Soil microbial composition and abundance influence the growth of Lotus japonicus
Chiharu Ota, Masaru Bamba, Shusei Sato, Takashi Tsuchimatsu
TL;DR
The growth of Lotus japonicus is influenced by the abundance of specific soil bacteria, particularly Mesorhizobium, which helps the plant through nitrogen fixation.
Contribution
This study identifies Mesorhizobium as the key bacterial group influencing Lotus japonicus growth in natural soil environments.
Findings
Plant growth varied significantly based on soil microbiota composition.
Mesorhizobium abundance was most strongly linked to improved plant growth.
Local presence of Lotus japonicus correlates with higher Mesorhizobium abundance.
Abstract
In mutualistic symbiosis between plants and bacteria, the abundance and composition of symbiotic bacterial groups in the soil microbiota can be important for plant growth. Here, we focused on the nitrogen-fixing mutualism between Lotus japonicus and nodule bacteria to investigate whether and how much the abundance of symbiotic rhizobia in the soil microbiota of natural environments contributes to variations in host plant growth. An inoculation experiment of soil microbiota revealed extensive variations in plant growth phenotypes, even between microhabitats. We found that the local presence of L. japonicus and the relative abundance of Mesorhizobium bacteria showed positive correlations with plant growth supported by both 16S amplicon sequencing and shotgun metagenome analyses. Among bacteria investigated, the abundance of Mesorhizobium was most strongly associated with plant growth…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —The University of Tokyo
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsLegume Nitrogen Fixing Symbiosis · Plant-Microbe Interactions and Immunity · Mycorrhizal Fungi and Plant Interactions
Introduction
Among various environmental factors affecting plant growth, soil properties, consisting of abiotic factors (e.g., pH, salinity, and heavy metal concentrations) and biotic factors (e.g., microfauna, bacteria, and fungi), are particularly important (Munns and Tester 2008; Neina 2019; Sharma et al. 2022; Trivedi et al. 2020). Since plants inhabit diverse soil environments and have evolved a wide range of traits to adapt to their habitats, they exhibit many differences in terms of which soil properties they are resilient to and which soil properties are critical in determining their growth and survival (Orsini et al. 2010; Schwachtje et al. 2012). Previous research has studied plant species or populations that inhabit harsh environments with extreme abiotic soil factors and revealed the extent of their resilience to the environments and the mechanisms involved (e.g., Busoms et al. 2021, 2024; Konečná et al. 2022; Ma et al. 2022; Selby and Willis 2018). Regarding the biotic factors, numerous fungi or bacteria are known to be important in affecting the growth of plant species that they interact with either positively (e.g., mycorrhizal fungi in most terrestrial plants) or negatively (e.g., plant pathogens). Specifically, in cases of highly specific interactions such as the mutualistic symbiosis, the abundance and composition of symbiotic bacterial groups in the soil microbiota can be highly important factors in determining plant growth (Rock-Blake et al. 2017; Vannette and Hunter 2013).
Moreover, because soil microbiota can influence plant growth, there is potential for an association between soil microbiota and the local distribution of plants (Cordovez et al. 2019). Plant–soil feedback may create a difference in soil microbiota depending on the local distribution and growth of plants (Bever et al. 1997, 2010; Thies et al. 1995). The Janzen–Connell hypothesis posits that conspecific pathogens accumulate near parent plants, reducing the survival of nearby seedlings (Connell 1971; Janzen 1970), and previous research has supported the existence of such a negative feedback for several plant species (Idbella et al. 2024; Luo et al. 2019). On the other hand, positive plant-soil feedback has also been debated. One hypothesis underlying the mechanism of positive plant-soil feedback is that litter decomposes faster in native soil than foreign soil because of enriched specialized decomposers in native soil, indirectly promoting the subsequently growing indigenous plants (De Long et al. 2023; Home-field advantage hypothesis). Another hypothesis emphasizes the importance of direct interactions of plants and soil microbes, especially mutualists in case of positive plant-soil feedback (Bever et al. 1997, 2010).
