Research Article

Legumes Symbioses: Absence of Nod Genes in Photosynthetic Bradyrhizobia

See allHide authors and affiliations

Science  01 Jun 2007:
Vol. 316, Issue 5829, pp. 1307-1312
DOI: 10.1126/science.1139548


Leguminous plants (such as peas and soybeans) and rhizobial soil bacteria are symbiotic partners that communicate through molecular signaling pathways, resulting in the formation of nodules on legume roots and occasionally stems that house nitrogen-fixing bacteria. Nodule formation has been assumed to be exclusively initiated by the binding of bacterial, host-specific lipochito-oligosaccharidic Nod factors, encoded by the nodABC genes, to kinase-like receptors of the plant. Here we show by complete genome sequencing of two symbiotic, photosynthetic, Bradyrhizobium strains, BTAi1 and ORS278, that canonical nodABC genes and typical lipochito-oligosaccharidic Nod factors are not required for symbiosis in some legumes. Mutational analyses indicated that these unique rhizobia use an alternative pathway to initiate symbioses, where a purine derivative may play a key role in triggering nodule formation.

Legume plants have developed symbiotic associations with specific soil bacteria, collectively referred to as the rhizobia, which allow plants to thrive and reproduce in nitrogen-poor environments. These plant-bacterial symbiotic associations typically result in the formation of root organs, termed nodules, in which the bacteria differentiate into nitrogen-fixing bacteroids. Initiation of nodule development involves molecular recognition between both symbiotic partners (14). Flavonoid molecules exuded by plant roots induce expression of bacterial nodulation (nod) genes leading to the synthesis of Nod factors, rhizobial lipochito-oligosaccharide signal molecules. Synthesis of the Nod-factor chitin oligomer backbone requires the activity of three specific enzymes, encoded by the nodABC genes, which are present in all rhizobia characterized thus far. Nod factor signal molecules are recognized by plant kinases of the LysM-RLKs family, which, in turn, initiate a developmental program in the legume host resulting in the formation of the nodule structure. The ubiquitous presence of nod genes and Nod factors in all rhizobia led to the development of a universal “lock-and-key” hypothesis (5, 6), which states that all symbiotic legumes and rhizobia have host nodulation determinants and homologs of the known nodulation genes, respectively.

The rhizobia belong to the alpha and beta subclasses of the Proteobacteria (7), and most species belong to the genera Rhizobium, Bradyrhizobium, Mesorhizobium, Sinorhizobium, and Azorhizobium. Among them, some photosynthetic Bradyrhizobium sp. strains specifically induce nodules on both the root and stem of the aquatic legume Aeschynomene (8). According to their host specificity within different species of Aeschynomene, two groups of photosynthetic Bradyrhizobium have been described (8): Group I strains contain the common nodulation genes nodABC and form nodules on all stem-nodulating Aeschynomene species (9), whereas nodulation ability in group II strains is restricted to a few species, including A. sensitiva and A. indica. The nod genes have not been detected in group II strains.

To unravel the genetic features contributing to the symbiotic properties of the stem-nodulating Bradyrhizobium, we determined the complete genomic sequences of two group II photosynthetic Bradyrhizobium strains, BTAi1 and ORS278 (10). We also performed a genome-wide screen for mutants unable to induce nodules on Aeschynomene plants.

General genome features. The photosynthetic Bradyrhizobium strains ORS278 and BTAi1 were isolated from stem nodules of two Aeschynomene species, A. sensitiva and A. indica, in Africa and North America, respectively. The ORS278 and BTAi1 strains were sequenced by Genoscope (France) and the U. S. Department of Energy (DOE) Joint Genome Institute (USA), respectively (see SOM). Manual annotation of both genomes was conducted using MaGe comparative genomic software (11).

