Experimental Evolution of a Plant Pathogen into a Legume Symbiont

1Bacteria known as rhizobia have evolved a mutualistic endosymbiosis of major ecological importance with legumes that contributes ca. 25% of global nitrogen cycling. Rhizobia induce the formation on legumes of root nodules that they colonize intracellularly and in which they fix nitrogen to the benefit of the plant. Rhizobia are taxonomically, metabolically, and genetically diverse soil bacteria,. They are currently distributed in 12 genera of α- and β-proteobacteria intermixed with saprophytes and pathogens. The occurrence of rhizobia in several distant genera is thought to have originated from repeated and independent events of horizontal transfer of key symbiotic functions in non symbiotic bacterial genomes,. Symbiotic plasmid/island transfer has been proven both in the field and in the lab,. However, horizontal gene transfer cannot solely account for the wide biodiversity of rhizobia, since only a few recipient bacteria—phylogenetically close to existing rhizobia –—turned into nitrogen-fixing legume symbionts. Which phylogenetic, genetic, or ecological barriers restrict evolution of symbiotic properties and how these barriers are overcome have not been investigated so far.

2Experimental evolution coupled with genome resequencing is a powerful approach to address the evolution of rhizobia. Ralstonia solanacearum and Cupriavidus taiwanensis are plant-associated β-proteobacteria with drastically different lifestyles. R. solanacearum is a typical root-infecting pathogen of over 200 host plant species. It intercellularly invades root tissues and heavily colonizes the vascular system, where excessive production of extracellular polysaccharides blocks water traffic, causing wilting,. Cupriavidus taiwanensis is the major nitrogen-fixing symbiont of Mimosa spp. in Asia, (see Figure 1A). Due to their phylogenetic and genomic distance (Figure S1), C. taiwanensis and R. solanacearum are ideally suited to act as symbiotic gene provider and recipient, respectively, in experimental evolution.

3Here, we report on the experimental evolution of R. solanacearum carrying the symbiotic plasmid of C. taiwanensis into Mimosa-nodulating and -infecting symbionts. Two types of key adaptive mutations are described that are crucial for the transition from pathogenicity to mutualism. One allows nodulation to occur, whereas the other allows intracellular infection of plant cells, a very rare event in plant-associated bacteria.

4To generate our starting material, we transferred the 0.55-Mb symbiotic plasmid pRalta of C. taiwanensis LMG19424 into R. solanacearum strain GMI1000, generating the Ralstonia chimeric strain CBM124. pRalta carries nitrogen-fixation genes and a full complement of nodulation genes required for the synthesis of lipochitooligosaccharide Nod factors (NFs) that trigger the plant developmental program of nodule organogenesis. Nevertheless, CBM124 was unable to nodulate the C. taiwanensis legume host Mimosa pudica and retained the pathogenic properties of R. solanacearum, i.e., pathogenicity on Arabidopsis thaliana and hypersensitive response (HR) induction on tobacco (Figure S2). Note that M. pudica is not a host plant for R. solanacearum. Several lines of evidence indicated that CBM124 had a symbiotic potential that, for an unknown reason, could not be expressed. First, a nodB-lacZ transcriptional fusion was induced by the nod-inducer luteolin in a similar way in CBM124 and in C. taiwanensis (Table S1). Second, mass spectrometry analysis demonstrated that CBM124 produced NFs structurally identical to those of C. taiwanensis (Figure S3). Third, CBM124 induced root hair proliferation and deformations on M. pudica, typical of those induced by NFs (see below), indicating that CBM124-produced NFs were active.

