Diversity analyses of Aeschynomene symbionts in Tropical Africa and Central America reveal that nod-independent stem nodulation is not restricted to photosynthetic bradyrhizobia


  • Present addresses: IMEP, Université de Provence, Marseille, France; LMBM, Université Mohammed V, Rabat, Maroc.

E-mail lucie.miche@univ-provence.fr; Tel. (+33) 4 91 28 91 02; Fax (+33) 4 91 28 86 68.


Tropical aquatic legumes of the genus Aeschynomene are unique in that they can be stem-nodulated by photosynthetic bradyrhizobia. Moreover, a recent study demonstrated that two Aeschynomene indica symbionts lack canonical nod genes, thereby raising questions about the distribution of such atypical symbioses among rhizobial–legume interactions. Population structure and genomic diversity were compared among stem-nodulating bradyrhizobia isolated from various Aeschynomene species of Central America and Tropical Africa. Phylogenetic analyses based on the recA gene and whole-genome amplified fragment length polymorphism (AFLP) fingerprints on 110 bacterial strains highlighted that all the photosynthetic strains form a separate cluster among bradyrhizobia, with no obvious structuring according to their geographical or plant origins. Nod-independent symbiosis was present in all sampling areas and seemed to be linked to Aeschynomene host species. However, it was not strictly dependent on photosynthetic ability, as exemplified by a newly identified cluster of strains that lacked canonical nod genes and efficiently stem-nodulated A. indica, but were not photosynthetic. Interestingly, the phenotypic properties of this new cluster of bacteria were reflected by their phylogenetical position, as being intermediate in distance between classical root-nodulatingBradyrhizobium spp. and photosynthetic ones. This result opens new prospects about stem-nodulating bradyrhizobial evolution.


Symbiotic interactions between legumes and rhizobia usually lead to the formation of nitrogen-fixing nodules on the roots of plant hosts. However, some tropical aquatic legumes – including many Aeschynomene species – also produce nodules on their stems at the sites of dormant adventious root primordia. This unusual property is shared only with a few other species of the genera Discolobium, Neptunia and Sesbania (Boivin et al., 1997). All these legumes have remarkable potentials as green manure for crop production under water logging conditions (Becker et al., 1990). The mode of infection of Aeschynomene spp. is rather primitive, as it does not occur via root-hair curling but by direct invasion of the epidermal cracks created by emerging lateral roots (Alazard and Duhoux, 1990). This ‘crack-entry’ infection is shared by other tropical legumes, including the closely related Arachis hypogaea (Boogerd and Rossum, 1997).

Bacterial microsymbionts isolated from Aeschynomene stem nodules exhibit the uncommon property to possess a photosynthetic activity (Eaglesham and Szalay, 1983; Evans et al., 1990). Several studies stressed the importance of photosynthesis for bacterial symbiotic lifestyle: a mutation in photosynthetic genes in the model strain ORS278 lead to a half decrease in stem-nodule numbers and consequently plant growth improvement, compared with inoculation with the wild type (Giraud et al., 2000). The work of Evans and colleagues (1990) also indicated that bacterial photosynthetic activity provides energy for nitrogenase activity inside nodules present on the plant stem.

From a phylogenetic point of view, stem-nodulating Aeschynomene symbionts belong to the genus Bradyrhizobium (Wong et al., 1994), but form a separate sub-branch distinct from the non-photosynthetic species, Bradyrhizobium japonicum and B. elkanii (Molouba et al., 1999). Besides photosynthetic activity, stem-nodulating bradyrhizobia possess several other phenotypic characteristics (e.g. diazotrophic activity or polysaccharide composition) that distinguish them from the other rhizobia (Ladha and So, 1994). Their divergence is thus expected to imply the presence of many more genes besides those involved in photosynthesis.

Not all Aeschynomene species and their symbiotic bacteria share the atypical properties described above. In a large study of 20 different Aeschynomene spp., Alazard (1985) and Molouba and colleagues (1999) identified three different cross-inoculation (CI) groups according to their associated symbionts (Table 1). Aeschynomene species belonging to CI-group 1 (e.g. A. elaphroxylon) are only nodulated on their roots by non-photosynthetic and weakly specific bradyrhizobia. Stem nodulation is thus restricted to CI-groups 2 and 3. Bacteria (e.g. strains ORS278 and BTAi1) nodulating Aeschynomene of the CI-group 3 (e.g. A. indica) are all photosynthetic and cannot nodulate plants from CI-groups 1 and 2. Bacteria nodulating Aeschynomene of the CI-group 2 (A. afraspera) are either photosynthetic, as with ORS285, or non-photosynthetic. But among these bacterial isolates, only the photosynthetic ones can nodulate plant stems (Molouba et al., 1999). Noticeably these stem-nodulating photosynthetic bradyrhizobia are also able to form effective nodules on CI-group 3 plants.

Table 1. Aeschynomene–Bradyrhizobium spp. cross-inoculation (CI) group properties (after Alazard, 1985; Giraud et al., 2007).
CI-groupsBacterial propertiesExamples of bacterial symbiontsExamples of Aeschynomene species
  1. Only representative bacterial strains but all Aeschynomene species analysed in this study are presented on the diagram. Each block indicates plant host range of non-photosynthetic (white) and photosynthetic (shaded) bacterial strains.