The ORS278 genome consists of a single, circular, chromosome of 7,456,587 base pairs (bp); the BTAi1 genome contains both a large chromosome of 8,264,689 bp and a single, circular, 228,826-bp plasmid, pBTAi1. Chromosomes from ORS278 and BTAi1 are almost identical in G+C content, 65.5 and 64.9%, respectively. In contrast, the G+C content of plasmid pBTAi1 is 60.7%. The genomes of ORS278 and BTAi1 contain 6752 and 7729 predicted coding sequences (CDSs), respectively, of which 59 and 60.1% could be assigned putative functions (Table 1).

Table 1.

General features of Bradyrhizobium genomes. The distribution of CDSs for B. japonicum is derived from the original annotation (19). IS, insertion sequence.

Genome featureBradyrhizobium sp. B. japonicum
Genome size (bp) 7,456,587 8,493,515 9,105,828
G+C content (%) 65.5 64.9 64.1
Ribosomal RNA operons 2 2 1
Transfer RNAs 50 52 50
Total protein-coding genes 6,752 7,729 8,317
Assigned function (%) 59 60.1 52.3
Conserved hypothetical (%) 27.7 31.8 30.1
Hypothetical (%) 13.3 8.1 17.6
Coding DNA (%) 90.5 91.0 91.3
Plasmid no. (size in bp) 0 1 (228,826) 0
IS elements (%) 0.5 1.4 2
Pseudogenes (%) 0.3 0.6 0

Comparative genomic analyses. BLAST analyses indicate that the genomes of BTAi1 and ORS278 share more genes with B. japonicum (4002) than with their closest photosynthetic neighbor, Rhodopseudomonas palustris (2563) (12). This agrees with phylogenies created using 16S ribosomal RNA sequences (13). A cross-comparison of the two photosynthetic Bradyrhizobium genomes identified a common pool of 1343 genes, whereas 1274 and 2133 genes were unique to ORS278 and BTAi1, respectively. Most of the 1343 genes in common were homogeneously distributed on the genomes (Fig. 1), except for genes involved in photosynthesis, which were clustered in a 50-kb region. This region, designated the photosynthesis gene cluster (PGC), is common in purple photosynthetic bacteria. The organization of the PGC is identical in the BTAi1 and ORS278 strains and is highly conserved with the PGC in R. palustris. In addition, the G+C content (67%) of the Bradyrhizobium PGCs is similar to the rest of the genomes. Bradyrhizobium strains BTAi1 and ORS278 also have additional genes specialized in light perception and response, including three putative bacteriophytochromes (biliprotein photoreceptors) and two kaiBC circadian clock operons. Taken together, these data strengthen previous hypotheses that photosynthetic capacity is an ancestral trait in these Bradyrhizobium strains that was subsequently lost in most rhizobial lineages (14).

Fig. 1.

Circular representation of the Bradyrhizobium strains ORS278 and BTAi1 chromosomes. Circles, from the inside out, show (1) GC skew (G+C/G–C using a 1-kb sliding window), (2) IS elements and transposases, (3) the coordinates in Mb beginning at 0 = oriC, and (4) GC deviation (mean GC content in a 1-kb window—overall mean GC). Regions with a GC deviation less than two times the standard deviation are highlighted in red. Circles 5 and 6 show the gene content comparison between the ORS278, BTAi1, and B. japonicum USDA110 genomes by using a similarity threshold of 40% identity and a ratio of 0.8 of the length of the smallest protein: blue, backbone genes found in all three strains; red, genes present in ORS278 and BTAi1, but absent from USDA110; cyan in (ORS278) show genes present in ORS278 and USDA110, but not in BTAi1, and in (BTAi1), genes present in BTAi1 and USDA110, but not in ORS278; green, genes specific to each strain. The different genomic islands identified in both chromosomes are highlighted by number. See tables S1 and S2.

