Sphingomonas sediminicola Is an Endosymbiotic Bacterium Able to Induce the Formation of Root Nodules in Pea (Pisum sativum L.) and to Enhance Plant Biomass Production

The application of bacterial bio-inputs is a very attractive alternative to the use of mineral fertilisers. In ploughed soils including a crop rotation pea, we observed an enrichment of bacterial communities with Sphingomonas (S.) sediminicola. Inoculation experiments, cytological studies, and de novo sequencing were used to investigate the beneficial role of S. sediminicola in pea. S. sediminicola is able to colonise pea plants and establish a symbiotic association that promotes plant biomass production. Sequencing of the S. sediminicola genome revealed the existence of genes involved in secretion systems, Nod factor synthesis, and nitrogenase activity. Light and electron microscopic observations allowed us to refine the different steps involved in the establishment of the symbiotic association, including the formation of infection threads, the entry of the bacteria into the root cells, and the development of differentiated bacteroids in root nodules. These results, together with phylogenetic analysis, demonstrated that S. sediminicola is a non-rhizobia that has the potential to develop a beneficial symbiotic association with a legume. Such a symbiotic association could be a promising alternative for the development of more sustainable agricultural practices, especially under reduced N fertilisation conditions.


Introduction
In the last fifty years, conventional agriculture has been based on mechanisation and extensive use of synthetic fertilisers and pesticides to increase yields [1], which had a negative impact on the environment [2]. More sustainable approaches are sought to compete these intensive practices. The approaches that focus on better management practices [3], precision-agriculture technologies [4], plant breeding strategies [5], or taking advantage of crop biodiversity [6] are the most suitable.
In particular, agriculture based on biological interactions, such as those with rhizospheric microorganisms, is very promising [7]. In recent years, it has been shown that inoculation with symbiotic or nonsymbiotic bacterial strains can significantly improve crop productivity [8][9][10][11]. These bacteria are generally described as plant-growth-promoting rhizobacteria (PGPR), which consist of free-living rhizobacteria and symbiotic bacteria. Freeliving rhizobacteria such as Acetobacter, Azospirillum, Bacillus, Pseudomonas, and Klebsiella can improve plant performance by producing growth regulators (e.g., auxin, cytokinins, gibberellin) or through their aminocyclopropane carboxylate deaminase (ACCD) activity, which lowers ethylene levels known to inhibit plant growth [8,10,11]. Other species, such

Molecular Analyses
From a pool of 3 to 5 surface-sterilised nodules, DNA was extracted using CTAB buffer (2% w/v cetyltrimethylammonium bromide, 100 mM Tris-HCl, 1.4 M NaCl, 20 mM EDTA). Five nanograms of DNA were used to perform a quantitative polymerase chain reaction (qPCR) with a LightCyler 480 system (Roche Diagnostics, Rotkreuz, Switzerland) using Sphingomonasand Rhizobium-specific primers and universal 16SrRNA primers (Table S1). Calibration ranges (0 to 4 ng µL −1 ) were performed for the 16SrRNA and Rhizobium markers from Rlv3841, while that for Sphingomonas was determined using S. sediminicola DNA. The 16S rRNA abundance was corrected to account for variations in gene copy-number according to the 16S copy number database rrnDB [49], i.e., 3 copies for R. leguminosarum and 1 copy for S. sediminicola.
S. sediminicola genome sequencing was performed by Beijing Genomics Institute (BGI, TaI Po, Hong Kong). S. sediminicola DNA from a single colony was isolated using the EZNA Tissue DNA Kit (Omega Biotek, Norcross, GA, USA). A negative control consisting only of DNA extraction solutions was included in the DNA extraction procedure to exclude cross-contamination. Genomic libraries were prepared with the Nextera XT Library Prep kit (Illumina, MA, Canton, OH, USA) and checked with the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) using the Agilent High Sensitivity DNA Kit (Agilent Technologies). Libraries were then pooled in equimolar amounts according to quantifica-Microorganisms 2023, 11, 199 4 of 22 tion by the Qubit ds-DNA HS Assay (Thermo Fisher Scientific, Waltham, MA, USA) and sequenced using a Hiseq 2500 (Illumina). The quality of the sequencing raw data was checked using FastQC (v0.11.8). Trimmomatic (v0.39) [50] was used to remove the adapters (ILLUMINACLIP option) and low-quality regions (HEADCROP:12 and SLIDINGWIN-DOW:5:30). Sequences with a length shorter than 80 base pairs (bp) were discarded. A total of 5 independent sequences were generated with an average of 592 Mb. Using the tools of the KBase platform [51], the assembly of the sequences was performed with SPAdes (v3.15.3). Basic assembly properties were assessed using QUAST (v4.4) and genome annotation using Prokka (v1.14.5). Annotations were also manually checked using the BLASTx program from the NCBI database (https://blast.ncbi.nlm.nih.gov/Blast.cgi (accessed on 7 December 2022)) and validated according to sequence similarity scores, E-values, and coverage. Prokka analysis also generated the functional categories of these annotations by comparing them in the SEED system ontology. Gene Ontology analysis and Metacyc pathway analysis were also performed using the QuickGO and MetaCyc databases, respectively [52,53]. Circular visualisations were created using the Circular Genome Visualization Tool (v0.0.2) from the KBase platform.
The similarity of S. sediminicola genome sequences compared to that of other bacteria was analysed by calculating the pairwise Average Nucleotide Identity (ANIb and ANIm) values using JSpecies software with the integrated BLAST algorithm [54]. In silico DNA-DNA hybridisations (DDH) values were obtained using the Genome-to-Genome Distance Calculator (GGDC 2.1; http://ggdc.dsmz.de/distcalc2.php (accessed on 7 December 2022), [55]) and the recommended BLAST + alignment. The bacterial genome sequences were downloaded from the GenBank database (https://www.ncbi.nlm.nih.gov/genbank/ (accessed on 7 December 2022)).

