![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Previous Article | Next Article 
Journal of Bacteriology, July 2007, p. 5223-5236, Vol. 189, No. 14
0021-9193/07/$08.00+0 doi:10.1128/JB.00105-07
Copyright © 2007, American Society for Microbiology. All Rights Reserved.
Laboratoire des Symbioses Tropicales et Méditerranéennes, UMR 113 IRD-Cirad-SupAgro-UM2/USC INRA, Campus de Baillarguet, 34398 Montpellier Cedex 5, France,1 Institut des Sciences de l'Evolution de Montpellier, UMR 5554 CNRS-UM2, Montpellier, France2
Received 18 January 2007/ Accepted 1 May 2007
 |
ABSTRACT |
|---|
 Top ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
First, sex can promote adaptive evolution through various mechanisms, some of which are specific to bacteria. For instance, illegitimate recombination or horizontal gene transfer (HGT), leading to the insertion of new DNA sequences, could enhance the ecological niche of a bacterial species through the acquisition of new metabolic capabilities (21). By placing homologous adaptive mutations in different genetic backgrounds, homologous recombination could uncouple the evolution of different genomic regions (30) and thus also facilitate adaptive evolution.
Second, the addition of a new DNA fragment through HGT could promote speciation among bacterial lineages, as this would lower the rate of homologous recombination among initially promiscuous lineages (55). It has also been suggested that disruptive selection, because it promotes sequence divergence at loci undergoing such selection, might lead to overall lower rates of recombination and thus to increased divergence among lineages (8). Yet homologous recombination might be an efficient cohesion force that prevents bacterial speciation in a neutral Fisher-Wright model (24). In this context, empirical studies focusing on the interplay between ecological adaptation and sex are needed to gain a better knowledge of the processes leading to bacterial diversification.
Nitrogen-fixing bacteria that interact with leguminous plants belonging to the genus Medicago appear to be an ideal biological model for such studies. Although only three bacterial species have been described as the usual symbionts of Medicago plants, Bena et al. (3) showed that the genus Medicago includes at least five groups of species featuring different symbiotic specificities, as follows. (i) The sister species Medicago laciniata and Medicago sauvagei interact efficiently with a specific ecotype of Sinorhizobium meliloti called S. meliloti bv. medicaginis (56). (ii) A paraphyletic group of Medicago species, which includes Medicago rigiduloides, Medicago noeana, and Medicago radiata, interacts with another ecotype of S. meliloti. Most species belonging to the last emerging clade of Medicago species can benefit from nitrogen fixation by interacting with either (iii) Sinorhizobium medicae (e.g., Medicago polymorpha) or (iv) S. medicae and S. meliloti bv. meliloti (e.g., Medicago truncatula) (38). (v) Finally, Rhizobium mongolense, which is also described as Rhizobium gallicum bv. orientale (43), is the specific symbiont of Medicago ruthenica (53).
Since the sequencing of the genome of S. meliloti strain 1021 (16), considerable efforts have been made to characterize the impact of sex on the genetic structure of S. meliloti bv. meliloti and S. medicae populations (1, 49, 54). Within the S. meliloti bv. meliloti strain 1021 genome, the chromosome harbors mainly core genes, encoding functions related to central metabolism and informational processing, whereas the pSymA and pSymB megaplasmids harbor most of the accessory genes, which commonly encode supplementary metabolic pathways and symbiotic factors. Specifically, bacterial genes involved in recognition by host plants through Nod factor synthesis, such as nod, noe, and nol genes, are localized on pSymA. Furthermore, several genes that encode surface polysaccharidic compound pathways (lipopolysaccharide, exopolysaccharide, and capsular polysaccharide), e.g., exo genes, are located on pSymB (Fig. 1). Using a comparative genomic hybridization approach, Giuntini et al. (18) found that the megaplasmids of S. meliloti, especially pSymA, behave as hot spots of gain and loss of genes. Moreover, MLSA showed that homologous recombination involving plasmidic loci is quite common within both S. meliloti bv. meliloti and S. medicae, leading to linkage equilibrium among the three replication units (1, 49) and within megaplasmids (1). On an interspecific scale, genetic studies suggested that S. meliloti bv. meliloti and S. medicae are sexually isolated, even though they can cooccur on the same individual host plants (1, 31, 54). Nevertheless, nothing is known about the genetic differentiation among S. meliloti ecotypes, in particular between S. meliloti bv. meliloti and S. meliloti bv. medicaginis, which are the only two S. meliloti biovars formally described so far as Medicago symbionts (56). Populations of these two biovars can be found in sympatry, as can the two species S. meliloti and S. medicae. Barran et al. (2) described polymorphism in the nodC gene as the main determinant of the variation in host range between the two S. meliloti biovars. Such specificity could induce disruptive selective pressures as well as reproductive isolation mediated by the host plant, thus promoting the coexistence of conspecific ecotypes at the community level. This process might then lead to the divergence of ecotypes for genes not involved in host specificity (i.e., speciation) due to selective sweeps acting on each ecotype and/or sexual isolation. Conversely, homogenizing selection acting on the core genome (i.e., purifying selection and global selective sweeps) and frequent recombination events might prevent strains specialized from different Medicago species, e.g., M. laciniata or M. truncatula, from diverging in genes not involved in specificity and thus from undergoing speciation (26).