1• Non-photosynthetic strainsORS301A. elaphroxylon
• Root nodulating  
• Some also nodulate CI-group 2 plants  
2• Cannot nodulate CI-group 1 plants A. afraspera
• Include:  
a. Non-photosynthetic strains that only nodulate rootsa. ORS354 
b. Photosynthetic strains that nodulate roots and stems. They moreover nodulate CI-group 3 plantsb. ORS285 
3• Photosynthetic strainsORS278A. indica
• Nodulate roots and stems  
• Cannot nodulate CI-groups 1 and 2 plants  
nodABC-independent nodulation  

Thus, some photosynthetic bradyrhizobia (ORS285) have a rather wide host range and can nodulate all stem-nodulated Aeschynomene, whereas others (e.g. ORS278 and BTAi1) are only able to nodulate plants belonging to CI-group 3. Recent studies have demonstrated that such a difference between the two groups of strains was due to the presence, or lack, of classical nod genes. Whereas CI-group 2 bradyrhizobia harbour these nodulation genes (Chaintreuil et al., 2001), bradyrhizobia exclusively nodulating Aeschynomene of CI-group 3 do not (Giraud et al., 2007). This result was surprising as it was the first time that rhizobia lacking classical nod genes were described.

Giraud and colleagues (2007) demonstrated that an isogenic nod mutant of strain ORS285 became impaired in A. afraspera nodulation, but could still nodulate A. indica. This result implied the coexistence of two nodulation mechanisms in the CI-group 2 photosynthetic strains: a classical nod-dependent one, and a nod-independent one with CI-group 3 plants such as A. indica.

A first step towards a better understanding of the mechanisms involved in CI-group 3 plant nodulation was the full-length sequencing of the two model strains ORS278 and BTAi1. Comparative genomic analyses revealed a high plasticity within the two-bradyrhizobial genomes, as reflected by large variations in genome sizes and composition (Giraud et al., 2007).

Accordingly, a phylogenetic analysis of several Aeschynomene symbionts originating from West Africa highlighted a wide genetic diversity among photosynthetic bradyrhizobia (Willems et al., 2000). All stem-nodulating bradyrhizobia (either from CI-group 2 or from CI-group 3) however grouped together within a single clade (Molouba et al., 1999).

To get a better insight of photosynthetic bradyrhizobial evolution, our goal was to analyse further to what extent nod-independent stem nodulation is widespread compared with the nod-dependent one and if a structuring of bacterial strain diversity according to their CI-group, plant hosts or geographical origin could be revealed. Since the two main areas of legume diversification, including the dalbergioid/aeschynomenoid clade, are concentrated within the lowland tropical forests of America and Africa (Lavin et al., 2000; Doyle and Luckow, 2003), we decided to broaden the previous work made on African strains by sampling various Aeschynomene species in central America.

The diversity of stem-nodulating Aeschynomene symbionts originating from these two continents was compared in terms of both genome composition and phylogeography. Phylogenetic analyses based on the recA housekeeping gene and whole-genome amplified fragment length polymorphism (AFLP) fingerprints were carried out on 56 bacterial strains isolated from nodules collected on a variety of native Aeschynomene species in Mexico, French Guiana and Guadeloupe, and were compared with a collection of 51 strains previously isolated in West Africa by Alazard (1991), Molouba and colleagues (1999) and Chaintreuil and colleagues (2000). The recombinase A housekeeping gene was used preferentially over the ribosomal genes traditionally used, because it is more polymorphic and gives phylogenetic organizations that are similar to 16S rRNA-based phylogeny of the Bradyrhizobium genus, but provides a much higher level of taxonomic resolution (Gaunt et al., 2001; Vinuesa et al., 2005).

According to the results obtained, a possible scenario of stem-nodulating bradyrhizobial evolution could be drawn, involving an ancestral nod-independent nodulation coupled with a photosynthetic trait, followed by (i) occasional acquisitions of nod genes that broadened the bacterial host range to local Aeschynomene species, and (ii) loss of photosynthetic ability within a cluster at some intermediate evolutionary time point between classical root-nodulatingBradyrhizobium and stem-nodulating ones.


Aeschynomene sampling and characteristics of the isolates

We focused on stem-nodulated plants growing, like in Senegal, in the north tropics and in marshes or rivers close to the coasts. The sampling sites followed a gradient from the region of Veracruz, Mexico (19°N), Guadeloupe island (West Indies) 16°N, to the more equatorial French Guinea (5°N) (Table S1).

The Mexican samples were the most diverse ones in term of plant species, with A. indica, A. rudis, A. ciliata (coast area of 60 km long, south from Veracruz) and A. scabra species (Tenango, south from Mexico city). The Guadeloupe samples consisted of only one monospecific site, with Aeschynomene sensitiva. More surprisingly, we only found this latter species in French Guiana on four different sites along the coast around Cayenne (50 km long). Consequently, all stem-nodulated Aeschynomene spp. of our tropical American collection belonged to CI-group 3.