Phylogenetic Analyses
Multiple sequence alignments of the nifH, parA, nodABCD, and nfeD sequences were performed individually using CLUSTALW (https://www.genome.jp/toolsbin/clustalw (accessed on 7 December 2022)). CLUSTALW first calculated a pairwise genetic distance between the sequences with degrees of similarity between each pair. Then, a phylogenetic tree was constructed for each alignment using the neighbour-joining algorithm with no distance corrections. The trees were generated in the Interactive Tree of Life (iTOL) programs (v6.5.2) [56]. The sequence accessions from NCBI (https://www.ncbi.nlm.nih.gov/ (accessed on 7 December 2022)) are indicated in the phylogenetic tree after the species name. The bootstrap values of the phylogenetic tree are indicated in blue.

Acetylene Reduction Assay
Acetylene (C 2 H 2 ) was prepared according to Postgate [57] from 1 g of calcium carbide (CaC 2 ; Sigma Aldrich) dissolved in 150 mL of distilled water producing 15.625 mmol of C 2 H 2 . ARA was performed in 250 mL flasks containing detached nodulated root system. Ten percent of the flask volume was replaced with C 2 H 2 [58], and after 1 h of incubation, a gas aliquot (1 mL) was analysed using an ethylene (C 2 H 4 ) analyser (F-950, Felix Instruments, Camas, WA, USA). Ammoniacal silver nitrate (10 g L −1 ) was used to precipitate residual acetylene [59]. Nase activity from the detached root system was expressed in µmol C 2 H 4 h −1 plant −1 .

Macroscopic and Microscopic Observations
To monitor root nodule colonisation, S. sediminicola Rif [pOPS0385] (2 10 6 CFU) was inoculated into 5-day-old etiolated peas growing in sterile vermiculite. Plants were grown under the same conditions as in the controlled growth chamber experiment and watered with N-free Fahraeus solution. The experiment was repeated three times independently. Detection of β-glucuronidase (GUS) enzyme activity (blue colour) in nodule sections of pea inoculated with S. sediminicola Rif [pOPS0385] was visualised with 5-bromo-4-chloro-3-indolyl glucuronide (X-Gluc; Sigma-Aldrich). The staining buffer was prepared under high stringency conditions for ferro-and ferricyanide concentrations (2.5 mM ferroferricyanide buffer, K 3 Fe(CN) 6 , K 4 Fe(CN) 6 , Sigma-Aldrich) to limit diffusion of GUS product in sodium phosphate buffer (0.1 M NaH 2 PO 4 , 0.1 M Na 2 HPO 4 , pH 7.0 and 0.1% (v/v) Tritton; Sigma-Aldrich) with X-Gluc (1 mg mL −1 ) previously dissolved in dimethylformamide (Sigma-Aldrich). Staining for GUS activity was performed on surface-sterilised nodules for 45 min at 37 • C in the dark. Macroscopic visualisation of nodules was performed using a stereomicroscope (ZEISS SteREO Discovery V20, Carl Zeiss, Göttingen, Germany). For the visualisation of GUS activity at the intracellular level, detached nodules were fixed for 1 h in 3% (w/v) glutaraldehyde in 0.1 M phosphate buffer (pH 7.4); dehydrated gradually using ethanol solutions from 15,30,50,70,80,90, and 100%; and then embedded in LR white resin (Sigma-Aldrich). Polymerisation was carried out in gelatin capsules for 10 h at 54 • C. Sections of 4 to 6 µm were cut with a histological diamond knife and floated on drops of sterile water on silanised slides at 60 • C. Sections were back stained with periodic acid-Schiff (PAS) [60] to visualise the insoluble polysaccharides (cell wall and starch). For both light and electron microscopy, nodules were fixed in 3% (w/v) glutaraldehyde dissolved in 0.1 M cacodylate buffer for 3 h at room temperature. Samples were then washed in the cacodylate buffer and post-fixed for 1 h in 1% (w/v) OsO 4 . The samples were then processed and embedded in resin as described above to visualise the GUS activity. For light microscopy, 1 µm thin sections were placed on silanised slides and stained first with PAS and then with Azur II (Agar Scientific, Stansted, Essex, England) 1% (w/v) in water. Observations were carried out on a Nikon Eclipse E800 microscope (Nikon, Tokyo, Japan). For electron microscopy, 70 nm ultrathin sections were mounted on 400-mesh nickel grids or 200-mesh nickel carbon formvar-coated grids, dried at 37 • C, counterstained with 5% uranyl acetate in water, and observed with a Tecnai F20 FEI Corpelectron microscope at 120 kV (FEI, Hillsboro, OR, USA).