|
|
MATERIALS AND METHODS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Each bacterial isolate was then grown in 20 ml YEM medium. We also grew the following in 20 ml YEM medium: (i) S. medicae strain STM 1605 and S. meliloti bv. meliloti strain STM 1604 (1); (ii) three S. meliloti strains which had been described as interacting specifically with M. noeana, M. radiata, and M. rigiduloides, namely, USDA 1613, USDA 1614, and USDA 1623 (3); and (iii) Rhizobium mongolense strain USDA 1844, which fixes nitrogen with M. ruthenica (53). We used these strains to obtain a nod gene phylogeny illustrating the diversity of the specificity groups described so far as being associated with Medicago species.
DNA extraction and sequencing. (i) Experimental procedures. One milliliter of liquid culture was washed twice in an Eppendorf microcentrifuge tube by centrifugation (15,000 x g, 4 min), and the pellet was resuspended in 750 µl of sterile water. One hundred microliters of this solution was incubated for 2 hours with 20 µl of 1-mg/ml proteinase K and 100 µl of Tris-HCl (10 mM, pH 8.3). After boiling, this mixture was used as a DNA matrix. DNA amplification was performed using a Perkin-Elmer 2400 thermocycler with 25-µl volume reaction mixtures containing 1 µl of DNA matrix, a 200 µM concentration of each deoxynucleoside triphosphate, a 0.8 µM concentration of each primer, 1.5 mM of MgCl2, 1x buffer supplied by the Taq polymerase manufacturer, and 1.25 U of Invitrogen Taq polymerase. We used a touchdown program including an initial denaturation stage (96°C, 4 min); 20 cycles of denaturation (96°C, 30 s), annealing (annealing temperature decreased steadily from 60°C to 50°C in 20 cycles, 30 s), and elongation (72°C, 1 min); and 20 further cycles of denaturation (96°C, 30 s), annealing (50°C, 30 s), and elongation (72°C, 1 min). PCR products were cut from a 1% (wt/vol) agarose electrophoresis gel and purified with a QIAquick gel extraction kit (QIAGEN). Sequencing reactions were performed on one strand by using a DYEnamic ET Terminator kit (Amersham Bioscience) and were analyzed on an Amersham Bioscience Megabace 1000 DNA sequencer.
(ii) Phylogenetic markers defined within the nod gene cluster of the symbionts of Medicago species. We aimed at obtaining the following two sequence data sets for the nod gene region of rhizobia that interact with Medicago plants (Fig. 1): (i) partial sequences of the nodABC gene cluster, which is ubiquitous to the genomes of all rhizobia; and (ii) partial sequences of the nodEG gene cluster, which seems to be specific to Medicago symbionts. The wide distribution of the nodABC gene cluster allowed us to obtain a phylogeny rooted by an appropriate outgroup that includes a "core collection" of the diversity of nod genes among Rhizobiaceae. Using the nodEG data set allowed us to perform an analysis that takes into account all the diversity we observed within the symbiotic cluster. Primers were designed based on homologous regions conserved among rhizobia, obtained from the GenBank database (Table 1).