Bacterial symbionts were recovered from stem nodules and, when available, root nodules of each Aeschynomene plant sampled, as described in Experimental procedures. Fifty-six purified bacterial strains were retained for further analyses. Their origin, host plant, and phenotypes are listed in Table S1. After 1 week of incubation at 37°C under aerobic conditions and light exposure, most bacterial colonies turned either light pink (LP), dark pink (DP), orange (O) or red (R), suggesting the presence of carotenoids and photosynthetic pigments (Lorquin et al., 1997). Accordingly, besides carotenoids absorption in the visible region of the spectra (400–500 nm), spectrophotometric analyses revealed the presence of an A870 peak for most of the light-induced strains (Fig. S1), which corresponds to the bacteriochlorophyll-containing light-harvesting polypeptide complex I (LHI-Bchl) (Fleischman and Kramer, 1998).

Contrarily to reference strains ORS278 and BTAi1, which produce Bchl only when exposed to light–dark cycles, or under a continuous 730 nm induction (Evans et al., 1990; Giraud et al., 2000), several of our strains did not display any light-dependent Bchl production (Table S1). This was particularly the case for the strains isolated from French Guiana, where 73% of the isolates constitutively produced Bchl (Fig. S1B). Stem-nodulating photosynthetic bradyrhizobia that were not sensitive to light regulation had also been isolated in north-east Argentina by Montecchia and colleagues (2002), but they made up smaller proportion (22%). Accordingly, our African collection also contained 33% of strains that were constitutively photosynthetic (Table S1).

Despite their isolation from A. sensitiva nodules (CI-group 3), four strains (STM3843, STM3844, STM3846 and STM3964) displayed white colonies even after light induction and no photosynthetic pigment could be detected in their absorption spectra (Fig. S1C). Degenerate primers designed to amplify pufLM genes from both the photosynthetic Bradyrhizobium spp. and their closest photosynthetic relative, Rhodopseudomonas palustris (Young et al., 1991), were used, but they proved unsuccessful for these four strains (Table 2). Since pufLM genes encode the reaction centre polypeptides of the bacterial photosynthetic apparatus, the lack of PCR amplification is in accordance with our results.

Table 2.  PCR amplification of photosynthetic (pufLM) and nodulation (nodA) genes from representative strains of Bradyrhizobium spp.
StrainCI-goupaTarget genes
  • a. 

    Aeschynomene–Bradyrhizobium cross-inoculation group.

  • b. 

    Rhodopseudomonas palustris CGA009.

  • +, positive PCR amplification; −, no amplification product detected.

R. palustrisb +

Phylogenetic relationships among Bradyrhizobium spp. isolates

The genetic diversity of bacterial isolates originating from African and American samplings was analysed by two approaches: recA phylogeny and AFLP whole-genome fingerprints.

recA phylogeny. recA gene sequences were determined for the 56 strains originating from French Guiana, Mexico and Guadeloupe, as well as for 51 strains previously isolated in West Africa (Table S1). African strains also included bradyrhizobia belonging to CI-groups 1 and 2. Sequences from various reference strains were added, which consisted of six Bradyrhizobium species as well as R. palustris. Sinorhizobium meliloti 1021 was chosen as an out-group strain to root the tree.

The maximum likelihood (ML) phylogenetic tree based on a 547 bp recA fragment is presented in Fig. 1A. All of the photosynthetic strains (including Bradyrhizobium denitrificans) formed a single cluster with a strong (88%) bootstrap support. Within this clade 11 different groups were detected, each with a bootstrap value support above 90%, and grouping together strains isolated from various sites and Aeschynomene species.

Figure 1.

Phylogenetic analyses of Bradyrhizobium isolates. Strains numbers followed by ‘-II’ indicate isolates belonging to cross-inoculation group 2. Photosynthetic clade is shaded in grey block. Strain numbers were coloured according to their geographical origin: blue, Guadeloupe; purple, Mexico; red/pink, Africa; green, French Guiana. Letters at the right of strain names indicate the plant of isolation: af, Aeschynomene afraspera; ci, A. ciliata; in, A. indica; ru, A. rudis; sc, A. scabra; se, A. sensitiva; ob, Oryza breviligulata (wild rice). Reference strains and non-photosynthetic African strains (ORS) are written in black (or grey for CI-group 2).
A. recA maximum likelihood (ML) tree was built using a GTR model (six types of substitution, discrete gamma model). Bootstrap support values (100 replicates with phyML) are provided as percentage at the corresponding nodes when > 50%. Strains not listed in Table S1 and their corresponding accession numbers were: Bradyrhizobium canariense BC-MAM11 (AY653748), B. denitrificans LMG 8443 (EU665419), B. elkaniiUSDA46 (AY591575), B. japonicumUSDA110 (BA000040), B. liaoningense ICMP 13639 (AY494833), B. yuanmingense CCBAU 10071 (AY591566), Rhodopseudomonas palustrisCGA009 (BX572605), Sinorhizobium meliloti 1021 (AL591788).
B. Composite dendrogram of AFLP analyses. The banding patterns obtained with four independent AFLP primers were pooled to produce a composite tree (see Experimental procedures). Phylogenetic relationships were performed by using upgma analysis and clusters were delineated at a Dice similarity level of 55%.