The large variation in genome sizes (1.6 Mb between ORS278 and B. japonicum, and 1 Mb between ORS278 and BTAi1), the overall low level of synteny (Fig. 2 and fig. S1), and the presence of numerous mobile genetic elements indicate that bradyrhizobial genomes are highly plastic. Indeed, the genomes of ORS278 and BTAi1 contain 21 and 29 putative horizontally acquired genomic islands (HAIs), respectively, displaying hallmarks of recent gene transfer events (Fig. 1 and tables S1 and S2). HAIs may confer functional advantages in the adaptation of these bacteria to their symbiotic or free-living ecological niches. Genes encode for important metabolic functions on HAIs, in both strains, including the following: (i) a ribulose 1,5-bisphosphate carboxylase (RuBisCo); (ii) enzymes involved in nitrogen metabolism, including urease; (iii) lipopolysaccharide (LPS)–modification enzymes; (iv) a type II secretion system; (v) a chemotaxis operon; and (vi) a multidrug efflux pump. In addition, the BTAi1 genome contains a specialized HAI that is made up of all the genes necessary for CO2 fixation, along with a hydrogenase gene cluster. Remarkably, this strain harbors three RuBisCOs, which assimilate CO2, and three uptake hydrogenase complexes, one of which is plasmid-borne; the other two are in an HAI and close to the PGC. The presence of these elements suggests that these bacteria have an exceptional ability to fix CO2 partly due to reducing power furnished by the photosynthetic activity and partly by the uptake hydrogenase enzymes, which potentially scavenge hydrogen produced during nitrogen fixation.

Fig. 2.

Synteny plot between the three Bradyrhizobial chromosomes. This line plot (11) was obtained by using synteny results between Bradyrhizobium sp. BTAi1 and Bradyrhizobium sp. ORS278, as well as results obtained from comparison of Bradyrhizobium sp. ORS278 and Bradyrhizobium japonicum USDA110. Synteny groups containing a minimum of five homologous genes are drawn in green for colinear regions, and in red for inverted regions.

No symbiotic genes (nod, nif, or fix) were found on islands in the chromosomes of ORS278 and BTAi1 or on plasmid pBTAi1 (tables S1 and S2). Although the nif and fix genes are clustered in a 45-kb region of the BTAi1 and ORS278 genomes, there is no evidence that this region was acquired by lateral transfer. This is in contrast to all rhizobial genomes characterized to date (1521), which have nodulation and nitrogen fixation genes clustered either on plasmids or in large chromosomal symbiotic islands, for example, as observed in B. japonicum USDA110 (see fig. S2).

Lack of canonical nodulation genes. BLAST analyses indicate that neither the BTAi1 nor the ORS278 genome contain homologs of NodA (acyl transferase) or NodC (oligomerization of N-acetyl-glucosamine), two of the three enzymes essential for the synthesis of lipochito-oligosaccharidic Nod factors in rhizobia. Although a homologous gene in both strains displayed a moderate level of amino acid identity (33 to 36%) to NodB from Rhizobium galegae, this homology was limited only to the polysaccharide deacetylase domain of NodB (Table 2), a motif found in other enzymes, including chitin deacetylases and endoxylanases. The low identity that these CDSs have to NodB and the absence of nodA and nodC gene homologs indicate that the canonical nodABC genes are absent in both the BTAi1 and ORS278 genomes. CDSs displaying some similarity to other Nod proteins are present in the genomes of both BTAi1 and ORS278 (Table 2); however, these homologs are well conserved in other nonsymbiotic prokaryotes.

Table 2.

Coding sequences in the ORS278 and BTAi1 genomes showing similarities with known nodulation (Nod) proteins. Absent refers to no hits with ≥30% amino acid identity using the BLASTP algorithm. Brado and BBTa refer to Bradyrhizobium sp. strains ORS278 and BTAi1, respectively, CDS with best BLASTP hits. Percent amino acid similarity to named species of the BLASTP hit. Acc no., GenBank accession numbers of the Nod protein. E value of the BLASTP hit. Abbreviations: R., Rhizobium sp.; M., Mesorhizobium sp.; R. leg., Rhizobium leguminosarum bv. viciae; and B., Bradyrhizobium sp.

