < 0.05); normality and homoscedasticity were checked with Shapiro-Wilk and Levene’s tests, respectively. Differences in rhizospheric community composition were visualized using non-metric multidimensional scaling (NMDS) with Unweighted UniFrac distances, and the significance of these differences was assessed using Permutational Multivariate Analysis of Variance (PERMANOVA). The Permutation test for homogeneity of multivariate dispersions was checked with the betadisper command of the microbiome package version 1.32.0.

The core microbiome of the entire dataset, as well as that of the different legumes, was calculated by considering only bacterial genera with relative abundances greater than 0.1% of the total community. When calculating the core microbiome at the family level, 0.5% was used as the abundance threshold.

3 Results

3.1 Basic parameters of the datasets obtained

We centered our analysis on eight economically relevant legume genera. After filtering the SRA query results, we obtained 19, 8, 6, 5, 5, 3, 1, and 1 datasets for Glycine max, Lotus spp., Arachis hypogaea, Medicago sativa, Trifolium spp., Lupinus spp., Mimosa pudica, and Desmodium canadiense, respectively (Table 1). Datasets from the genus Lotus included five datasets from L. japonicus, two from L. corniculatus, and one from L. tenuis. Within the genus Trifolium, we found three datasets belonging to T. repens and two to T. pratense. Within the genus Lupinus, we found two datasets belonging to L. montanus and one to L. angustifolius. The number of samples that each dataset comprised, as well as the lowest and highest number of reads on each dataset, are detailed in Table 1. Considering the minimum number of reads among samples, and the need to perform a rarefaction, we decided to exclude from the analysis those samples that retained less than 3,000 named sequences after taxonomy assignment. The largest dataset in our analysis was LJA (from L. japonicus), which initially contained 340 samples (304 passed all parameters), whereas the dataset from T. repens contained only three samples. The disparity in the number of samples per study, as well as differences in settings, rRNA region analyzed, and study site, may introduce bias into the results and should be accounted for when interpreting them.

3.2 Alpha diversity indices of the rhizospheric bacterial communities of different legume genera

Analysis of diversity indices can provide insight into the expected values of these parameters in legume rhizosphere communities. For instance, at the phylum level, the highest value of Chao1 (30.2) was observed for a Lotus corniculatus study (LCB, n=11) while the lowest Chao1 value obtained (3.3) was for a Trifolium repens study (TRB, n=3) (Supplementary Figure 1). At the genus level, which was the lowest taxonomic rank analyzed, the highest Chao1 value observed (223.8) was for the same L. corniculatus study that had the highest phylum Chao1 (LCB, n=11). The lowest Chao1 value observed was an average of 9.73 for a L. japonicus study (LJA, n=270) (Supplementary Figure 2). In this way, these results suggest that the number of observed taxa in legume rhizospheres varies across taxonomic ranks, with roughly 3–30 phyla and 10–200 genera.

When analyzing the Chao1 and Shannon indices across taxonomic levels for the included legume genera, some genera exhibit higher rhizosphere diversity than others. At high taxonomic ranks, Desmodium plants showed significantly higher Chao1 index (Figure 2), while Mimosa, Arachis, and Lupinus samples have significantly higher Shannon index values (Figure 3). On the other hand, Lotus samples exhibited significantly lower Chao1 and Shannon indices than those of the other rhizospheres. These differences tend to narrow and even disappear at lower taxonomic levels. For instance, at the genus level, only Lotus samples exhibit significantly lower Chao1 and Shannon indices.

3.3 Bacterial community structure in the rhizospheres across studies

When visualizing all data available through NMDS, most samples cluster tightly along the left side of the two-dimensional space. Despite this clustering, PERMANOVA results show that rhizospheric community composition between legumes genus are significantly different from each other (Supplementary Table 1). Only some samples are located outside of this cluster. Notably, Lotus japonicus samples from the LJA study and some from the LJD study form a cluster to the right of the NMDS space (Figure 4; Supplementary Figure 3); this separation is also evident when analyzing the Lotus samples separately (Supplementary Figure 4). These two studies (LJA and LJD) were performed by sequencing the V5-V7 region of the 16S rRNA gene, in contrast to the more commonly used V3-V4. Even though this could have included a bias in the identification of the members of the communities, the fact that most samples of the LJD study do not cluster with LJA, and that there were other studies, outside this cluster, that also used V5-V7, indicates that the region used for amplification is not the sole cause of this result.