5To isolate clones expressing symbiotic potential, we took advantage of specific traits of the rhizobium–legume symbiosis, (i) legume plants act as a trap by selecting rare, nodulation-proficient mutants in an otherwise non-nodulating population, (ii) a single bacterium enters and multiplies within the nodule, which implies that a rare nodulation-conferring mutation in a population is rapidly fixed, and (iii) nodulation, infection, and nitrogen fixation, are phenotypically clear-cut symbiotic stages. Both the original chimera CBM124 and a gentamicin-resistant derivative, CBM124GenR, were used to repeatedly inoculate sets of ca. 500 M. pudica seedlings grown in nitrogen-free conditions, as previously described. Whereas no nodules were obtained using CBM124 as an inoculum, three nodules, which appeared 3–4 wk after inoculation, were recovered from three independent CBM124GenR inoculation experiments. One bacterial clone was isolated from each nodule, generating CBM212, CBM349, and CBM356. These three clones nodulated M. pudica with different kinetics and efficiencies (Figure 1). Their nodulation ability was, however, reduced relative to C. taiwanensis (Figure 1D and Figure S4), and all three clones were unable to fix nitrogen (Fix−).

6We re-sequenced the three experimentally evolved clones as well as their immediate ancestor, CBM124GenR, using paired-end Illumina/Solexa sequencing technology (http://www.illumina.com/). Sequence data were mapped to the reference genome (6.37 Mb) based on the known genome sequences of R. solanacearum GMI1000 and C. taiwanensis LMG19424, and analyzed using the SNIPER software (S. Cruveiller and C. Medigue, unpublished data). We identified indels, SNPs (single nucleotide polymorphisms), and large deletions in the evolved clones relative to the CBM124GenR ancestor (Table S2). Among them, we focused on a large deletion as well as three SNPs that affected the HrpG-controlled virulence pathway in all three clones (Table 1). We confirmed the deletion and the SNPs by PCR amplification and Sanger resequencing.

7The ca. 33-kb deletion (Rsp0128–Rsp0154) of the R. solanacearum chromosome 2 removed 27 genes, including the pme gene coding for a pectin methylesterase involved in virulence and genes encoding a putative type II secretion system. This deletion was reconstructed in the chimera CBM124 by using the cre-lox system (see Material and Methods). The resulting strain did not nodulate M. pudica, indicating that this deletion either was not adaptive or alone could not account for nodulation. This region probably corresponds to an unstable region of the genome.

8The regulatory protein HrpG controls the expression of many virulence determinants in R. solanacearum. These include a type III secretion machinery (T3SS) and associate effector proteins that are regulated via the intermediate regulator hrpB as well as a large ensemble of genes that are modulated by hrpG in an unidentified circuitry. A stop mutation in the hrcV gene, which encodes a structural inner membrane protein at the base of the T3SS apparatus, was observed in CBM356, whereas both CBM212 and CBM349 harboured a stop mutation in the master regulator hrpG gene itself (Table 1). Consistently, all three clones exhibited a typical T3SS-defective phenotype, i.e., loss of HR induction on tobacco leaves (Figure S2). Although C. taiwanensis also possesses a T3SS of unknown function, it is not located on pRalta, thus ruling out the possibility that the impact on nodulation of the R. solanacearum virulence pathway was due to a modulation of indigenous C. taiwanensis T3SS. To assess the possible role of hrcV and hrpG gene inactivation in M. pudica nodulation, we inactivated the hrcV and hrpG genes in the original Ralstonia chimeric strain CBM124. Both CBM125 (hrcV) and CBM664 (ΔhrpG) were indeed found to nodulate M. pudica. Nonpolar disruption of hrcS, another T3SS structural gene, as well as independent hrpG inactivation by site-directed Tn5 mutation in the CBM124 background further confirmed the role of the T3SS and the hrpG gene in nodulation of M. pudica. The hrcV and hrpG mutants had symbiotic behaviours similar to that of the hrcV and hrpG evolved clones, respectively (Figure 1C and 1D).