Apart from B. denitrificans, all of the Bradyrhizobium reference species fell outside the photosynthetic clade. It is worth noting that African non-photosynthetic strains isolated from CI-group 1 and 2 Aeschynomene plants clustered with those references strains. The four non-photosynthetic strains originating from French Guianan A. sensitiva nodules (CI-group 3) formed a separate cluster (XI), at an intermediate distance between the stem-nodulating photosynthetic clusters and classical non-photosynthetic strains.

recA nucleotide sequence comparisons revealed that ORS278 and BTAi1 shared only 90% sequence identity and 88% with that of B. japonicum USDA110, while Bradyrhizobium reference species shared from 91% to 93% recA sequence identity with each other. Bacterial diversity within Aeschynomene symbionts thus appeared as important as between the other known species of Bradyrhizobium, which may indicate that photosynthetic strains include closely related but different (sub)species, as suggested by Ladha and So (1994) and Willems and colleagues (2000).

Comparison of bacterial genomic fingerprints by AFLP. AFLP fingerprints analysis was used to assess bacterial diversity at the genomic level, on 110 isolates (Fig. 1B). The technique was reproducible, with a mean of 88.6 ± 0.7% similarity obtained for the five strains that were repeated for normalization (not shown). The AFLP study confirmed the high level of genomic diversity among the isolates, but all the photosynthetic strains still fell into a single clade. Above a mean Dice similarity coefficient (SD) value of 55%, 11 main clusters were delineated. Those were highly correlated with recA phylogeny, thereby confirming that sequence comparison of this single housekeeping gene gives a very good estimation of the bacterial strains evolutionary relationships (Vinuesa et al., 2005).

The cut-off level of 55% was arbitrarily chosen as the best correlation with clusters obtained on recA phylogeny, but interestingly similar cut-off level of diversity had been used in previous AFLP studies (Willems et al., 2000). Compared with the latter study, the dendrogram presented in Fig. 1B exhibited greater resolution; for example, strains ORS303, ORS380, ORS372 and ORS393 were all grouped in cluster IV in this analysis and with a similarity coefficient of 59%, while they were included in two different clusters in the previous study, and with a weaker support (SD = 40%).

Few differences between recA and AFLP fingerprint phylogenies could be detected: strain STM3956 was placed in AFLP group III while it was in a distinct clade in the recA phylogeny (Table S1). Such results underline the pitfall of using phylogenetic information from a single gene, as genes can be subjected to mosaicism or horizontal transfer (Vinuesa et al., 2005). AFLP was a more powerful tool for analysing strain relationships at the intraspecific level, in that it successfully grouped together African clusters IVa and IVb in a unique clade, and included strains ORS341 and ORS386 into cluster III. These results stress the importance of taking into account whole-genome composition when analysing the phylogenetic relationships of bacterial strains that are adapted to similar ecological niches.

Comparison of diversity patterns with geographical and host plant origins of isolates

Both recA and AFLP analyses highlighted the monophyly of all photosynthetic stem-nodulating bradyrhizobia. Bacterial strain distributions inside this clade were however not correlated with their respective CI-group, country and/or plant species of isolation, as all the clusters mixed the strains together regardless of their origin: this was particularly the case for clusters I and II (Fig. 1). On the other hand, some strains tended to group together according to at least one of those parameters. Indeed some clusters were exclusively composed of strains from the same location, regardless of their plant of origin, such as clusters V (from Mexico), IV and VI (West Africa) and XI (French Guiana). Conversely, clusters VIII, IX and X harboured strains that originated from both French Guiana and Senegal, but were all isolated from the same plant species: A. sensitiva.

Figure 1 results also highlighted a phylogenetic distribution of CI-group 2 bradyrhizobia according to their photosynthetic status. On the one hand, the non-photosynthetic strains ORS336 and ORS354 grouped with non-photosynthetic Bradyrhizobium reference species and all CI-group 1 isolates. On the other hand, strains ORS285, ORS300, ORS303, ORS320, ORS324 and ORS380 were distributed among CI-group 3 photosynthetic bacterial clusters IV and VI. As the genome composition of the strains belonging to these clusters appears to be similar (cf. AFLP data), this raises again the question about what makes the difference between photosynthetic stem-nodulating bradyrhizobia belonging to CI-groups 2 and 3.

Bradyrhizobium spp. host specificity: search for nod genes

All of the bacterial strains isolated in the present work originated from Aeschynomene species belonging to CI-group 3 (A. indica, A. sensitiva, A. ciliata, A. scabra and A. rudis). Their host specificity was still uncertain, since CI-group 2 strains could also be isolated from CI-group 3 plants in Senegal [e.g. strains ORS300, ORS320 and ORS380 were isolated from A. sensitiva and A. indica stem nodules (Table S1)]. To test for their host range specificity, these strains were inoculated on A. indica (CI-group 3) and A. afraspera (CI-group 2), with ORS285 used as a positive control for both species. All of our strains were able to nodulate A. indica but none nodulated A. afraspera (Table S1). Thus, they all belong to the highly specific CI-group 3, including also the four non-photosynthetic strains from French Guiana.