Statistical Analysis
Means were compared between plants inoculated with S. sediminicola or R. leguminosarum and non-inoculated plants using the Kruskal-Wallis test (p < 0.05) followed by pairwise Wilcoxon rank sum tests with Holm's p-adjust method for multiple comparisons. All statistical testing was carried out in R (v4.0.4; http://www.r-project.org/) using agricolae [61] and ggpubr [62] packages.

Sphingomonas sediminicola Improved Pea Biomass Production
First, the effect of S. sediminicola inoculation on the main phenotypic traits of pea plants was investigated. On a sterile NPK 18-10-20 potting soil (N-rich substrate), we observed that both shoot and root dry weights of 30-dpi-old plants inoculated with S. sediminicola were at least twice higher those of non-inoculated plants (p < 0.001; Figure 1a,b). Although shoot height was not altered in the inoculated plants (Figure 1c), we observed that the main root was significantly longer compared to that of non-inoculated plants (p < 0.001; Figure 1d).
To compare the effect of S. sediminicola inoculation on the different phenotypic traits measured, a rhizobia species (R. leguminosarum) that develops a typical N 2 -fixing association in pea was used as a positive control. We observed an increase in both shoot and root biomass production when plants were inoculated with Rhizobium (p = 0.027 and <0.001, respectively, Figure 1a,b) compared to non-inoculated plants. We also observed that shoot height was higher in plants inoculated with R. leguminosarum (p < 0.001), while root length was not changed. On a sterile NPK 4-4-4 potting soil (N-limited substrate) and on sterile vermiculite (N-free substrate), an increase in both plant biomass production and shoot height was observed in the presence of R. leguminosarum. In all non-inoculated plants (>100), the presence of nodules was never observed. In the plants inoculated with R. leguminosarum, the percentage of plants with nodules increased under N-deficient conditions and reached 70% under N-free conditions with 5 to 88 nodules in nodulated plants (Figure 1e). To compare the effect of S. sediminicola inoculation on the different phenotypic traits measured, a rhizobia species (R. leguminosarum) that develops a typical N2-fixing association in pea was used as a positive control. We observed an increase in both shoot and root biomass production when plants were inoculated with Rhizobium (p = 0.027 and <0.001, respectively, Figure 1a,b) compared to non-inoculated plants. We also observed that shoot height was higher in plants inoculated with R. leguminosarum (p < 0.001), while root length was not changed. On a sterile NPK 4-4-4 potting soil (N-limited substrate) and on sterile vermiculite (N-free substrate), an increase in both plant biomass production and shoot height was observed in the presence of R. leguminosarum. In all non-inoculated plants (>100), the presence of nodules was never observed. In the plants inoculated with R. leguminosarum, the percentage of plants with nodules increased under N-deficient conditions and reached 70% under N-free conditions with 5 to 88 nodules in nodulated plants ( Figure  1e).
When plants growing on the N-rich substrate were inoculated with S. sediminicola, root nodules were never observed. However, 37.5% and 76.5% of the plants inoculated When plants growing on the N-rich substrate were inoculated with S. sediminicola, root nodules were never observed. However, 37.5% and 76.5% of the plants inoculated with S. sediminicola and grown under N-limited and N-free conditions, respectively (Figure 1e), developed numerous nodule-like structures similar to those observed when the plants were inoculated with R. leguminosarum (Figure 2a). The Sphingomonas-specific gene was detected in these detached nodules. In contrast, no amplification was observed when Rhizobiumspecific primers were used. Conversely, in nodules that developed after inoculation with R. leguminosarum, PCR amplification products were detected with the Rhizobium-specific primers but not with the Sphingomonas-specific primers (Figure 2b). In the detached root system of plants inoculated with S. sediminicola, nitrogenase activity was detected at levels comparable to those measured in plants inoculated with R. leguminosarum ( Figure 2c).
1, x FOR PEER REVIEW 7 of 22 detached root system of plants inoculated with S. sediminicola, nitrogenase activity was detected at levels comparable to those measured in plants inoculated with R. leguminosarum ( Figure 2c). (c) Measured nitrogenase activity using the ARA with the noninoculated and non-nodulating root system (black) of pea plants, as well as the nodulated root system of peas inoculated with R. leguminosarum (red) and S. sediminicola (green). Statistical differences were based on Wilcoxon rank sum tests with Holm's p-adjust. ***, p < 0.001; ns, not significant.

Sphingomonas sediminicola Genome Contains the Genetic Information Necessary to Induce Nodulation
de novo whole-genome sequencing showed that S. sediminicola had a chromosome of 2,756,125 bp with a GC content of 62.88%. Thus, the genome coverage of the different sequencing runs ranged from 165 to 260X. The bacterial genome was composed of 2764 complete coding sequences (CDSs) and 64 tRNA sequences (Figure 3a). The similarity of S. sediminicola genome sequence compared to that of S. sediminicola KACC 15039, Sphingomonas azotifigens, and some rhizobia was analysed using ANI-Blast (ANIb), ANI-MUM- and R. leguminosarum (red). (c) Measured nitrogenase activity using the ARA with the noninoculated and non-nodulating root system (black) of pea plants, as well as the nodulated root system of peas inoculated with R. leguminosarum (red) and S. sediminicola (green). Statistical differences were based on Wilcoxon rank sum tests with Holm's p-adjust. ***, p < 0.001; ns, not significant.