9Like most rhizobia, C. taiwanensis invades Mimosa roots by means of transcellular infection threads (ITs), which are initiated from microcolonies entrapped within the curled root hairs known as shepherd's crooks (Figure S5). Later on, ITs elongate into emerging nodules delivering bacteria into the plant cells. Each infected cell houses thousands of symbiosomes composed of internalized bacteria (called bacteroids) surrounded by plant-derived peribacteroid membranes. Mature Mimosa nodules induced by C. taiwanensis have the typical histology of indeterminate nodules, i.e., a single distal persistent meristem and peripheral vascular bundles (Figure S5). The ancestral chimeras CBM124 and CBM124GenR promoted root hair proliferation and deformations as well as shepherd's crooks. However, these chimeras showed a clear defect in IT initiation and elongation (Figure 2A and Figure S6). In contrast, well-elongated ITs were observed with the hrcV mutant CBM125 (Figure 2B). Nodules formed, which displayed the typical nodule structure (Figure S7), although often of irregular shape compared to those induced by C. taiwanensis. However the hrcV mutant only partially and extracellularly invaded the nodule (Figure 2C and 2D). A two-step inoculation experiment using differently labelled (gfp and lacZ) strains confirmed that extracellular bacteria inside nodules originated from ITs and did not result from intercellular penetration of bacteria from the nodule surface. A necrotic dark brown zone around which bacteria were distributed was often observed in the distal part of the infected zone of the nodules (Figure S7). Plant cell wall thickening next to extracellular bacteria was also suggestive of a plant structural defence response. This could act as a physical barrier to intracellular infection. In a similar way, the evolved clone CBM356 (hrcV) was able to form elongated ITs but did not permit invasion of nodule cells (Figure S7), strongly indicating that the hrcV mutation indeed accounted for the symbiotic phenotype of CBM356. We observed that a double mutant of the PopF1 and PopF2 translocons, which do not inhibit the formation of the T3SS apparatus in R. solanacearum but are required for protein effector injection in plant cells, had a similar phenotype (Figure S4), thus suggesting that a T3SS effector(s) is involved in blocking nodulation and early infection. Most interestingly, some rhizobia have been shown to use specialized host-targeting type III or type IV secretion systems to either extend or restrict legume host range (reviewed in ). Expression of these secretion systems is coordinated to nodulation gene expression. Effectors have been identified that can either be rhizobium specific or pathogen related. They have been proposed to modulate host (signalling) pathways, including plant-defence reactions triggered by the presence of infecting rhizobia. Because R. solanacearum has more than 70 effectors, identification of the effector(s) responsible for blocking nodulation requires further work. Either nodulation is inhibited by effector-triggered immunity or a T3SS effector(s) specifically interferes with the NF-signalling pathway.

10The hrpG mutant of CBM124 (CBM664), as well as the hrpG evolved clones CBM212 and CBM349, formed nodules on M. pudica that looked similar to those induced by C. taiwanensis (Figure S8). In young nodules, plant cells were massively intracellularly invaded (Figure 3A and 3B, and Figure S8), although the infected zone was restricted, compared to N2-fixing nodules formed by C. taiwanensis. Intracellular bacteria were surrounded by a peribacteroid membrane forming typical symbiosomes (Figure 3C). Nodules, however, showed early signs of degeneration generally 3 wk postinoculation, i.e., loss of cell-to-cell contact, cytoplasmic structure desegregation of nodule cells and degradation of the internalized bacteria (Figure S8). A few extracellular bacteria were found in nodules formed by the hrpG chimeric mutant and CBM212 and CBM349 clones (Figure 3B), which is never seen with C. taiwanensis. In these cases, no plant cell wall thickening could be observed in proximity to extracellular bacteria, suggesting that they did not induce plant defence reactions. To summarize, hrpG mutants and evolved clones were able to intracellularly invade nodule cells, contrary to hrcV mutants, although bacteroids were impaired for long-term maintenance. The regulatory gene hrpG thus controls one or several T3SS-independent functions interfering with plant cell entry. In plant-associated bacteria, massive intracellular infection is restricted to nodule bacteria. Hence, there is a paradox between the rarity of intracellular infection in plants and the ease with which this trait was acquired by a strictly extracellular pathogen. Mechanisms of plant cell entry in C. taiwanensis and in rhizobia in general are largely unknown, although it has been established that surface polysaccharides play a key role in host invasion. Identification of the gene(s) downstream of hrpG controlling intracellular infection should shed light to this key, but still obscure, step of the symbiotic interaction.