The inability of strain ORS278 to nodulate A. afraspera had been linked to the absence of canonical nodABC genes in its genome (Giraud et al., 2007). This feature may be shared among CI-group 3 strains, as suggested by the failure of Chaintreuil and colleagues (2001) to detect nodABC genes in all CI-group 3 African strains they analysed.

To check whether the absence of canonical nod genes was widespread among CI-group 3 strains, we searched for nodulation genes in our strains. No PCR product could be detected for any CI-group 3 isolate, regardless of the primers used (not shown). Those results were confirmed by Southern dot-blots analyses, where only the positive controls (CI-groups 1 and 2 strains) gave hybridization signals above background (Fig. S2). This further indicates that, like the two reference strains ORS278 and BTAi1, none of our bacteria harbours the canonical nod genes, a trait that thus appears to be shared by all strains belonging to CI-group 3.

As expected we obtained a PCR amplification of nodA in the positive controls from both the CI-group 1 and the CI-group 2 strains (Table 2). Analyses of their sequences showed that the nodA genes from all the non-photosynthetic bradyrhizobia (either from CI-group 1 or from CI-group 2) were similar to the nodA sequences from various other Bradyrhizobium species (Fig. 2). nodA sequences from photosynthetic CI-group 2 strains were however highly divergent from the other bradyrhizobia and fell in two distinct subclusters that strictly followed AFLP clusters IV and VI.

Figure 2.

Maximum likelihood (ML) phylogeny of nodA sequences. The tree was built using a GTR model (six types of substitution, discrete gamma model). Bootstrap support values (100 replicates with phyML) are provided as percentage at the corresponding nodes when > 50%. Sequences accession numbers are given in parentheses. A. caulinodan: Azorhizobium caulinodans. Strains numbers written in grey and followed by ‘-II’ indicate isolates belonging to cross-inoculation group 2. Shaded block highlights the photosynthetic clade.

Symbiotic phenotype of cluster XI strains

The discovery of non-photosynthetic bradyrhizobia that stem-nodulate A. indica was rather puzzling, since Giraud and colleagues (2000) had established a link between photosynthetic ability and stem nodulation. In order to better characterize their symbiotic phenotype, cluster XI strains were either root- or stem-inoculated on A. indica seedlings. Four weeks after inoculation, the nitrogen-fixation efficiency of the bacteria was clearly visible, with plants having a dark-green colour and three times more leaves than the un-inoculated negative control (not shown). Cluster XI strains formed a number of root and stem nodules on A. indica plants comparable to that induced by the control strain ORS278 (Table 3). Nodule morphology was also similar, with dark green nodules appearing at the base of lateral roots, or at the sites of dormant stem–root primordia (Fig. S3), as previously described (Alazard and Duhoux, 1988).

Table 3.  Symbiotic phenotypes of cluster XI non-photosynthetic strains compared with those of reference strains.
StrainRoot inoculationStem inoculation
No. of root nodules plant−1Plant fresh weight (mg)Fixation activity (nmol h−1 plant−1)No. of stem nodules plant−1Plant fresh weight (mg)Fixation activity (nmol h−1 plant−1)
  • a. 

    Strain USDA 110.

  • Results are the means of four replicates, 30 days post inoculation. Values of each column followed by identical letters are not significantly different (Fisher test, P < 0.05).

ORS27821.0 (A, B)473 (A, B)164 (B)3.8 (B)530 (B)158 (B)
STM384331.8 (A)518 (A)212 (A)5.0 (A, B)560 (A, B)170 (A, B)
STM384429.8 (A, B)456 (A, B)185 (A, B)5.5 (A)501 (B)169 (A, B)
STM384631.8 (A)525 (A)174 (A, B)6.0 (A)620 (A)213 (A)
STM396418.5 (B)379 (B, C)164 (B)4.5 (A, B)512 (B)171 (A, B)
B. japonicuma0 (C)336 (C)21 (C)0 (C)410 (C)0 (C)
No inoculum0 (C)333 (C)0 (C)0 (C)373 (C)0 (C)

Nitrogen fixation rates of the strains in planta were also statistically not different (or sometimes stronger) from those of the control strain ORS278 (Table 3). This indicates that, contrary to ORS278, cluster XI strains were not affected by the lack of photosynthetic activity to efficiently stem-nodulate Aeschynomene. As expected the negative control B. japonicum USDA 110 (CI-group 1) could neither nodulate roots nor nodulate stems of A. indica. It is however noteworthy that a weak nitrogenase activity could be detected on cut roots (Table 3), which probably resulted from a diazotrophic activity of the bacteria on root surface.


Phylogenetic relationships among Bradyrhizobium spp. isolates

Bradyrhizobial diversity had been extensively studied on a collection of African strains, most of them originating from various sampling sites in Senegal. By using DNA–DNA hybridization and 16S−23S ribosomal DNA (rDNA) internal transcribed spacer region sequence analyses, Willems and colleagues (2001a; 2003) identified more than 11 genospecies among bradyrhizobia, with photosynthetic Aeschynomene isolates forming a single cluster composed of at least two distinct but closely related (sub)species. To better understand the evolution and phylogenetic structure of stem-nodulating bradyrhizobia, our aim was to compare the diversity of strains from West Africa (Table S1) with isolates from tropical America, this latter region concentrating – with Africa – the highest diversity of the dalbergioid/aeschynomenoid clade of tropical legumes (Lavin et al., 2001).