Nod proteinBradyrhizobium sp. strains
CDSSimilarity (%)Acc. no.E valueCDSSimilarity (%)Acc no.E value
NodA Absent Absent
NodB Brado4564 33% to R. galegae P50354 2e-16 BBta4792 36% to R. galegae P50354 2e-16
NodC Absent Absent
NodD Brado3695 39% to R. Q53061 1e-54 BBta1932 38% to R. Q53061 1e-53
NodE Absent BBta0068 31% to R. leg. P04684 1e-47
NodG Brado3311 66% to M. P72332 4e-86 BBta3818 65% to M. P72332 2e-85
NodI Brado0517 39% to R. tropici Q933C0 5e-37 BBta7658 39% to R. tropici Q933C0 4e-35
NodJ Absent Absent
NodL Brado2093 42% to R. leg. P08632 2e-11 Absent
NodM Brado3763 78% to B. Q9AQ10 1e-277 BBta4166 76% to B. Q9AQ10 1e-269
NodP Brado5192 75% to B. elkanii BAB55898 1e-133 BBta0328 73% to R. P72338 2e-125
NodQ Brado5193 66% to B. elkanii BAB55899 5e-236 BBta0327 55% to R. 007309 2e-195

Consistent with genomic information revealed by our analyses, two other observations support our statement that ORS278 and BTAi1 are unique among rhizobia. First, infection of A. sensitiva by ORS278 (fig. S3) revealed the absence of Nod factor–mediated root-hair deformation that classically precedes the entry of rhizobia into legumes. Despite this, the nodules induced by ORS278 (14) on the stems and roots of Aeschynomene displayed the functional characteristics of classical nodules induced by other rhizobia, such as the synthesis of leghemoglobin, nitrogenase activity, and transfer of fixed N2 into the plant. Second, the production of classical Nod factor–like compounds by strain ORS278 could not be shown when genistein was used as a nod-gene inducer and with a Nod factor isolation method currently shown to work with bradyrhizobia (22, 23). To show more definitively that classical Nod factors are not involved in the A. sensitiva symbiosis, we examined the nodulation phenotype of a nodB mutant of the broader host-range Bradyrhizobium sp. strain ORS285 (group I). This strain nodulates both A. sensitiva and A. afraspera, whereas the ORS278 and BTAi1 strains do not nodulate A. afraspera. As shown in Fig. 3, this mutant failed to induce root and stem nodules on A. afraspera, although it maintained its ability to form nitrogen-fixing nodules on roots and stems of A. sensitiva with an efficiency similar to that of the wild-type strain. These observations and results demonstrate that the canonical common nodulation genes and, thus, typical lipochito-oligosaccharidic Nod factors are not required for the symbiotic interaction of the photosynthetic bradyrhizobia with A. sensitiva.

Fig. 3.

Efficiency of stem nodulation by Bradyrhizobium sp. strains ORS278, ORS285, and the nod gene deletion mutant ORS285ΔnodB (285Δnod) on A. afraspera (A) and A. sensitiva (B). The NI refers to the noninoculated control. The ORS285ΔnodB mutant was obtained by homologous recombination as described (14) after insertion of the lacZ-KanR cassette (pKOK5) in the unique Xho I site of nodB. Plants were inoculated and cultivated as described (14).

The absence of nodABC genes in Bradyrhizobium strains ORS278 and BTAi1 raises the question of the nature of the bacterial signal used to induce nodule formation on A. indica and A. sensitiva plants. Analysis of the ORS278 and BTAi1 genomes revealed several other genes involved in plant-microbe symbiotic and pathogenic interactions (table S3). These include genes involved in the following: (i) synthesis or degradation of phytohormones (24), (ii) modification of O-antigen (or LPS) (25, 26), and (iii) biosynthesis of exopolysaccharides (5, 27). Nevertheless, the symbiotic role of these candidate genes remains highly speculative, and we cannot rule out the possibility that the mechanism of interaction involves genes of unknown function.