11How rhizobia have emerged is a fascinating, but so far only partly documented, question. Although pioneering work 15 y ago established the role of lateral transfer in rhizobia evolution,, we and others, have observed that in many instances, transfer of symbiotic loci did not increase symbiotic competence. Here, we show that a recipient genome—that is not immediately converted to a rhizobium upon transfer of a symbiotic plasmid—could rapidly evolve two specific symbiotic traits, i.e., nodulation and intracellular infection, under plant selection pressure. Although in our case, nitrogen fixation—and hence mutualism—was not achieved and evolved clones could be considered as cheaters, evolution of nodulation and infection capacities is the first step in the evolutionary process of reciprocal cooperation. Extant rhizobial lineages diverged long before they acquired symbiotic properties, i.e., after legumes appeared on earth 60 million years ago. Our results show that adaptive genomic changes indeed allow effective dissemination of symbiotic traits over large phylogenetic and ecological distances. The fact that a single gene played a major role in the shift from extracellular pathogenesis to endosymbiosis reinforces previous reports that global regulators are preferred targets for evolution and supports fluid boundaries between parasitism and mutualism.

12Our knowledge of the rhizobium–legume symbiosis mainly comes from gene inactivation studies. Although a gain-of-function approach was first initiated ca. 25 y ago on Agrobacterium,, and used thereafter,, the experimental evolution approach we describe here is novel, as it consists of the progressive and dynamic acquisition of symbiotic ability under plant selection pressure. Evolved clones gained symbiotic traits to different degrees, allowing for a future fine dissection of unexplored aspects of nodulation and intracellular infection. Serial in planta passages using the nodulating clones described here as ancestors should allow improvement of their symbiotic capacities, i.e., bacteroid maintenance and possibly nitrogen fixation. Other symbiotic stages, such as rhizosphere colonization, host specificity of nodulation, and nitrogen fixation, could similarly benefit from coupled experimental evolution and genome resequencing approaches.

13Bacterial strains and plasmids used in this work are listed in Tables 2 and 3. C. taiwanensis strains were grown at 28°C on TY medium supplemented with 6 mM CaCl2 or quarter-strength minimal medium (MM) supplemented with 10 mM disodium succinate and vitamin solution (1 µg/ml nicotinic acid, 1 µg/ml thiamine hydrochloride, 1 µg/ml pyridoxine hydrochloride, 100 µg/ml myo-inositol, 1 µg/ml calcium pantothenate, 1 µg/ml riboflavin, 1 µg/ml ascorbic acid, 1 µg/ml folic acid, 1 µg/ml cyanocobalamin, 1 µg/ml D-biotin). R. solanacearum strains were grown at 28°C on rich BG medium or MM supplemented with 28 mM glucose. Antibiotics were used at the following concentrations (in micrograms per millilitre): streptomycin 600, spectinomycin 40, trimethoprim 100, tetracycline 10, gentamicin 25, chloramphenicol 50 for E. coli and 200 for C. taiwanensis, and kanamycin 50 for E. coli, and 30 for R. solanacearum.