Sphingomonas sediminicola Genome Contains the Genetic Information Necessary to
Induce Nodulation de novo whole-genome sequencing showed that S. sediminicola had a chromosome of 2,756,125 bp with a GC content of 62.88%. Thus, the genome coverage of the different sequencing runs ranged from 165 to 260X. The bacterial genome was composed of 2764 complete coding sequences (CDSs) and 64 tRNA sequences (Figure 3a). The simi-larity of S. sediminicola genome sequence compared to that of S. sediminicola KACC 15039, Sphingomonas azotifigens, and some rhizobia was analysed using ANI-Blast (ANIb), ANI-MUMmer (ANIm), and in silico DDH. High ANIb, ANim, and DDH values (>96%) were found between the genomes of S. sediminicola and S. sediminicola KACC 15039, whereas comparisons with the other selected genomes gave much lower values depending on the three types of analysis (Table 1). Among these CDSs, the most abundant were those encoding proteins involved in amino acids and related metabolism (6.8%), membrane transport (5.8%), protein fate (5.3%), carbohydrate metabolism (4.3%), and some small molecules (including cofactors, vitamins, prosthetic groups) (4.2%) (Figure 3b). More precisely, and despite the detection of 1018 CDSs encoding proteins of unknown function, we found that some show strong homology with genes involved in iron uptake (fur, piuB), storage (bfrA, bfrD, bfd), transport (feoB, fecR, fieF, tonB), scavenging (entS, yfiZ), phosphorus transport (pstABCS), and signal transduction (phoBHRU). Genes encoding proteins involved in tryptophan biosynthesis (trpABCD) and, interestingly, in N 2 fixation (fixJKL, nifSU-like) were also identified ( Figure 3a, Table S2). between the genomes of S. sediminicola and S. sediminicola KACC 15039, whereas comparisons with the other selected genomes gave much lower values depending on the three types of analysis (Table 1).  The presence of a plasmid of 519,958 bp was also detected. It contained a number of CDSs showing strong similarity to genes involved in the replication of low-copy-number plasmids (parABC), nodule formation and development (nodABCDEFLMPQ, nolKVT; noeLJ), N 2 fixation (nifBHRUX, fixGHIJNOPQS), and nitrate assimilation (narGHIJKQ, napA, nirBD). In addition, this structure harboured two clusters of CDSs sharing strong similarities with genes of type three (T 3 SS, escRSTUV; sctDOPQ) and four (T 4 SS, virB2,3,4,6,9,10,11) encoding proteins involved in secretion systems. We also identified other CDSs sharing strong homology with bacA (Bacteroid development protein A), acdS genes, and a number of genes involved in the regulation of bacterial motility when chemical attraction occurs in the rhizosphere (cheCY, fliL, flbD, flgDF) (Figure 3c, Table S3).
Phylogenetic analysis showed that the nifH sequence of S. sediminicola clustered with that of another Sphingomonas species (S. azotifigens) in the clade that included Azorhizobium caulinodans. In addition, we found that S. sediminicola nifH sequence was also closer to those of the selected β-rhizobia and Bradyrhizobium species compared with the sequences of other α-rhizobia ( Figure 4). Conversely, the partitioning system (parA) sequences of S. sediminicola and S. azotifigens were more closely related to that of Rhizobium than that of other Sphingomonas species ( Figure S1a). Unlike those of nodC, the nodA and nodB sequences of S. sediminicola were found to be quite similar to those of Rhizobium ( Figure S1b-d). The nodD sequence of S. sediminicola was close to that of Azorhizobium caulinodans, whereas the nfeD sequences formed a cluster unique to Sphingomonas species and distinctly different from those of other bacterial genera (Figure S1e,f). virB2, 3,4,6,9,10,11) encoding proteins involved in secretion systems. We also identified other CDSs sharing strong homology with bacA (Bacteroid development protein A), acdS genes, and a number of genes involved in the regulation of bacterial motility when chemical attraction occurs in the rhizosphere (cheCY, fliL, flbD, flgDF) (Figure 3c, Table S3). Phylogenetic analysis showed that the nifH sequence of S. sediminicola clustered with that of another Sphingomonas species (S. azotifigens) in the clade that included Azorhizobium caulinodans. In addition, we found that S. sediminicola nifH sequence was also closer to those of the selected β-rhizobia and Bradyrhizobium species compared with the sequences of other α-rhizobia ( Figure 4). Conversely, the partitioning system (parA) sequences of S. sediminicola and S. azotifigens were more closely related to that of Rhizobium than that of other Sphingomonas species ( Figure S1a). Unlike those of nodC, the nodA and nodB sequences of S. sediminicola were found to be quite similar to those of Rhizobium ( Figure S1bd). The nodD sequence of S. sediminicola was close to that of Azorhizobium caulinodans, whereas the nfeD sequences formed a cluster unique to Sphingomonas species and distinctly different from those of other bacterial genera (Figure S1e,f).