Other samples that do not form a cluster, but lie outside of the two major groups, include TRB (all three samples), AHE (12 out of 13 samples), GMH (2 out of 6 samples), GMM (1 out of 7 samples), and one sample from LCA (out of 12 samples). Only TRB and AHE contain all or most of their samples in this group. These samples are from the rhizospheres of T. repens and A. hypogaea, respectively; no clear rationale is apparent for their placement outside the main cluster or for their separation from other samples of their respective legume species. Both are from samples collected from field settings, and the rRNA 16S region analyzed was V4 and V3-V4, respectively. The only distinguishing feature of these studies relative to the others in the same legume is the country of origin. Of the 6 A. hypogaea studies analyzed, AHE is the only one from India, while the remaining are all from China. At the same time, TRB is from the USA, while the other two studies of T. repens are from China. Because all these studies analyzed plants collected from field settings, it is possible that the geographic region in which they were conducted played an important role in shaping rhizospheric communities.

3.4 Core microbiome of the studied legume rhizospheres

There were 18 prokaryotic genera with a relative abundance greater than or equal to 0.1% at least in half the samples (Table 2). Among these, Mycobacterium was the most ubiquitous, present in 86% of samples, followed by Pseudomonas, Devosia, and Streptomyces, each present in 79%, 76%, and 70% of samples, respectively. Not only was Mycobacterium particularly ubiquitous in all samples, but it was also highly present in the rhizosphere of all the legume genera analyzed. In Medicago sativa, Mycobacterium was present in 69% of samples (n=45), representing the lowest frequency observed across all analyzed legume rhizospheres. Pseudomonas, on the other hand, was particularly present in all samples. However, when analyzing its presence by legume genera, it can be observed that it was absent (threshold 0.1% of relative abundance) from more than half the samples of Trifolium (present in 45% of the samples) and Arachis (present in 33% of the samples).

Table 2. List of the most ubiquitous and abundant bacterial genera in the different legume rhizospheres. Numbers indicate the fraction of samples in each plant in which that genus is present with a relative abundance greater than 0.1%.

To avoid missing microorganisms that may be ubiquitous but present at low abundance, we also examined prokaryotic genera with relative abundance >0.01%. In that case, the number of genera reached 28, but the top 6 most ubiquitous genera remained unchanged compared with the 0.1% threshold (Figure 5).

4 Discussion

Regarding the alpha-diversity results, a difference is observed between low and high taxonomic ranks. These differences may be relevant, as diversity at higher taxonomic ranks is often more meaningful than at lower taxonomic levels for functional and niche diversity (16).

The appreciations about alpha-diversity should be interpreted mainly as descriptive. Based on the data analyzed, no conclusion can be drawn regarding the influence of different legume species on the diversity of their rhizosphere communities. On the one hand, we gathered data from experiments designed for different purposes, conducted in the field or in pots, and at distant locations. On the other hand, the different number of studies and samples for each legume species makes it particularly difficult to perform valid comparisons regarding α-diversity parameters. For instance, Desmodium comprised a single dataset (study) with 11 samples, while eight studies contained Lotus spp. rhizosphere samples. When analyzing the Lotus samples alone, we found significant differences among them. For instance, at the genus level, the highest average value obtained for Chao1 was 223.8 (study_id=LCB, n=11), but the lowest was 9.73 (study_id=LJA, n=270) (Supplementary Figure 1B); the average Chao1 of Lotus samples from all studies was 43.8 (n=426) (Figure 1E). On the other hand, the Desmodium canadiense study had a Chao1 average of 111.2 (n=2) (Figure 1E). Thus, the high diversity values observed for Desmodium may be biased by the limited data available. Nevertheless, our results shed light on the expected levels of richness and diversity across different rhizospheres. In the future, as global databases expand, the greater volume of data will enable more robust comparisons between species rhizospheres.

The persistence of certain taxa above the chosen abundance threshold is meaningful. Not only are these genera present in most samples, regardless of the plant species analyzed, but they are also present across different settings and locations, and when rhizospheres were analyzed by sequencing different 16S rRNA regions. To identify the core taxa of the rhizospheres, we used the abundance-occurrence approach (17) to avoid the inclusion of spurious data. One criticism that can arise from this is that low-abundant but ecologically meaningful taxa are overlooked. However, the validation of the presence of sequences with a low number of reads is at present troublesome, and we think a cautious approach is the best fit. The top four most ubiquitous bacterial genera in this study are highly prevalent in all legume genera, even in those with few samples, such as Desmodium and Mimosa.