14Transfer of pRalta to R. solanacearum was performed in three consecutive conjugation steps. Step 1. C. taiwanensis CBM832 was randomly transposon mutagenised using pMH1801 possessing the Tn5-B13S transposon which carries the mob site (oriT), an npt-sacB-sacR cassette and Tet-resistance. Step 2. Mutants were selected on TY supplemented with Tet and Str, and the helper plasmid, RP4-7, was individually introduced into each C. taiwanensis mutant. Step 3. C. taiwanensis::Tn5-B13S mutants carrying RP4-7 were then conjugated with R. solanacearum. Transconjugants were selected on MM supplemented with glucose and Tet. One Tn5-B13S mutagenised C. taiwanensis clone, CBM61, was successful in producing Tet-resistant R. solanacearum transconjugants. A selected transconjugant, CBM62, was verified as R. solanacearum containing pRalta by 16SrDNA and nifH gene amplification, and a seemingly intact pRalta was confirmed by a modified Eckhardt gel analysis. The Tn5-B13S insertion in pRalta of CBM62 was found located within a putative transposase (see DNA Manipulation), and thus had not disrupted any gene essential for symbiosis, as confirmed by nodulation tests and microscopic observation of the mutagenised C. taiwanensis strain CBM61 used as donor for pRalta transfer. The Tn5-B13S, which contains sacRsacB genes that might interfere with plant tests, was exchanged in CBM62 with a trimethoprim (Tri) resistance cassette (see DNA Manipulation), giving rise to the Ralstonia chimeric strain GMI1000(pRalta::Tri), or CBM124.

15The ancestral strain CBM124GenR was obtained by natural transformation of CBM124 with genomic DNA from the R. solanacearum GRS412 strain (containing the GenR plasmid pCZ367 inserted in the Rsp1236 gene). Correct insertion of pCZ367 in CBM124GenR was verified by using a primer located upstream of the inactivated gene and a primer located in the lacZ gene of pCZ367. Transfer of pRalta::Tn5-B13S from CBM61 to R. solanacearum mutants was performed as indicated in step 3.

16To construct CBM351, a CBM124 derivative deleted for the Rsp0128–Rsp0154 region, PCR fragments from the Rsp0125 and Rsp0157 genes (Rsp0126, Rsp0127, Rsp0155 and Rsp0156 are transposases) were amplified using oCBM494–oCBM495 and oCBM496–oCBM497 as primers and cloned into the EcoRI/NcoI and SacI/SacII restriction sites of pCM184, respectively. The modified plasmid was introduced into CBM124 by conjugation. Transconjugants resistant to kanamycin and sensitive to tetracycline were screened. The replacement of the Rsp0126–Rsp0156 region by the kanamycin resistance cassette in strain CBM351 was verified by PCR.

17To construct CBM125, a hrcV mutant of CBM124, pRalta::Tn5-B13S, was transferred by conjugation from C. taiwanensis CBM61 to the R. solanacearum hrcV mutant GMI1694. The Tn5-B13S transposon was then replaced by the trimethoprim resistance cassette as described above.

18To construct CBM142 and CBM145, the hrcS mutation and the popF1 popF2 double mutation were introduced into CBM124 by natural transformation of CBM124 with genomic DNA from the R. solanacearum hrcS mutant GMI1596 and the popF1 popF2 double-mutant GMI1667, respectively. The presence of an inserted cassette in hrcS, popF1, and popF2 was verified by PCR.

19To construct the hrpG mutants, CBM663 and CBM664, two different methods were used. First the CBM124 strain was transformed with genomic DNA from R. solanacearum hrpG::Tn5-B20 mutant GMI1425. Transformants were selected on BG medium supplemented with trimethoprim and kanamycin. The Tn5-B20 insertion in hrpG was verified by PCR in strain CBM663. Second, PCR fragments upstream and downstream from hrpG were amplified using oCBM622–oCBM623 and oCBM624–oCBM625 as primers and cloned into the EcoRI/KpnI and SacII/HpaI restriction sites of pCM184, respectively. The resulting plasmid was introduced into CBM124 by conjugation. Transconjugants resistant to kanamycin and sensitive to tetracycline were screened. The replacement of hrpG by the kanamycin resistance cassette was verified by PCR in strain CBM664.

20Primers used for DNA amplification are listed in Table S3.

21To determine the precise location of the Tn5-B13S insertion point in pRalta of CBM61 and CBM62, tail-PCR was performed with arbitrary primer AD1 or AD4 in combination with three sequential Tn5-specific primers designed from the terminal arms of the Tn5 transposon, oCBM183, oCBM184, and oCBM185.