The legume–rhizobia interaction is a representative example of mutualistic symbiosis, wherein rhizobia induce nodule formation on plant roots to reside in, fix atmospheric nitrogen within the nodules, and supply part of the fixed nitrogen to the host plant (Oldroyd and Downie 2008). In return, the plant provides photosynthates to support bacterial growth within the nodules. Due to host specificity in this interaction, only a limited group of rhizobial strains can establish symbiosis with a given legume species (e.g., Lotus japonicus–Mesorhizobium: Handberg and Stougaard 1992; Glycine max–Bradyrhizobium: Xu et al. 1995; Medicago truncatula–Ensifer: Badri et al. 2007). In legumes—particularly species that form determinate nodules and do not terminally differentiate rhizobia within nodules—it is known that rhizobia can be released back into the soil when nodules senesce (Denison 2000). Therefore, positive plant–soil feedback may be established in which effective symbionts proliferate within nodules and are subsequently enriched in the local soil after release. Furthermore, legumes are known to preferentially select high-quality rhizobial strains and form more or larger nodules (Gubry-Rangin et al. 2010; Heath and Tiffin 2009; Simms et al. 2006), while penalizing “cheater” strains that fail to fix nitrogen, for example, by accelerating nodule senescence (Kiers et al. 2003; Regus et al. 2017). These mechanisms may lead to the selective enrichment of high-quality symbionts in the soil surrounding a legume species. Although many previous studies have investigated the specificity of the interactions between various legumes and their symbionts (Andrews and Andrews 2017), the relationship between the abundance of symbiotic rhizobia in the soil microbiota of natural environments, the local distribution of legume species and variations in host plant growth remains largely unexplored (Vuolo et al. 2022).
In this study, we focused on the mutualism between Lotus japonicus (Regel) K. Larsen and its symbiotic rhizobia. L. japonicus is a model legume plant for investigating the molecular or genetic basis of legume–rhizobia symbiosis (Handberg and Stougaard 1992; Soyano et al. 2013, 2024; Szczyglowski et al. 1998). Additionally, since L. japonicus is a wild legume species that is widely distributed in East Asia, it has been an ideal material for ecological research, including for investigations of climate adaptation (Bamba et al. 2019, 2020; Mustamin et al. 2023; Shah et al. 2020). Although Bamba et al. (2020) revealed massive partner quality variation among rhizobial bacteria associated with L. japonicus, how the soil bacterial abundance and composition influence the growth of L. japonicus remains unknown*.*
To investigate the effect of soil bacterial composition on L. japonicus in its natural habitats, particularly with respect to symbiotic rhizobia, we aimed to: (1) identify which bacterial taxa within the soil microbiota influence the growth of L. japonicus, and (2) determine whether and how the distribution of L. japonicus is related to differences in plant growth and soil microbial communities.
Materials and methods
Soil sampling in L. japonicus’ natural habitats
We sampled soil materials from a depth of approximately 10 cm at three sites around L. japonicus habitats in the Misaki Marine Biological Station, School of Science, the University of Tokyo (Kanagawa Pref. Japan, 35°16’N, 139°61’E) in 2023 (Fig. 1a). At each site, we collected two soil samples each of two different conditions according to the local presence of L. japonicus plants: near L. japonicus (< 30 cm) and distant from L. japonicus (> 1 m; Fig. 1b). Among the soil sampling sites used in this study, site A exhibited richer vegetation than the other two sites, with Solidago altissima and Xanthium strumarium subsp. sibiricum observed exclusively at this site and absent from site B and C. Soil microbiota were extracted within a day of sampling, and the remaining samples were preserved at –80 ℃ until DNA extraction. The outline of the analyses is illustrated in Fig. 1c.Fig. 1. Sampling and analyses of this study. a Three sampling sites around L. japonicus habitats in Misaki Marine Biological Station (Kanagawa Prefecture, Japan). b Representative picture of collection sites. Scale bar: 2 m. c Overview of the analysis and experiments
DNA extraction from soil samples and sequencing
We extracted DNA from the soil samples using the NucleoSpin® Soil system (Macherey–Nagel), according to the manufacturer’s instructions. We used Lysis Buffer SL2 for soil samples S4, 5, 6, and 7, and Lysis Buffer SL1 for the other samples. We performed 16S rRNA amplicon sequencing analyses and shotgun metagenome sequencing analyses using the Illumina MiSeq (2 × 300 bp) and Illumina Novaseq 6000 systems (2 × 150 bp), respectively. For the 16S rRNA amplicon sequencing analysis, the V3–V4 hypervariable regions were amplified using 341F and 805R primers (Table S1). The libraries were prepared with the two-step tailed PCR method. For shotgun metagenome sequencing analysis, the libraries were prepared using the NEBNext® Ultra™ II DNA Library Prep Kit (New England Biolabs, Inc.). The library preparation for Illumina MiSeq and Illumina Novaseq 6000, as well as the sequencing run, was performed by the Bioengineering Lab. Co., Ltd. (Kanagawa, Japan) and Rhelixa Co., Ltd. (Tokyo, Japan), respectively. The sequence data underlying this article are available in the Sequence Read Archive database (BioProject ID: PRJNA1302742).