The four most ubiquitous genera in the samples analyzed were Mycobacterium, Pseudomonas, Devosia, and Streptomyces. The relationship between Mycobacterium and plants has long been recognized and remains an active area of study. Members of the genus Mycobacterium are commonly found in the rhizosphere of many plants, such as tomato (18), barley (19), or medicinal plants (20). Mycobacterium spp. have also been found inside the root nodules of M. sativa, alongside Sinorhizobium and eight other genera (21). In addition to being identified by DNA sequencing, isolates have been obtained from nodules of indigenous legumes (22).

Pseudomonas is a well-known soil inhabitant and has long been studied as a rhizosphere inhabitant (23). A higher abundance of Pseudomonas in the rhizosphere compared to bulk soil is well recognized (24). Therefore, the inclusion of this genus as one of the top core genera should not be surprising. Recently, it has served as a model organism for studying the genetic determinants of rhizosphere competence (25). As noted for Mycobacterium, Pseudomonas has also been found inhabiting legume root nodules (26).

As well as Mycobacterium, Devosia was present in samples from every legume species analyzed, with a prevalence of more than half the samples of each legume. Some members of the genus Devosia have been primarily isolated from the rhizosphere of plants such as rice (27) or legumes (28). Devosia has also been found in the rhizosphere of maize (29) or millet (30). Not only is it a common inhabitant of many legume nodules (31, 32), but some members of the genus are known to have nitrogen fixation ability, for instance, D. neptuniae (28, 33, 34). Devosia belongs to the order Rhizobiales, and many aspects of its lifestyle are congruent with those of rhizobia. Various aspects of the association between Devosia and legumes warrant further investigation.

On the other hand, Streptomyces is a common soil inhabitant (35), and a competent rhizosphere colonizer in many plants, where its ability to degrade diverse carbon compounds, plays a key role (36). Legume rhizospheres have served as the source of many Streptomyces isolations in the search for plant growth promoters. Such is the case for M. sativa (37), soybean (38), or L. tenuis (39).

As we analyzed a set of legume species that are typically nodulated by different rhizobium symbionts, it is not surprising that no rhizobium genus is present among the top core genera. Even the family Rhizobiaceae is present in only 71% of samples (Supplementary Table 3), primarily due to its low presence in G. max samples, which is expected, as the G. max symbiont belongs to the family Nitrobacteraceae.

It is noteworthy that the analyzed datasets, derived from DNA amplicon sequencing results, provide important information on the presence of these taxa in the rhizosphere of these legumes. However, to gain a deeper understanding of the functionality of these communities, other analyses, such as shotgun metagenomics, should be considered (40).

The four genera mentioned above were the most ubiquitous in the rhizosphere of the legumes analyzed. They share certain aspects that may explain their capacity for colonization and also exhibit traits that can benefit plant growth, suggesting that the presence of these microorganisms in legume rhizospheres is not a consequence of random events. Given the plant’s role in creating an environment that ultimately selects for specific microorganisms in its rhizosphere, it is notable that across eight legume species, certain bacterial genera are consistently present. The existence of a set of bacterial genera ubiquitous in the legume rhizospheres is rather meaningful, considering that the rhizospheric communities in these legumes are significantly different in terms of composition.

These results demonstrate that, despite the highly diverse conditions of the systems explored, a significant collection of bacterial taxa is ubiquitously found in legume rhizospheres. A different approach would be needed to elucidate which plant-microbe and microbe-microbe interactions are occurring in these systems. The identification of ubiquitous taxa in the rhizosphere of legumes opens the possibility of gaining a deeper understanding of the interaction between plants and microbes. Consistent selection of soil microbiota would indicate conservation in the recruitment strategies of these species, giving rise to further, more specific hypotheses about plant-microbe interactions. For instance, by studying and analyzing isolates from these taxa, physiological and genomic traits can be assessed to reveal key aspects of rhizospheric colonization. Future cross-studies can analyze other plant rhizospheres to determine whether the ubiquity of the taxa found in our study is specific to legume rhizospheres. Besides, as public databases expand, the approach presented in this paper can be applied to search for ubiquitous fungi in rhizospheres or to analyze functional genes to identify core microorganisms belonging to relevant bacterial guilds, such as diazotrophs or nitrifiers.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fsoil.2026.1861827/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Simone Raposo Cotta, University of São Paulo, Brazil

Reviewed by: Kusum Raj Tamang, University of Vermont, United States

Armando Cavalcante Franco Dias, University of São Paulo, Brazil