Sphingomonas sediminicola Induced the Formation of Nodules
To unequivocally demonstrate that S. sediminicola is responsible in pea plants for nodule formation, inoculation of pea plants with a modified S. sediminicola strain containing a GUS reporter gene (S. sediminicola Rif [pOPS0385]) was performed in sterile vermiculite and with an N-free nutrient solution. This experiment was performed to determine if the

Sphingomonas sediminicola Induced the Formation of Nodules
To unequivocally demonstrate that S. sediminicola is responsible in pea plants for nodule formation, inoculation of pea plants with a modified S. sediminicola strain containing a GUS reporter gene (S. sediminicola Rif [pOPS0385]) was performed in sterile vermiculite and with an N-free nutrient solution. This experiment was performed to determine if the reporter gene activity could be detected in the nodule tissues. Compared with the wild-type strain, the changes in shoot and root biomass production and in the number of nodules per plant were similar ( Figure S2). We also verified that the wild-type strain did not exhibit GUS activity, either under free-living conditions or in the nodules of inoculated plants. In mature nodules (30 dpi), GUS activity (blue staining) was detected only in the central area of the nodules (Figure 5a, arrow), whereas roots (R), vascular bundles (VB), cortical cell layers (C), and apical area (AA) remained unstained. Thinner nodule sections confirmed that GUS activity was present in the majority of cells in the central zone of the nodule (Figure 5b-d). In this area, we also observed that some cells containing large vacuoles remained unstained (stars) (Figure 5b,c), similar to those of the cortical area (C) (Figure 5c). PAS counterstaining also revealed the presence of numerous starch granules (SG) appearing as dark-red dots. At higher magnification (Figure 5d), GUS staining was visible in small particles corresponding to bacteroids (B) with an expected size of 1 µm.
To refine the cellular structure of the central zone of the nodules and to determine whether the infected cells contained bacteroids, light and electron microscopy experiments were performed on thin and ultrathin sections of nodule tissue, respectively. An overview of a nodule longitudinal thin section (Figure 6a) confirmed that three distinct cell types were clearly observed, consisting of an apical area (AA) characteristic of indeterminate pea nodules [61] containing small cells without starch, an intermediate area (IA) composed of larger blue-stained cells without starch, and a fully differentiated and mature area (MA) represented by large blue-stained cells with numerous starch granules (Figure 6a). Infection threads (ITs) were frequently observed in the intermediate area. These ITs contained small spherical or rod-shaped, blue-stained particles resembling bacteria (Figure 6b). These blue-stained particles were also detected in the cytoplasm of the adjacent cells (Figure 6b, arrows). Magnification of the central area allowed us to detect within the cell cytoplasm the presence of many almost unstained ovoid particles (Figure 6c) surrounded by a white halo (Figure 6d, arrows). It is likely that a membrane surrounding these particles was present, thus indicating that they could be bacterial symbiosomes.
Transmission electron microscopy (TEM) experiments confirmed that the particles observed by light microscopy were symbiotic bacteria and that they were polymorphic. In the intra-or intercellular IT near the apical area visible in Figure 6b, the average surface of the bacteria (B) was 0.46 ± 0.11 µm 2 (Figures 6e and S3). We also observed that the bacteria were surrounded by an electron-transparent matrix (arrows), which was itself included in a slightly denser fibrous matrix with no separation of membranes between the two matrices. In many cells of the apical area and in cells containing IT, the size of the bacteria ranged from 0.5 to 1 µm, and as in the central zone of the nodule, we observed many membrane structures, suggesting that an endocytosis process had occurred (Figure 6f,g,  arrows). This process probably resulted in the formation of symbiosomes called bacteroids, each consisting of a single symbiotic bacterium and a peribacteroid membrane (Figure 6h, arrow). The size of this structure changed while the nodule developed to reach an average surface of 1.62 ± 0.66 µm 2 in the intermediate area and 4.37 ± 1.56 µm 2 in the mature area ( Figures S3 and 6i).
confirmed that GUS activity was present in the majority of cells in the central zone of th nodule (Figure 5b-d). In this area, we also observed that some cells containing large vac uoles remained unstained (stars) (Figure 5b,c), similar to those of the cortical area (C) (Fig  ure 5c). PAS counterstaining also revealed the presence of numerous starch granules (SG appearing as dark-red dots. At higher magnification (Figure 5d), GUS staining was visibl in small particles corresponding to bacteroids (B) with an expected size of 1 µm. bacteria (Figure 6b). These blue-stained particles were also detected in the cytoplasm of the adjacent cells (Figure 6b, arrows). Magnification of the central area allowed us to detect within the cell cytoplasm the presence of many almost unstained ovoid particles ( Figure  6c) surrounded by a white halo (Figure 6d, arrows). It is likely that a membrane surrounding these particles was present, thus indicating that they could be bacterial symbiosomes.