Surprisingly, we did not detect any obvious genetic diversity structure of the bacterial isolates according to their origin. This lack of geographical structuring suggests that strains associated to Aeschynomene are widely scattered all over the distribution area of this genus. Accordingly van Berkum and colleagues (1995) could isolate Aeschynomene-nodulating bradyrhizobia from soils collected all over the world, regardless of whether the host plant was present or not. Moreover, the type strain of Blastobacter denitrificans[now renamed Bradyrhizobium denitrificans (van Berkum et al., 2006)] was included inside the photosynthetic Aeschynomene stem-nodulating clade, while this bacterium had been isolated from the surface water of a lake in Germany (Hirsch and Müller, 1985). This suggests that the bacterial ecology and spatial distribution of Aeschynomene-associated symbionts are far from being strictly dependant on the presence of their host plant. This can also be related to the results of Vinuesa and colleagues (2005), who suggested that strains belonging to various Bradyrhizobium species generally present a weak structuring according to the distances, with a wide dispersion potential among continents. Figure 1 shows however that although bacterial symbionts were spread out among various sub-branches of the phylogenetic trees, several clusters presented a tendency of microstructuring according to plant species and/or country of origin. Thus, a fine structuring linked to the ecology of the strains may exist, although a broader sampling would be required to confirm these observations.

Selective pressure on photosynthetic traits

The monophyly of all photosynthetic stem-nodulating bradyrhizobia isolated from various Aeschynomene species, regardless of their country of origin, suggests that the photosynthetic trait was acquired only once and was ecologically important enough to be conserved during the diversification of this cluster. It had long been suggested that present rhizobia arose from a photosynthetic ancestor (Jarvis et al., 1986; Sprent, 1990) and rooting of the Bradyrhizobium spp. clade with R. palustris is consistent with this hypothesis (Vinuesa et al., 2005). Like other aquatic aerobic phototrophic bacteria, the use of light as a supplementary source of energy may confer a selective advantage for bacterial survival ex planta; but photosynthesis may also facilitate bacterial infectivity and symbiotic effectiveness (Yurkov and Beatty, 1998; Giraud and Fleischman, 2004).

In a worldwide study of bradyrhizobia capable of nodulating A. indica, van Berkum and colleagues (1995) isolated several strains that did not synthesize bacteriochlorophyll and carotenoids pigments. Yet, most of them formed nodules that were ineffective for nitrogen fixation. This could indicate that bradyrhizobia have a preponderantly free-living lifestyle, and that in the absence of selective pressure by a host plant, those strains lost their ability to induce the formation of functional nodules. Life in the soil may also have led to the loss of any photosynthetic advantage, compared with a light-regulated aquatic environment. The enrichment of ineffective and non-photosynthetic strains may thus be a bias resulting from the indirect strain isolation from sampled soils and by plant trapping. In contrast, Molouba and colleagues (1999) established a strong correlation between naturally occurring stem-nodulating Aeschynomene symbionts and their photosynthetic phenotype.

Accordingly, most of the strains we isolated from Aeschynomene samples were photosynthetic, but we also found a cluster of naturally occurring non-photosynthetic symbionts in French Guiana. The phylogenetic position of these strains in a distinct cluster at an intermediate distance between photosynthetic and classical bradyrhizobia was confirmed by 16S rDNA sequence analyses (not shown), which indicates that cluster XI strains may represent a new Bradyrhizobium species. Those strains were able to nodulate both stem and roots of A. indica and were at least as effective as the model strain ORS278 at fixing nitrogen (Table 3). This constitutes a paradox compared with the work of Giraud and colleagues (2000), which demonstrated that a mutant impaired in its photosynthetic activity displayed a weaker symbiotic interaction.

Since the four cluster XI strains were all recovered from root nodules, this may explain the lack of selective pressure to keep photosynthetic traits. Yet, Montecchia and colleagues (2002) did isolate two non-pigmented bacterial strains (out of 46) from naturally occurring stem nodules of A. rudis in north-east Argentina (although no data about nitrogen fixation efficiency of these strains in planta are available). The question thus still remains opened about the importance of the photosynthetic trait for stem-nodulating bradyrhizobia, whether it is associated with symbiosis with the plant or a free-living lifestyle survival in the environment. A fine analysis of molecular traces of selection on photosynthetic genes might help answering this question.

Acquisition of nod genes

All strains isolated from our sampling of stem-nodulated Aeschynomene in Tropical America were able to nodulate A. indica, but not A. afraspera. Moreover no classical nod genes could be detected, which is consistent with their classification within CI-group 3. It is quite surprising that no bacteria belonging to CI-group 2 could be found in our study, as those bacteria (like strain ORS285) may nodulate Aeschynomene species belonging to both CI-group 2 and CI-group 3 (Alazard, 1985). In addition, several photosynthetic strains belonging to CI-group 2 could be isolated by Alazard (1985) and Molouba and colleagues (1999) from environmental samples of A. indica and A. sensitiva (CI-group 3) in Senegal.