Extraction of soil microbiota and inoculation experiments
We performed the extraction of soil microbiota and inoculation experiments following Bamba et al. (2024) with slight modifications. A total of 100 g of each soil sample and sterilized phosphate-buffered saline (PBS) were homogenized in a sterilized blender for 3 min and centrifuged at 1,000 g for 10 min at 10 ℃, and the supernatant was collected. We then added PBS to the precipitated soil and repeated the same procedure two more times. The supernatant was vacuum-filtered using Advantec 5A filter paper (particle size > 7 μm; ash content < 0.01%), and the filtered liquid was centrifuged at 8000 g for 20 min at 10 ℃. Fungi and protists were removed in the filtration and precipitation processes. The precipitated bacterial communities were resolved in 100 mL of PBS and stored at 4 ℃ until soil microbiota inoculation.
Partly scrubbed L. japonicus MG20 seeds were surface sterilized by immersing them in 2% sodium hypochlorite for 3 min and rinsing five times with sterile distilled water. After overnight imbibition, the swollen seeds were sown onto 1% agar plates, incubated in the dark for 3 days at 20 ℃, and grown at 20 ℃ under 16/8 light/dark conditions for 48 h. The rooting plants were transplanted into Leonard jars with lids (Leonard 1943) filled with sterilized vermiculite with 250 mL of sterilized B&D medium with 0.2 mM KNO_3_ (Broughton and Dilworth 1971). A total of 10 mL of soil-extracted microbiota resolved in PBS was precipitated using centrifugation at 8000 g for 3 min and washed with 10 mL of sterilized B&D medium. We inoculated 2.5 mL of the inoculant into Leonard jars and grew the plants at 20 ℃ under the same lighting conditions for 28 days. The Leonard jars were closed with lids to prevent cross-contamination between pots. We used two pots for each soil-extracted microbiota inoculation. The positions of pots in the growth chamber were determined randomly. The number of seedlings per pot was five.
We then harvested the whole plant bodies, imaged all individuals with a scanner, and separated them into shoots and roots. The shoots were dried for 48 h at 60 ℃, and then the dry weights were measured. For root phenotypes, we measured the numbers and projected areas of nodules from the scanned data. When we grew L. japonicus without inoculation, the plants did not form any nodules (data not shown).
Estimation of bacterial abundance in soil and β diversity with 16S rRNA amplicon sequencing analysis
The pre-processing of Illumina sequencing data was performed by Bioengineering Lab. Co., Ltd (Kanagawa, Japan). The obtained raw reads were first demultiplexed using the fastx_barcode_splitter tool in FASTX-Toolkit (ver. 0.0.14, Hannon 2010), and only reads with an exact match to the primer sequences were retained. Subsequently, the primer sequences were removed using fastx_trimmer, and sequences with a Phred quality score below 20 were filtered out using Sickle (ver. 1.33, Joshi and Fass 2011). Paired-end reads shorter than 130 bp were discarded along with their corresponding mate pairs. The remaining high-quality reads were then merged using FLASH (ver. 1.2.11, Magoč and Salzberg 2011) with a merged read length of 410 bp, a read length of 280 bp, and a minimum overlap of 10 bp.
To remove chimeric and noisy sequences, the DADA2 plugin in Qiime2 (ver. 2023.7, Bolyen et al. 2019) was employed, which generated representative sequences and an amplicon sequence variant (ASV) table. Taxonomic classification of the representative sequences was performed using the feature-classifier plugin in Qiime2 by comparing them to the Greengenes database (ver. 13_8, DeSantis et al. 2006) at 97% operational taxonomic unit (OTU) similarity. The proportion of each bacterial genus in each soil sample was then estimated as the number of reads classified as the genus divided by the total number of reads in the dataset. Weighted and unweighted unifrac distances were calculated using the diversity plugin in Qiime2.