Nodulation defective mutants. In order to identify new genes involved in the interaction between the group II photosynthetic bradyrhizobia and A. sensitiva and A. indica plants, a library of ORS278 transposon (Tn5) mutants was screened for strains unable to induce nodule formation on A. sensitiva. Of the 9500 mutants tested, 27 were found to be severely defective in symbiosis, most eliciting only a few pseudonodules on a small number of plants (Table 3). Some plants displayed an apparent nodulation-minus phenotype, and microscopic examination often revealed a small number of pseudonodules. No completely nodulation-deficient mutant could be found; we tentatively attribute this phenotype to one or more factors: (i) there are redundant genes controlling this initial step in the symbiotic process; (ii) the appropriate mutation may be fatal in this bacterium; or (iii) the mutagenesis procedure used was not saturating, and an essential gene may have been missed.

Table 3.

Bradyrhizobium sp. strain ORS278 mutants affected in symbiosis with Aeschynomene sensitiva. Minitransposon mTn5-GNm (43), containing a gusA reporter gene and the nptII gene encoding neomycin resistance, was used for the construction of the insertion mutants. After mutagenesis, the mutants were selected on YM-modified rich medium (14) containing kanamycin (50 μg/ml) and nalidixic acid (35 μg/ml). Each of the 9500 mutants was initially individually tested on one A. sensitiva plant. Mutants having an alteration in their nodulation efficiency were retested using three plants per mutant, and subsequently tested again in a third analysis with six additional plants. The localization of the site of Tn5 insertion in symbiotically deficient mutants was ascertained as previously describe (44). Mutant number describes the number of independent mutants identified in the corresponding CDS. Phenotype shows the number of plants not nodulated (–); number of plants having pseudonodules (PN); and number of plants having classical nodules (+).

CDSGeneFunctionMutant no.Phenotype
Amino acid auxotrophs
Brado0374 leuC Leucine biosynthesis 5 9-, 1PN
Brado0365 leuD Leucine biosynthesis 2 8-, 2PN
Brado3572 hom (metL) Homoserine biosynthesis 1 4-, 2PN, 4+
Brado0430 proA Proline biosynthesis 1 9-, 1PN
Brado6238 gltB Glutamate biosynthesis 2 8-, 2PN
Nitrogen regulation system
Brado0530 glnD Nitrogen regulation system 1 8-, 2PN
Pyrimidine or purine biosynthesis
Brado0160 pyrF Pyrimidine biosynthesis 1 4-, 6PN
Brado3280 purF Purine biosynthesis 1 5-, 5PN
Brado6088 purD Purine biosynthesis 2 5-, 5PN
Brado2785 purL Purine biosynthesis 3 6-, 4PN
Brado3344 purM Purine biosynthesis 1 8-, 2PN
Brado1206 purE Purine biosynthesis 1 2-, 8PN
Brado2830 purB Purine biosynthesis 2 9-, 1PN
Brado0338 purH Purine biosynthesis 1 9-, 1PN
Brado5935 purA Purine biosynthesis 1 8-, 2PN
Brado3165 guaB Purine biosynthesis 1 2-, 8PN
Brado3317 gmk Purine biosynthesis 1 7-, 3PN

The site of Tn5 insertion in each of the 27 symbiosis-deficient mutants was determined, and none were found to have a mutation in a CDS that could mimic the action of the NodABC proteins. Four classes of mutants were distinguished. Class I comprised 11 independent amino acid–requiring (AA) auxotrophic mutants. Numerous AA auxotrophs of rhizobia were found to be defective in symbiosis (2831), and generally, these have symbiotic capacity rescued with the addition of the missing amino acid(s). This suggests that the absence of symbiosis stems from a growth deficiency, rather than an alteration in the infection process (30, 32). The single class II mutant had an insertion in glnD, a uridyltransferase/uridyl-removing enzyme constituting the sensory component of the nitrogen regulation (ntr) system, similar to a Rhizobium leguminosarum glnD mutant impaired in cell viability (33). Similarly, class III contained a single mutant with an insertion in a key enzyme (PyrF, orotidine 5′-phosphate decarboxylase) of the pyrimidine biosynthesis pathway.