Discussion
Enrichment of Sphingomonas has been frequently observed in agricultural soils after crop rotation, ploughing, cover crops establishment, and in relation to N fertilisation [29,63]. However, these studies dealt with metabarcoding characterisation of bacterial communities and did not directly address the function of Sphingomonas species [64][65][66]. In particular, the predominance of S. sediminicola in the pea rhizosphere raised the question of its role as a biological indicator in conventionally tilled agricultural soils [29]. In the present study, we showed that S. sediminicola improved plant biomass production similarly to other PGPR [8,11,67], such as Azospirillum brasilense [68], Bacillus licheniformis [69], Pseudomonas aeruginosa [70], or other Sphingomonas species. For example, S. LK11 was able to stimulate growth and biomass production in soybean [71] and tomato [40]. A similar effect was observed in Arabidopsis thaliana with S. Cra20 [64]. The mechanisms underlying these beneficial effects mainly involve facilitation of nutrient acquisition (phosphorus, iron), modulation of plant hormone levels and ACCD activity [72].
The S. sediminicola genome sequence allowed us to predict the potential capabilities of this strain. First, the size of the S. sediminicola genome (2.75 Mb) is among the smallest compared to other Sphingomonas genomes, ranging from 2.88 Mb in Sphingomonas sp. W1-2-3 [66] to 6.58 Mb in S. sanxanigenens [73]. As ANI values below 95-96% or 70% for DDH are the cut-off points for delineating bacterial species, the genome of S. sediminicola analysed in the present study was found to be quite different from that of rhizobia species and to be similar to that of Sphingomonas sediminicola KACC 15,039 [43]. However, we identified an extrachromosomal element of 520 kb containing a CDS that has strong homologies with low-copy-number plasmids carrying a partitioning locus, called the ParABC cassette [74]. Therefore, this extrachromosomal element could be a plasmid that we named pSs01. Within the Sphingomonas genus, many plasmids have been identified, and most of them have this ABC partitioning system [75,76] and are also large, such as pCAR3 in Sphingomonas sp. KA, pISP0 in Sphingomonas sp. MM-1, and pSWIT01 in Sphingomonas wittichii RW1 [77]. We found some CDSs on the chromosome of S. sediminicola that show homology with genes involved in iron absorption (fur), iron storage (bacterioferritins bfrA and ferrodoxin bfrD), and iron transport (ferrous iron transporters, feoAB), as well as a gene encoding a pump involved in ferrous ion efflux (fieF) [78]. In parallel, we identified CDSs close to genes related to the synthesis of enterobactin siderophores (entS) and siderophore transporters (yfiZ, tonB). These genes are present in many PGPR such as Enterobacter sp. J49 [79], Bradyrhizobium yuanmingense [80], or R. cellulosilyticum [81] and Sphingomonas pokkalii [82]. CDSs involved in phosphorus metabolism, such as specific inorganic phosphorus transporters (pstABCS) and transcriptional regulators of phosphorus metabolism (phoBHRU), have also been identified on the chromosome of S. sediminicola. These components are found in several PGPRs, such as Burkholderia cenocepacia [83], Paenibacillus sonchi [84], or Pseudomonas psychrotolerans [85]. CDSs sharing homologies with genes related to tryptophan biosynthesis (trpABCD), which may act as a biosynthetic precursor of auxins, were found in the S. sediminicola genome. Many PGPR, such as Acetobacter, Azospirillum, Bacillus, Bradyrhizobium, Burkholderia, Klebsiella, Pseudomonas, Rhizobium, Xanthomonas [86,87], and Sphingomonas LK11 [37,66], are auxin producers. Interestingly, we identified a CDS in pSs01 that shows strong homology with acdS, a gene encoding the enzyme ACCD. This bacterial enzyme catalyses the reduction and cleavage of the ethylene precursor (ACC) produced by plants [88]. Therefore, the next step will be to further characterise these bacterial properties that may explain the improved plant performance when S. sediminicola is inoculated not only with pea, but also with other monocotyledonous and dicotyledonous species.
One of the most striking results of our study was that S. sediminicola was able to induce nodulation on pea roots. Root nodule development is a process restricted to very specific plant-bacteria interactions, usually involving rhizobia and legumes [16][17][18]. During this interaction, and under N-deficient conditions, flavonoids are excreted by the plant into the rhizosphere and interact with the bacterial transcription factor NodD [89], which activates the transcription of a set of nodulation (nod) genes. In the well-characterised Rhizobium-legume symbiosis, nodABC transcription leads to the synthesis and secretion of lipochitooligosaccharides, the so-called Nod factor (NF), which triggers the entry of the bacteria and nodulation [89,90]. Interestingly, we found several CDSs with strong similarity to nodABC and nodD in plasmid pSs01. The presence or absence of specific nodulation genes (nod, noe, and nol) determines NF structure [91] by controlling host specificity [17]. The NF structure is also specific to the host plant species and can also be altered according to environmental conditions [90]. In R. leguminosarum, nodEL are involved in the formation of root nodules in pea and clover [92,93], while in R. meliloti, nodHPQ are required for alfalfa nodulation. It is also likely that nodHPQ are involved in the process of soybean nodulation in Bradyrhizobium japonicum [93]. A number of CDSs identified in the genome of S. sediminicola shared strong homology with genes controlling host specificity, such as nodELMPQ [92,94,95]. Such a finding suggests that S. sediminicola could be able to induce nodule formation in temperate legume species other than pea, as well as in tropical legumes.
It is known that other biological components such as root colonisation and symbiont recognition and suppression of the plant immune system are involved in the establishment of a rhizobia-legume symbiosis. All these steps require specific secretion systems, namely, types 1, 3, and 4 [96], which translocate effectors into the host plant. In R. leguminosarum bv. viciae, the type 1 protein secretion system (T 1 SS), which is encoded by the prsD and prsE genes, is involved in the secretion of EPS-glycanases [97]. These enzymes play a key role in biofilm formation both during root colonisation and in the initial steps of symbiotic interaction [96]. In the chromosome of S. sediminicola, CDSs encode proteins similar to prsD and prsE, suggesting that EPS-glycanases can be excreted by the bacteria.
Host-produced flavonoids induce expression of the nodD gene encoding NF, which also activates expression of type 3 secretion genes (T 3 SS). We identified in the plasmid pSs01 that many CDSs share strong similarities with the members of the T 3 SS [98] that form a cluster of genes similar to that found in the T 3 SS core components of the Rhc-I subgroup [99]. T 3 SS is the only secretion system involved in the establishment of symbiosis with legumes [100]. This type of T 3 SS is present in many rhizobia, such as R. elti, Ensifer (Sinorhizobium) fredii, and Bradyrhizobium japonicum [99]. Close to the T 3 SS cluster in pSs01, we also identified nine CDSs sharing strong similarities with T 4 SS-b, which is functionally similar to the T 3 SS-Rch-1 [100]. T 4 SS has been identified in some rhizobia, such as Mesorhizobium loti R7A [101] and R. etli CFN42 [102].