Estimation of bacterial abundance in soil with shotgun metagenome analysis
The raw reads were first processed by removing adapter sequences using Cutadapt (ver. 4.7, Martin 2011). Taxonomic classification was then performed using Kraken2 (ver. 2.0.8-beta, Wood et al. 2019) with the Standard database (as of June 5, 2024, from https://benlangmead.github.io/aws-indexes/k2). The proportion of reads assigned to each genus per total number of reads was calculated to estimate the relative abundance of each genus.
To estimate the abundance of nodulation related genes (nodA, nodB and nodC) and a nitrogen fixation gene (nifH) of L. japonicus-associated Mesorhizobium, we performed a mapping of the short reads that were pre-processed and classified as Mesorhizobium with Kraken2. We also investigated the abundance of three single-copy housekeeping genes (atpD, dnaK and recA) as references. We retrieved the genome of Mesorhizobium sp. 113-1-2 from GenBank as reference genome, previously reported as L. japonicus-associated by Bamba et al. (2020). Reads were mapped to this reference using bwa-mem2 (ver. 2.2.1, Vasimuddin et al. 2019) with default parameters. The alignment results were subsequently sorted and indexed using Samtools (ver. 1.9, Danecek et al. 2021). We then calculated the average depth of mapped reads for each gene relative to the total number of sequenced reads based on the alignment results.
Statistical analysis of soil microbiota inoculation experiments
To analyze the effects of local presence of L. japonicus, soil sampling sites and their interactions on the plant growth in the soil microbiota inoculation experiments, we used a generalized linear model (GLM). In the GLM, each measured plant phenotype was the response variable, and the local presence of L. japonicus (i.e., near or distant), soil sampling sites (i.e., site A, B, or C), and their interactions were the explanatory variables. For the total number of nodules, we chose the quasipoisson distribution as an error distribution and log link function. For the total size of the nodule, we chose the Tweedie family as an error distribution with the var.power parameter estimated using the cpglm function and a log link function. For the shoot dry weight, we chose the gamma distribution as an error distribution and log link function. We used the glm function from stats (ver. 4.2.2) for all phenotypes and the cpglm and Tweedie functions from cplm (ver. 0.7.11, Zhang 2013) and statmod (ver. 1.5.0, Dunn and Smyth 2018; Hasan and Dunn 2012; Joergensen 1987; Smyth 1996, 1999; Tweedie 1984) of R packages in R Core Team (2024), respectively.
To further analyze the bacterial determinants of host plant growth in the soil microbiota inoculation experiments, we analyzed the results using generalized linear mixed model (GLMM) and added the abundance of each bacterial genus in soil calculated from 16S rRNA amplicon sequencing analysis or shotgun metagenome analysis as a fixed effect instead of the soil sampling site. We used soil samples used for inoculation as the random effect. For the total number of nodules, we chose the zero-inflated Poisson model and a log link function. For the total size of the nodule, we chose the zero-adjusted gamma distribution as an error distribution, and used a log link function. For the shoot dry weight, we chose the gamma distribution as an error distribution and log link function. Statistical significance was evaluated using Wald tests. As for the bacterial abundance estimated with 16S rRNA amplicon sequencing analysis, only the genera detected in more than two soil samples were tested to see if their abundance correlated with plant growth. As for the bacterial abundance estimated with shotgun metagenomic analysis, we only used genera with more than 2000 reads detected from more than two samples. P-values to observe the statistical significance of the abundance of each bacterial genus were adjusted using the Benjamini–Hochberg (BH) method (Benjamini and Hochberg 1995). We used the glmmTMB function of the R package glmmTMB (ver. 1.1.8; Brooks et al. 2017).
Results
Microbial communities of soil samples estimated with 16S rRNA amplicon sequencing analysis and shotgun metagenomic analysis
In the 16S rRNA amplicon sequencing analysis, we obtained an average of 37,519 merged high-quality reads per sample, with an average of 2471 ASVs per sample and 20,581 ASVs across all samples (Fig. S1; Tables S2–S4). We were able to classify 3060 ASVs (14.87% of total ASVs) at the genus level across all samples, with 275 genera.
As for the shotgun metagenomic analysis, Kraken2 processed an average of 29,370,776 individual reads, including both forward and reverse reads from paired-end data, treating each read independently for classification (Tables S2–S4). Of these, an average of 16,689,420 reads (56.82%) were classified. Within the classified reads, an average of 12,276,200 reads (43.33%) were classified to the genus level with an average of 3275 genera detected per sample (Tables S2–S4).