But no Aeschynomene plant species belonging to CI-group 2 could be detected either in our sampling, and all CI-group 2 plants identified so far seem to be restricted to the African and Asian continents (http://www.ildis.org; Boivin et al., 1997). Those results would indicate that the area of distribution of CI-group 2 bacterial strains is tightly linked to the presence of their host plants, which is in contradiction to the high dispersion rate and the independence of Aeschynomene symbionts from their host plant, as previously suggested.

It is however interesting to note that the African strains belonging to CI-group 2 are not clustered according to their host range on the phylogenetic trees, but to their photosynthetic status (Fig. 1). Moreover AFLP studies revealed that strains like ORS285 and ORS303, belonging to clusters VI and IVa, respectively, are more closely related to CI-group 3 strains than they are with each other. As a consequence they are expected to have many genes in common besides photosynthetic ones. The most probable hypothesis would be that photosynthetic bradyrhizobia of CI-group 2 have acquired nod genes through lateral gene transfer, which then broadened their host range from CI-group 3 to CI-group 2 plants. The apparent lack of these strains in America would mean that acquisition of nod genes is favoured or selected by the presence of Aeschynomene plant species belonging to CI-group 2.

Phylogenetic analyses revealed that the nodA gene sequences of photosynthetic stem-nodulating bradyrhizobia were divergent from all other nodA genes already described (Fig. 2). Two subclusters that follow AFLP dendrogram could be detected, possibly indicating two independent nod gene acquisition events and further followed by nod gene loss in strain ORS2006 (Fig. 1B, cluster VI). Alternatively, the present phylogenetic tree may reflect a more ancestral acquisition event, followed by numerous losses in the progeny. In all cases the selective pressure to keep nodulation genes seems rather low, since most strains lack those genes. Furthermore, the nodA phylogenetic tree shows that they have undergone an important nucleotide divergence, which is indicative of a relaxation of selective pressure. This may indicate that nodulation of CI-group 2 plants by photosynthetic bradyrhizobia is more an alternative ecological lifestyle, their main host being Aeschynomene belonging to CI-group 3, and would explain the absence of these strains in locations where Aeschynomene species belonging to CI-group 2 are absent.

Concluding remarks

Our results showed that the lack of nod genes in bradyrhizobial–Aeschynomene symbionts is shared by all CI-group 3 strains. nod-independent symbiosis was spread in strains of both Tropical Africa and Central America and seemed to be linked to the CI-group 3 Aeschynomene spp., but was not strictly dependent on photosynthetic ability, as exemplified by the cluster XI strains. As CI-group 3 Aeschynomene spp. and their associated symbionts are widespread all over the tropics, we can hypothesize that nod-independent stem nodulation of Aeschynomene is an ancestral interaction, and occasional acquisition of nod genes by lateral gene transfer has led to the adaptation of some CI-group 3 strains to local Aeschynomene CI-group 2 species, thereby broadening their host range.

The discovery of an intermediate cluster of efficiently stem-nodulating bradyrhizobia that are not photosynthetic and surprisingly lacking nodA gene greatly increases the genomic diversity of the bacteria belonging to CI-group 3 that was first described by Alazard (1985). This high diversity of nod-independent symbionts thus opens great opportunities for making comparative genomic studies on representative isolates in order to highlight which genetic traits are involved in this very peculiar model among rhizobia–legumes interactions.

Experimental procedures

Bacterial strains isolation and growth conditions

The phenotypic and genomic diversity of African bradyrhizobia nodulating various species of Aeschynomene had been previously studied on a collection of isolates originating mostly from Senegal (Table S1). The sampling area consisted of eight sites, most of them being on the coast south of Dakar, Senegal, but also extending to the eastern part (350 km away, sampling sites ranging from 16°N to 13°N and 17°W to 13°W). Photosynthetic stem-nodulating bradyrhizobia included in our study were isolated from A. indica, A. sensitiva (CI-group 3) and A. afraspera (CI-group 2), but also from wild rice growing in temporary marshes that are also colonized by CI-group 3 Aeschynomene species (Chaintreuil et al., 2000).

Strains isolated from Aeschynomene spp. in Tropical America are also listed in Table S1. Bacterial isolates were obtained from root and stem nodules of Aeschynomene species collected from natural fields in different regions of French Guiana, Guadeloupe and Mexico. Nodule endosymbionts were isolated and purified on Yeast Extract Mannitol medium, as described by Sy and colleagues (2001). For nodules that were too hard for a direct isolation from the stems, symbionts were isolated by plant trapping: stem pieces containing the nodule were placed in Jensen medium with A. indica seedlings, and trapped symbionts were recovered 3 weeks later from A. indica root by grinding the nodules.