In temperate legumes that develop indeterminate nodules, N 2 fixation occurs in infected cells located in the zone III. This zone is visible in the root nodules of pea plants inoculated with S. sediminicola ( Figure 5). It is known that the leghaemoglobin in this zone prevents the inhibition of the NAse enzyme by O 2 while maintaining cellular respiration [103,104]. Thus, ATP can be produced via oxidative phosphorylation by a cytochrome c oxidase (Cbb3-type) in the mitochondria [104][105][106]. The genes fixNOQP [16,105], which encode cytochrome c oxidase, are located near the fixGHIS genes that enable the assembly of Cbb3. We found two clusters of CDSs in the S. sediminicola genome that have very strong homologies with fixNOQP and fixGHIS. These two clusters were located next to each other on pSs01. They were also found in a similar position as in the symbiotic plasmids pSymA and p110 of Ensifer (Sinorhizobium) meliloti [17] and R. leguminosarum, respectively [105], and in the symbiotic islands of Bradyrhizobium japonicum USDA 110 [105]. The regulation of the fixNOQP and fixGHIS operons is usually controlled by the fixJKL genes, which were also present in pSs01. As in some rhizobia, such as R. leguminosarum 3841 [105], the fixKL regulatory genes were located on the chromosome of S. sediminicola, while the fixJ gene was found on the plasmid pSs01. As in Ensifer meliloti [105], fixNOQP were also located in S. sediminicola next to the CDSs, exhibiting strong homology to genes encoding proteins involved in the control of N metabolism, such as napA (a periplasmic nitrate reductase), nirDB, and narGHIJKQ (involved in nitrite and nitrate reduction, respectively).
Regulation of O 2 partial pressure during symbiotic N 2 fixation by rhizobia is also an essential process for maintaining a fully active Nase. In these bacterial species, the nifAB and nifHDKEN genes are involved in the synthesis and assembly of the enzyme [17,107].
nifA encodes a transcriptional regulator required for the transcription of nifB, nifN, and nifHDKEX. nifH encodes a reductase that enables the production of NH 4 + from atmospheric N 2 , while nifB is required for the biosynthesis of a specific cofactor located at the active site of Nase. The assembly of this cofactor, composed of iron molybdenum (FeMo-co), is in turn initiated by the gene products nifS and nifU [107,108]. Several CDSs with similarity to nifBHU were identified in the megaplasmid of S. sediminicola. The nifS gene was located on its chromosome as well as other CDSs sharing homologies with nifX, nifE, and nifN, which encode the NAse FeMo-cofactor [16]. Such a distribution of the different nif genes on a plasmid and on the chromosome has also been found in other rhizobia, such as Ensifer meliloti [17]. The analysis of the S. sediminicola genome allowed the identification of several genes encoding proteins involved in the onset of atmospheric N 2 fixation. However, some of the genes essential to this process, such as nifD and nifK [109], were not identified in this species. As this analysis also revealed the presence of several genes encoding proteins of unknown function, further work is required to determine whether some of these genes are essential for a fully functional Nase enzyme. Our phylogenetic analysis showed that the acquisition of nodulation genes by S. sediminicola involved dozens of genes and was thus rather complex. Therefore, it can be hypothesised that the sequences of S. sediminicola nifD and nifK are so distant from those of other rhizobia that they could not be identified on the basis of sequence homologies alone.
At the ultrastructural level, different cell types were identified in the mature root nodules of pea plants inoculated with S. sediminicola, which are characteristic of indeterminate nodules. From their apical to their basal area, these nodules are composed of an initial zone of meristematic cells that give rise to cells that form the infection zone with some cell layers full of starch granules, followed by cells in which the bacteria progressively differentiate [20,110]. Light and electron microscopic observations allowed us to follow the different steps in the establishment of the symbiotic association at the cellular level, including the penetration of the bacteria into the root cells and the formation of a symbiosome. In particular, the presence of IT and an electron-translucid matrix probably corresponding to an infection droplet that allows a direct contact between the bacteria and the root cell membrane was clearly visible in the apical zone. We were also able to visualise the process of bacterial endocytosis through the plasma membrane in the infected zone, followed by the development of a symbiosome compartment in which the bacteria differentiate into bacteroids [111]. This differentiation is known to be plant host dependent [112], and in legumes belonging to the inverted repeat-lacking clade (IRLC), this process leads to terminal differentiation of their bacterial endosymbionts, resulting in endoreduplication of the genome, cell enlargement, and loss of cell division ability [113]. In nodules of pea inoculated with S. sediminicola, we identified bacteria in the infection zone that were similar in size to those previously observed in free-living bacteria [43] (i.e., 0.5 µm diameter by 1 µm length), while in zone III, the size of the bacteroids increased fivefold and is characteristic of the differentiation of bacteroids when they begin to fix atmospheric N 2 [20]. This observation suggests that in the symbiotic association of S. sedimincola and pea, the resulting indeterminate nodules contain E-type bacteroids. However, it would be useful to analyse the expression of NCR-like genes by RT-qPCR and the DNA content by flow cytometry.
In conclusion, our results rekindle current concepts on nodulation of legumes both biological and evolutionary perspectives, especially for further agronomic applications. For almost a century, it was generally assumed that all legumes could form nodules only when inoculated with bacteria of the order Hyphomicrobiales (=Rhizobiales), which belong to the α-Proteobacteria (α-rhizobia). Then, in the last two decades, the discovery of modulating bacteria belonging not to the classical α-rhizobia but to the βor γ-Proteobacteria (β-or γ-rhizobia) has overturned the postulate that only α-rhizobia are able to develop N 2 -fixing symbiotic associations with legumes [16][17][18]. By showing that within the α-Proteobacteria, nodulation of legumes is not restricted to the Hyphomicrobiales but can also occur in the Sphingomonadales, it can be suggested that the classification of α-rhizobia taxa could be revised to include S. sediminicola or other species belonging to the Sphingomonadales.