We calculated unweighted and weighted unifrac distances, which indicate qualitative and quantitative β diversity metrics, respectively, using the data from the 16S rRNA amplicon sequencing analysis (Fig. 2). Principal coordinate analysis of the two metrics indicated that the overall similarity of soil microbiota was explained largely by sampling sites rather than the local presence of L. japonicus.Fig. 2. Principal coordinate analysis of unweighted (a) and weighted (b) unifrac distance of soil microbial communities estimated with 16S rRNA amplicon sequencing analysis. The color of the dots indicates the local presence of L. japonicus in sampling sites (cyan: near L. japonicus, pink: distant from L. japonicus), and the shape indicates the soil sampling sites (circle:** A**, triangle:** B**, square:** C**)
The difference in soil-extracted microbial communities in the effect on L. japonicus growth
To quantify the difference in soil-extracted microbiota in the effect on L. japonicus growth, we measured the phenotypic traits of L. japonicus seedlings (i.e., shoot dry weight, total number of nodules, and total size of nodules) inoculated with different soil-extracted microbial communities 4 weeks after inoculation (Fig. 3). We observed significant variation among the different inoculants in the L. japonicus growth for all three traits investigated (Steel–Dwass test, p < 0.05, Fig. S2). The number of nodules per plant ranged from zero (S8, S9, S10, and S11) to 13 (S2).Fig. 3. The difference in the effect of soil-extracted microbial communities on plant growth. Total number of nodules (a), total size of nodules (b), and shoot dry weight (c) of L. japonicus seedlings inoculated with soil-extracted microbiota of different conditions. The color of the boxplots indicates the local presence of L. japonicus in sampling sites (cyan: near L. japonicus, pink: distant from L. japonicus), and the color of the dots indicates the soil sampling sites (red:** A**, green:** B**, blue:** C**). Horizontal lines indicate median, boxes include second and third quartiles, and whiskers extend to points that lie within 1.5 times the interquartile range
We found that soil sampling sites, the local presence of L. japonicus, and their interaction had significant effects on all the measured plant phenotypes (GLM, p < 0.05, Table 1). Soil samples from different sites varied significantly in their effects on plant growth; the effect of soil-extracted microbiota from site C was generally smaller regardless of the local presence/absence of L. japonicus (Fig. S2). Since the microbial communities of soil samples were similar within the same sampling sites (Fig. 2), the influence of soil sampling sites on plant growth may be attributable to overall differences in microbial community composition. In addition to the effect of sampling sites, soil samples with the local presence of L. japonicus showed significant positive effects for plant growth. The effect size compared with that of the sampling site differed among the phenotypes; it was relatively high for shoot dry weight and low for the total number of nodules and the total size of nodules. The effect of the local presence of L. japonicus on plant phenotypes was particularly prominent in sites A and B rather than in C (Fig. S2).
Table 1. Generalized linear model for growth and nodulation of plants in soil-extracted microbiota inoculation experiments.* 0.01 < p ≤ 0.05, ** 0.001 < p ≤ 0.01, *** p ≤ 0.001.
Influence of bacterial relative abundance on plant growth
We hypothesized that the relative abundance of specific bacterial genera explains such a striking variation in the L. japonicus growth among locally collected soil samples. To address this possibility, we obtained bacterial relative abundance data at the genus level using 16S amplicon sequencing analysis and shotgun metagenome analysis. We investigated whether the abundance of any bacterial genera correlated with variations in plant growth phenotypes. From the 16S amplicon sequencing analysis, 166 genera detected in more than two samples were used. From the shotgun metagenome analysis, 738 genera detected in more than two samples with more than 2000 reads were used.