Plant culture and determination of symbiotic effectiveness

Seeds of A. indica and A. afraspera were scarified and surface sterilized with concentrated H2SO4 for 30 min, then 5 min in 3% Ca(OCl)2 after thorough washings. Plant cultivation and inoculations were carried out in nitrogen-free Jensen medium as previously described (Alazard, 1985). For each strain nodulation tests were performed in triplicates and in two independent experiments. Nodules formation was checked 3 weeks after inoculation. Several samples were randomly taken to re-isolate bacteria from surface-sterilized nodules and check the inoculated strain was recovered by PCR-RFLP analyses of their recA gene (not shown).

The ability to form nodules on stems of A. indica was tested on 2-week-old plants by applying the liquid inoculum (109 cells ml−1) onto the non-flooded area of the stem with a sterile cotton tip. Acetylene reduction assays (ARA) were performed 4 weeks after inoculation using standard techniques (Eaglesham and Szalay, 1983) on either excised stems or roots. Statistical multiple pair-wise comparisons of the results (anova) were performed using XLSTAT 8.0 software (AddinSoft).

Detection of bacterial photosynthesis apparatus

Bradyrhizobium spp. strains were aerobically grown for 5–7 days in a modified yeast extract mannitol agar medium (Giraud et al., 2000) in sealed Petri dishes at 37°C in either complete darkness or continuous illumination conditions provided by light-emitting diodes with a wavelength of 730 nm and irradiance of 6.6 μmol of photons m−2 s−1. Absorption spectra of intact cells were measured with a Cary 50 UV-Vis spectrophotometer (Varian) as described previously (Giraud et al., 2000).

DNA isolation, amplification, sequencing and analysis

Bacterial DNA was phenol-extracted using standard protocols. Primers used for PCR amplification and sequencing are listed in Table S2. PCR reactions were performed as previously described (Sy et al., 2001), except nodAB amplification, for which touchdown PCR conditions were modified (annealing temperature decreasing from 60°C to 50°C in 20 cycles) and a new forward primer designed for a better amplification of both CI-group 1 and CI-group 2 strains. Nearly full-length 16S rDNA was amplified using the universal eubacterial 16S rDNA primers FGPS6 and FGPS1509 (Normand et al., 1992) and the resulting 1500 bp fragments were subsequently sequenced by using two additional primers (16-1080r and 16S-870f). All sequences were corrected and assembled using ChromasPro v1.33 (Technelysium Pty). Multiple alignments were first performed with MEGA 4 (Tamura et al., 2007) then submitted to the MUSCLE/PHYML/TreeDyn online package of Phylogeny.fr (Dereeper et al., 2008) to perform ML analyses with the GTR substitution model.

Southern dot-blot hybridizations

One microgram of genomic DNA from each Bradyrhizobium strain was denatured in 0.4 N NaOH and spotted on positively charged nylon membranes (Roche Diagnostics, Germany). Mixtures of probes were prepared to cover nodA nucleotide sequences diversity (Fig. 2): nodA genes were PCR amplified from all Bradyrhizobium strains belonging to CI-groups 1 and 2 (Table S1), and from B. japonicum USDA 110. Probes were labelled and membranes were hybridized (in triplicate) using the DIG-high prime DNA labelling and detection starter kit II from Roche Diagnostics, according to the manufacturers' instructions. Moderate stringency conditions were used to allow up to 20% mismatches between each probe and target.

Amplified fragment length polymorphism (AFLP)

Previous investigations on bradyrhizobial diversity identified AFLP as the most powerful tool to provide information at the intraspecific level (Willems et al., 2001b). In this work, we used the single restriction enzyme-based AFLP method described by Mueller and colleagues (1996), with modifications. Briefly, for each sample, 1 μg of DNA was digested with PstI and ligated to adapters. Ten nanograms of the adapted PstI fragments were then selectively amplified using primers having a sequence homologous to the adapters, with the following 2 bp extensions at the 3′ end: -AA, -AC, -AT or -GT. The best primer extensions to be used were selected after analyses of BTAi1 and ORS278 genome sequences, using the AFLP-PCR in silico tool (http://insilico.ehu.es/) developed by Bikandi and colleagues (2004).

PCR products were run on 1% agarose gels in 0.5× TBE buffer. Gel pictures were analysed with the GelCompare II (v 3.0) software (Applied Maths, Kortrijk, Belgium), based on binary band-matching tables from which pair-wise similarity matrices were calculated using Dice's similarity coefficient. The banding patterns from the four primers were combined to produce a composite upgma similarity tree. Five samples were repeated at least twice to check for repeatability and adjust band matching optimization parameters (not shown).

Nucleotide sequences accession numbers

The recA partial gene sequences determined in this study have been deposited in the GenBank database under Accession No. EU665322 to EU672454 (Tables 2 and 3). Those for the partial 16S rRNA and nodA genes has been assigned Accession No. EU781678 to EU781685 and FJ150398 to FJ150405 respectively.


We thank Bernard Dreyfus for providing Aeschynomene samples from Guadeloupe, Jean-Jacques de Granville and Sophie Gonzales from ‘Herbier de Guyane’ (IRD) for collaboration in French Guiana, Erika Yashiro for her assistance with manuscript correction. This work was supported by ANR (Agence Nationale de la Recherche – France) through ‘young researcher grant’– BOA (JC05_52208).