Conclusions
In summary, our work demonstrates that S. sediminicola possesses strong potential with respect to legume N nutrition, yield, and their ability to produce green manure, as this soil bacterium can develop functional root nodules when inoculated with pea. To further promote the biological and agronomic importance of the Sphingomonas-legume association, more research is needed to (1) decipher the structural and regulatory elements involved in such an association both during its establishment and during the onset of N 2 fixation; (2) evaluate the benefits of such an association in agricultural systems where peas or other legumes are grown either as the main crop, as green manure, or as an associated crop in the presence or absence of Rhizobium; and (3) investigate its short-or long-term effects on the microbial community in the rhizosphere of the surrounding plant and on soil properties.

Supplementary Materials:
The following supporting information can be downloaded at https://www. mdpi.com/article/10.3390/microorganisms11010199/s1. Table S1: Primers used to target specific gene sequences. Table S2: List of genes in the genome of Sphingomonas sediminicola. The coordinates in bp are given for each CDS identified by the Prokka analyses with their orientation as well as their description and symbol. Table S3: List of genes in the plasmid pSs1 of S. sediminicola. The coordinates in bp are given for each CDS with their orientation. The sequence similarity scores, Evalues, and coverage from Blast analysis are indicated, as well as the blast target sequences. Figure S1: Phylogenetic tree of the multiple sequence alignment of (a) parA, (b) nodA, (c) nodB, (d) nodC, (e) nodD, and (f) nfeD sequences from atmospheric N 2 fixing symbiotic and non-symbiotic bacterial species. Sphingomonas sediminicola is in red. Bootstrap values of the phylogenetic tree are indicated in blue. Figure S2: Effect of inoculation with Sphingomonas sediminicola Rif [pOPS0385] (purple) or Rhizobium leguminosarum (red) on some phenotypic traits of peas compared to uninoculated peas (black) when plants were grown in the absence of nitrogen. (a) Shoot height, (b) root length. (c) Shoot and (d) root dry weight (DW). (e) Number of root nodules per plant. Statistical differences were based on Wilcoxon rank sum tests with Holm's p-adjust. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns, not significant. Figure S3: Measurement of the surface area of S. sediminicola in three different root nodule zones. Surface areas were measured by TEM image analysis on 30 to 50 longitudinally sectioned bacteria or bacteroids; partial and transverse sections were not included in the measurements. Significant differences (different letters) between infection threads and intermediate and mature zones were determined by ANOVA followed by Tukey-Kramer HSD at a 95% confidence limit. Funding: This research was funded by AgroStation, the Région Hauts de France, and the European Regional Development Fund (ERDF) within the framework of the ERDF/FSE 2014-2020 through PhD grant (CM). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data Availability Statement: Whole-genome sequencing data associated with this study has been deposited in the NCBI Sequence Read Archive under accession number PRJNA818132.