While several bacterial genera showed significant correlations with the growth indexes (Table S5), the relative abundance of Mesorhizobium bacteria exhibited the strongest correlation. In the GLMM analysis based on the 16S rRNA amplicon sequencing analysis, Mesorhizobium bacteria was the only genus among the tested genera whose relative abundance exhibited a significant correlation with shoot dry weight (p = 0.0391, adjusted with the BH method, Fig. 4a). While no bacterial genera were significantly correlated with the total number of nodules, we detected 21 genera including Mesorhizobium whose abundance was significantly correlated with the total size of nodules (Table S5). We found a similar trend when using the results of the shotgun metagenome analysis. Although no genera were significantly correlated with the total number of nodules and shoot dry weight, we detected 42 genera that were significantly correlated with the total size of nodules, with only Mesorhizobium (Fig. 4b) and Kribbella detected by both methods (Table S5). These results indicate that the relative abundance of Mesorhizobium can positively affect the growth of L. japonicus. The total size of the nodules showed a significant positive correlation with the shoot dry weight (Fig. 5), which also supported the importance of rhizobia in determining L. japonicus growth. We note that several other genera might also partially influence plant growth. For example, there were several rhizobial (e.g., Rhizobium and Ensifer) and legume root–nodule endophytic (e.g., Bosea; Pulido-Suárez et al. 2022) genera whose abundance significantly correlated with the total size of nodules when using shotgun metagenomic data.Fig. 4. Effects of the relative abundance of Mesorhizobium bacteria and the local presence of L. japonicus on plant growth (total number of nodules, total size of nodules, and shoot dry weight). Relative abundance of Mesorhizobium was estimated with 16S rRNA amplicon sequencing analysis (a) and shotgun metagenomic analysis (b). Shaded areas represent the 95% confidence intervals for the fitted regression lines, with random effects accounted for by including soil sample as a random intercept. Regression lines are shown when the effects were significant (p < 0.05)Fig. 5. The correlation of the total size of nodules and the shoot dry weight in soil-extracted microbiota inoculation experiments. The two growth indexes showed a significant positive correlation (Kendall’s rank correlation test: p-value = 4.047 × 10^–11^, Kendall’s rank correlation tau = 0.445)
We also found that the local presence of L. japonicus had positive effects on plant growth along with the abundance of Mesorhizobium, specifically the total size of the nodules and shoot dry weight (Fig. 4a; total size of nodules: p = 4.57 × 10^–5^; shoot dry weight: p = 0.07553). A similar trend was observed when using shotgun metagenome data (Fig. 4b; p = 1.25 × 10^–7^). Soil samples collected near L. japonicus plants showed better performance when the abundance of Mesorhizobium bacteria was controlled, indicating that the proportion of L. japonicus-associated microbes that belong to the Mesorhizobium genus may differ according to the local presence of L. japonicus. Soil around L. japonicus plants may harbor more abundant Mesorhizobium symbionts, whereas soil distant from L. japonicus plants may harbor fewer symbionts. Furthermore, we note that the relative abundance of Mesorhizobium bacteria was not significantly associated with the local presence of the host plant, L. japonicus (16S rRNA amplicon sequencing analysis: p = 0.7771; shotgun metagenomic analysis: p = 0.5887; Wilcoxon rank-sum test for both analyses). We found that the abundance of nodulation genes (nodA, nodB and nodC) and a nitrogen fixation gene (nifH) was not significantly associated with the local presence of L. japonicus (Wilcoxon rank-sum test, p > 0.05; Fig. S3). We note that the abundance of these genes was substantially lower than that of several single-copy housekeeping genes (atpD, dnaK and recA; Fig. S3).
Discussion
While it has been common to quantify the partner quality of rhizobia using inoculation experiments (Bamba et al. 2020; Heath and Tiffin 2007), an attempt to quantify the effect of the whole soil microbiota on plant growth and to elucidate the factors determining the growth has been uncommon. In this study, we performed inoculation experiments on the whole soil-extracted microbiota and evaluated the influence of soil sampling conditions, which were the sampling sites and the local distribution of L. japonicus (i.e., near or distant). We found that soil sampling sites, the local presence of L. japonicus, and their interaction significantly correlated with all the measured plant phenotypes (Fig. 3; Table 1).
The plant phenotypes in inoculation experiments were highly variable between sampling sites (Figs. 2 and 3; Table 1). β diversity analyses of the soil microbiota indicated the difference in the overall patterns of microbial composition between sampling sites, which may have influenced the phenotypic variation observed in inoculation experiments. Vegetation may have contributed to the difference in microbial composition between sites. A previous study indicated that plant species identity and plant-induced soil physicochemical properties affect the assembly of the soil microbiome (Byers et al. 2023). Among the soil sampling sites we used in this study, site A exhibited richer vegetation than the other two sites (see Materials and Methods). Abiotic soil properties such as pH and salinity (Fierer and Jackson 2006; Li et al. 2021), and the degree of human disturbance may have also influenced the microbial composition and plant phenotypes between sites.