In contrast, class IV contained 14 mutants that were affected in the biosynthesis of purines. The mutated genes (10) were scattered throughout the ORS278 genome, which indicated that there was no Tn5 integration hotspot and that purines play a pivotal role in the symbiotic process. Symbiotically defective purine auxotrophs of several Rhizobium species have been previously reported (3436). The intriguing finding that purine intermediates, such as 5-aminoimidazole-4-carboxamide riboside (AICAR), rather than purine itself, can partially restore nodulation deficiency in purine mutants of Rhizobium leguminosarum or R. etli suggested that a purine precursor might act as a signal molecule during infection (35, 37).

To further examine this, a ORS278 purL mutant, which only produced a few pseudonodules, was evaluated for its ability to normally nodulate A. sensitiva in the presence of the purine biosynthesis precursors AICAR, inosine monophosphate (IMP), and adenosine monophosphate (AMP) (fig. S4). In the absence of precursors, the mutant bacterium proliferated on the root surface, which indicated that the host plant can supply the required nutrient(s) for mutant growth. In addition, although cytological analysis showed that pseudonodules elicited by the purL mutant contained bundles of infection threads progressing intercellularly around the meristematic zone, the bacteria were not released into meristematic cells. In contrast, the symbiotic ability of the mutants was partially restored when the plant culture medium was supplemented with AICAR, IMP, or AMP at 0.5 mM. Treatment of plants with these compounds allowed the mutant to proliferate in meristematic cells, and plants had normal size nodules.

Similar observations were obtained using purH and purB mutants, except that the partial rescue of the purH mutant was observed only in the presence of IMP and AMP, and the nodulation phenotype of the purB mutant was only rescued in the presence of AMP. Taken together, these data suggest that a purine derivative, which may be a cytokinin-like molecule, plays a key role in the formation of nodules on A. sensitiva, and presumably A. indica, by the photosynthetic bradyrhizobia. This hypothesis is consistent with previous observations that (i) several Bradyrhizobium strains were shown to produce cytokinin compounds (38, 39); (ii) a Nod factor–deficient Sinorhizobium strain carrying a constitutive transzeatin secretion (tzs) gene involved in cytokinin production induced nodule-like structures on alfalfa plants (40); and (iii) a plant cytokinin-receptor was shown to play a key role in the activation of nodule organogenesis in Lotus japonicus (41, 42). We were unable to identify a homolog to the tzs gene in the ORS278 or BTAi1 genomes; this suggests that these two strains may use a different pathway for cytokinin production. This is likely the case, as B. japonicum USDA110, which produces cytokinin molecules (39), also lacks a tzs homolog.

Our data show that nodulation of some legumes by rhizobia occurs in the absence of the nodABC genes and lipochito-oligosaccharidic Nod factors, which indicates that other signaling strategies can trigger nodule organogenesis in some legumes. The nodABC genes have been identified in all the rhizobia thus far characterized, consisting of hundreds of symbiotically distinct strains isolated worldwide. This indicates that the nodulation of legumes in the absence of canonical nod genes is likely an unusual phenomenon. However, such a discovery is highly significant and raises several interesting questions: Is the first recognition step initiated by Nod factor simply bypassed in some legume plants, especially those allowing infection via crack entry? What is the signal molecule(s) directly triggering nodule organogenesis in Aeschynomene? Is there any relation between the molecular mechanism used by photosynthetic bradyrhizobia and other symbiotic organisms, such as Frankia which induces nodulation on nonleguminous plants? Our discovery should stimulate research on plant-bacteria interactions by providing insight into the diversity and evolution of symbiotic strategies used by both partners.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Tables S1 to S3

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article