22For Tn5-B13S insertion exchange by TriR cassette, a 2-kb PCR fragment, corresponding to approximately 1 kb each side of the Tn5-B13S insertion point, was amplified from LMG19424 using primers oCBM196 and oCBM198 and cloned into pGEM-Teasy (Promega). The TriR cassette isolated from p34E-Tp digested by BamHI was then introduced in the BglII site of the fragment, generating pMG02. This BglII site was located only 6 bp from the Tn5-B13S insertion point in CBM62. ScaI linearized pMG02 DNA was used to transform naturally competent R. solanacearum chimeric strains containing pRalta::Tn5-B13S. The exchange of Tn5-B13S with the trimethoprim cassette was verified by establishing that the strain had lost resistance to tetracycline and could grow on 5% sucrose.

23For the construction of pCBM01, the promoter region of nodB was amplified using oCBM203 and oCBM211 as primers and cloned into pGEM-Teasy (Promega), cleaved from pGEM-Teasy with HindIII and PstI, and then directionally cloned into the same sites of the lacZ transcriptional fusion in pCZ388. pCBM01 was introduced in C. taiwanensis and R. solanacearum strains by conjugation.

24The lacZ- and gfp-derived strains were obtained by natural transformation with genomic DNA from strains GMI1485 and GMI1600, respectively.

25Sequence data production was performed by the C.E.A/IG/Genoscope (Evry). Paired-end libraries were prepared following the protocol recommended by Illumina Inc. (http://www.illumina.com). For each strain, more than 5 million paired-end reads (L = 72 bp = 2×36 bp) were generated with Genome Analyzer sequencing system, leading to a ca. 60× total coverage of the reference genome (Table S4). Taking advantage of the local production of raw sequencing data, a bioinformatic pipeline called SNiPer (S. Cruveiller and C. Medigue, unpublished data) and based on ssaha2 alignment software (Sequence Search and Alignment by Hashing Algorithm has been implemented. This pipeline allows the detection of small variations (SNPs and InDels) between a collection of short reads and a reference sequence, this latter being either a consensus produced by assemblers or a previously published one.

26SNiPer is a shell script that automatically sets the alignments parameters depending on the kind of reads (ABI-3730/454-GSFLX/Solexa/SOLiD) being used, launches the various parts of the detection pipeline, and controls for all tasks having been completed without errors. The detection of SNPs and indels is achieved in four main steps: (1) The data preparation, which consists in (i) the conversion of sequencing raw data (i.e., reads files) into Sanger Institute FastQ formatted files; (ii) the removal of duplicated reads (quite common when using Solexa platform) so as to keep exactly one copy of each read; and (iii) the split of paired-ends reads into single-end reads when required. (2) Reads mapping onto a reference molecule using the ssaha2 package. This package combines the SSAHA searching algorithm (sequence information is encoded in a perfect hash function) aiming at identifying regions of high similarity, and the cross_match sequence alignment program (http://www.phrap.org/phredphrapconsed.html), which aligns these regions afterwards using a banded Smith-Waterman-Gotoh algorithm,. (3) Based on the characteristics of reads alignments onto the reference molecule, a file containing the lists of all possible events is generated. (4) Each event is then scored so as to keep only significant ones. This score takes into account the reference base coverage (i.e., the number of reads mapping a given location) and the quality of bases of reads displaying a change at that particular location as well.

27The ca. 5 million paired-end reads were split into single reads and mapped on the reference genome (the two replicons of Ralstonia solanacearum GMI1000 [RefSeq acc. NC_003295.fna and NC_003296.fna for the chromosome and the megaplasmid respectively]+the nodulation plasmid of Cupriavidus taiwanensis LMG19424 [RefSeq acc. NC_010529.fna]) using SNiPer. Among the 10 million single reads, around 7 millions were successfully mapped, leading to an effective coverage of the three reference molecules higher than 30× (Table S4), hence warranting a reliable detection of changes. The remaining unmapped reads (3 million on average) correspond to reads that could neither be mapped unambiguously (i.e., repeat regions, insertion sequences, rDNA, etc.) nor be mapped at all (i.e., fragment of sequences not present in the references).