By integrating the data of the soil-extracted microbiota inoculation experiments and soil metagenomic analysis, we found that the relative abundance of genus Mesorhizobium is the factor most strongly associated with L. japonicus growth (Fig. 4; Table S5). Mesorhizobium was the only genus whose abundance exhibited a significant correlation with the shoot dry weight when we used 16S rRNA amplicon sequencing data to estimate bacterial abundance, and the genus also showed a significant positive correlation with the total size of the nodule. Since Mesorhizobium bacteria have previously been shown to establish symbiosis with L. japonicus (Bamba et al. 2019; Niwa et al. 2001), these results support the role of Mesorhizobium bacteria as the primary symbiotic rhizobium that positively influences L. japonicus growth in natural environments. It is noteworthy that the variations in the Mesorhizobium abundance within the range observed in natural soils seem to underlie the substantial variations in the growth of L. japonicus. Although several studies have detected variations in the effect on host growth among rhizobial strains (e.g., Bamba et al. 2020; Sachs et al. 2010), those have mostly been based on the inoculation of each isolated bacterial strain. Nonetheless, we acknowledge a few caveats for the methods employed in this study. First, the soil slurry used for inoculation may not fully preserve the microbial community present in the original soil. Second, the analysis based on bacterial abundance at the genus level may also represent a constraint, as not all strains belonging to a genus have the same influence on plant. Although we examined the abundance of symbiosis-related genes of L. japonicus-associated Mesorhizobium in the soil samples, the abundance was substantially smaller than housekeeping genes, providing little evidence about whether there is a difference in the abundance of symbiotic Mesorhizobium among soil samples (Fig. S3).
In addition to the positive effect of Mesorhizobium abundance in soil, samples collected near L. japonicus exhibited a significant positive effect on plant growth, at least on the shoot dry weight and total size of the nodules (Fig. 4), suggesting a higher quality microbiome surrounding L. japonicus as symbionts. Given the specific interactions and the selectivity of legume plants for favorable rhizobia (Andrews and Andrews 2017; Gubry-Rangin et al. 2010; Heath and Tiffin 2009; Simms et al. 2006), together with the reports that rhizobia within nodules are released back into soil when nodules senesce (Denison 2000), legume–rhizobia interactions could trigger a positive plant–soil feedback that enriches favorable rhizobia in the surrounding soil of legume plant habitats. While our observation on the positive effect of microbiota extracted from soil near L. japonicus plants would be consistent with positive plant–soil feedback, the underlying process remains unclear. To elucidate the processes driving the difference in the effect of plant growth among soil microbial communities, a direct test of the positive plant–soil feedback hypothesis, such as transplantation experiments and long-term observations of soil microbiome composition in natural L. japonicus habitats, would be required.
Other than Mesorhizobium, several other genera whose abundance significantly correlated with the total size of the nodules may also influence plant growth, albeit not primarily. For example, the genus Lysobacter was identified as positively correlated with the total size of the nodules when 16S rRNA sequencing data were used. Lysobacter enzymogenes has been reported to secrete the heat-stable antifungal factor that exhibits broad-spectrum antifungal activity and prevents plant infection by multiple fungal pathogens such as Uromyces appendiculatus, which causes bean rust (Yuen et al. 2001). The relative abundance of Lysobacter showed a positive correlation with L. japonicus growth, indicating that this genus might also play a role in promoting plant growth. Nonetheless, the total size of the nodules showed a significant positive correlation with the shoot dry weight in the inoculation experiments (Fig. 5), indicating that the abundance of Mesorhizobium is the major factor influencing L. japonicus growth. We note that the abundance of several rhizobial (e.g., Rhizobium, Ensifer) and legume root–nodule endophytic (e.g., Bosea; Pulido-Suárez et al. 2022) genera also correlated with the total size of the nodules; however, none of these have been reported as being a typical symbiont of L. japonicus. Detailed analysis, such as inoculation experiments of these strains, would be interesting to quantify the effect on plant growth.
Supplementary Information
Below is the link to the electronic supplementary material.Supplementary file1 (PDF 4765 kb)Supplementary file1 (XLSX 4765 kb)
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Hannon GJ (2010) Fastx-toolkit. http://hannonlab.cshl.edu/fastx_toolkit/. Accessed 25 Dec 2025
- 2Joshi N, Fass J (2011) Sickle: A sliding-window, adaptive, quality-based trimming tool for Fast Q files (Version 1.33) https://github.com/najoshi/sickle. Accessed 25 Dec 2025.