28Pathogenicity assays with M. pudica and Arabidopsis thaliana ecotype Col-0 were performed according to Deslandes et al.. Root inoculations used the method of cutting 2 cm from the bottom of Jiffy pot–grown plants, followed by immersion for 5 min in a suspension of bacteria grown overnight and diluted to an OD600 of 0.1 in water. R. solanacearum and derivatives were tested for the HR ability by infiltrating a bacterial culture adjusted to 108 cells/millilitre into tobacco (cultivar Bottom Special) leaf parenchyma as described previously.

29For M. pudica nodulation assay and cytology, seeds were surface sterilised and planted under sterile conditions using the tube method of Gibson as previously described, (except tubes contained Fahraeus slant agar and liquid water). For the selection of nodulating evolved clones, 107 bacteria par tube were used as inoculum. Otherwise, 104 bacteria were routinely inoculated per tube unless specified. Nitrogen fixation was estimated by visual observation of the vigour and foliage colour of 40/60-d-old plants on at least 20 plants. For reisolation of nodule bacteria, nodules were surface sterilised 10 min with 2.6% sodium hypochlorite, rinsed five times, then crushed and dilutions plated on the appropriate solid medium. For each M. pudica tube, ex planta number of bacterial generations is estimated at a maximum of 5, and in planta generation number is calculated using the formula log(number of bacteria/nodule)/log2.

30LacZ-tagged infecting bacteria were stained according to the standard procedure. Briefly, roots were fixed in glutaraldehyde 1.5% in K phosphate buffer for 30 min under vacuum condition followed by 1 h at room temperature. After washing, roots were incubated overnight with the staining solution at 28°C (0.1 M K phosphate [pH 7.4], 2 mM K ferricyanide, 2 mM K ferrocyanide, and 0.08% of X-gal in dimethylformamide). Roots were washed and used for microscopic analysis. To analyse infection of gfp-tagged bacteria, root and nodules were fixed in paraformaldehyde 3.7% in phosphate buffered saline (PBS) for 30 min under vacuum, then washed and used directly or cut for nodule sections 60-µm thick using a Leica VT1000S vibratome. Samples were observed by using a fluorescence (Zeiss Axiophot Fluorescence microscope) or confocal microscope (Leica SP2).

31For fine histological examination, nodules were fixed in glutaraldehyde (2.5% in phosphate buffer 0.1 M [pH 7.4]), osmium treated, dehydrated in an alcohol series, and embedded in Epon 812. Semithin nodule sections were observed by brightfield microscopy after staining in 0.1% aqueous toluidine blue solution and observed under a Zeiss Axiophot light microscope. Ultrathin sections were stained with uranyl acetate and observed with a Hitachi EM600 electron microscope.

32For the two-step infections, we proceeded as follows. M. pudica plants, grown as described above, were first infected with the lacZ-tagged hrcV chimeric strain. After 9 d of infection, once nodules were formed, a secondary infection was performed by using the gfp-tagged hrcV chimeric strain. Two weeks after, nodules were fixed in paraformaldehyde 3.7% as previously described and used for cytological analysis.

33Strains were grown overnight at 28°C in MM supplemented with the appropriate carbon source, vitamins, and tetracycline. Overnight cultures were then diluted to an OD600 of 0.005–0.01 in MM with tetracycline ±15 µM final concentration of luteolin and grown a minimum of 16 h until an OD600 of 0.7 was reached. The cultures were then assayed for β-galactosidase activity (Miller units) according to Miller, 1972. The β-galactosidase activities represent an average of quadruplicate samples from two separate experiments.

34NFs were produced, purified, and characterized as previously described.