|
Data analyses based on the nod gene clusters of the symbionts of both Medicago plants and other Rhizobiaceae. (i) Alignment of sequence data sets. Nucleotidic sequences of the nodABC-nodEG gene cluster were obtained from S. medicae strain STM 1605; S. meliloti bv. meliloti strain STM 1604; S. meliloti bv. medicaginis isolates STM 2835 and STM 2836, from this study; S. meliloti strains USDA 1613, USDA 1614, and USDA 1623, which interact efficiently with Medicago rigiduloides, Medicago noeana, or Medicago radiata (3); and Rhizobium mongolense strain USDA 1844. These sequences were aligned using Clustal X 1.83 (51), together with both sequences available for the marker IGSNOD and with those of S. meliloti bv. meliloti strain 1021 (nodABC-nodEG; GenBank accession number [AN], NC_003037), strain 042B (nodABC; AN, AF038577), strain CC2093 (nodA; AN, AJ300226), strain CC2017 (nodA; AN, AJ300222), strain 102F28 (nodA; AN, AJ300221), and strain L5-30 (nodA; AN, AJ300220); S. meliloti bv. medicaginis strain USDA 1170 (nodABC; AN, AF522456); S. medicae strain A321 (nodA, nodB, and nodC; AN, AJ300224, AY929593, and EF428921, respectively) and strain M3 (nodA; AN, AJ300225); Rhizobium galegae strain HAMBI 1174 (nodABC; AN, X87578) and strain HAMBI 1207 (nodA; AN, AJ300240); Rhizobium tropici strain CFN 299 (nodABC; AN, X98514); Rhizobium leguminosarum bv. viciae strain 3841 (nodABC; AN, AM236084) and strain 248 (nodABC; AN, Y00548); R. leguminosarum bv. trifolii strain TA1 (nodA; AN, AY904443); R. mongolense strain USDA 1832 (nodB; AN, AY929577), strain USDA 1834 (nodB; AN, AY929578), strain USDA 1836 (nodB; AN, AY929579), strain USDA 1849 (nodB; AN, AY929580), strain USDA 1877 (nodB; AN, AY929581), strain USDA 1890 (nodB; AN, AY929582), strain USDA 1904 (nodB; AN, AY929583), and strain USDA 1929 (nodB; AN, AY929584); Rhizobium yanglingense strain SH22623 (nodB; AN, AY929585); Rhizobium giardini strain H152 (nodA and nodC; AN, AJ300238 and AF217267), strain Ro84 (nodA; AN, AJ300239), and strain H251 (nodC; AN, AF217264); R. gallicum strain PhD12 (nodA, nodB, and nodC; AN, AJ300237, AY929556, and AF217265, respectively), strain R602sp (nodA, nodB, and nodC; AN, AJ300236, AF29022, and AF217266, respectively), and strain FL27 (nodB and nodC; AN, AF29021 and AF217270); Rhizobium etli strain CFN42 (nodABC; AN, U80928) and strain GR-12 (nodB; AN, AY929587); Sinorhizobium strain NGR234 (nodABC; AN, U00090); Sinorhizobium terangae strain ORS1073 (nodA; AN, AJ300229) and strain ORS604 (nodA; AN, AJ300228); and Sinorhizobium saheli strain ORS611 (nodA; AN, AJ300227). Coding sequences were aligned based on amino acid translations, while intergenic spacers (IGS) were aligned directly. Alignments were edited with BioEdit 5.0.9 (19).
(ii) Detection of recombination.
We searched for traces of recombination in a concatenated data set including full sequences of nodABC and nodEG markers belonging to Medicago symbionts, using three methods implemented in RDP2 (i.e., MaxChi, Chimaera, and Geneconv), as these methods were the most accurate for detecting recombination events (29). The first two methods use a sliding window to detect significant discrepancies in the segregation of polymorphisms on either side of each sequence site. These are based on the computation of 
2 distributed statistics obtained from comparisons of pairs and triplets of sequences, respectively. Geneconv is an extension of Sawyer's method (40). It compares sequence pairs and scores unusually long series of polymorphic sites for which two sequences share a strong similarity. The null hypothesis of a lack of gene conversion events (i.e., homologous recombination) is tested through the randomization of the location of aligned sites. A P value is given as the proportion of permuted alignments which result in a greater score than the original alignment. For all these methods, insertion-deletion blocs were analyzed as single polymorphisms, we performed 1,000 permutation steps, and we used a P value cutoff of 0.05, using a Bonferroni correction for multiple tests (i.e., tests performed on several sequence sets and several sequence sites). We also searched consensus daughter sequences and consensus recombination breakpoints. For MaxChi and Chimaera, we used a variable window size of 20 polymorphic sites and 10% of variable sites, respectively. For Geneconv, similarity scores were computed using a gscale (i.e., a parameter which scales the impact of mismatches on similarity scores) set to 1.
(iii) Inference of a supertree. Taking into account the heterogeneity of our data set, we applied a supertree approach (see reference 5 for a review) to combine the information available from the different strains. The analysis of a concatenated data set in a likelihood framework is the best alternative for obtaining an accurate supertree (10). We therefore inferred a phylogeny from the alignments obtained from the nodABC-nodEG sequences by using MrBayes 3.1 (20). We used a Bayesian approach because its implementation in MrBayes 3.1 allows the use of mixed models for partitioned data analyses. Sequence evolution was modeled as follows: (i) each IGS or gene was considered a single partition; (ii) we assumed a different model for each partition, with parameters of the different models being estimated independently; (iii) the rate of evolution was allowed to vary among partitions; (iv) a GTR + I + G (general time reversible + invariant + gamma) model was used for IGS; (v) for coding sequences, the analysis was performed based on a codon model that uses a GTR transition probability matrix to model single-nucleotide evolution and in which we assumed variation in selection across sites (i.e., NY98 option). Two independent Metropolis-coupled Markov chain Monte Carlo searches were performed, using three chains with a temperature parameter set to 0.1, starting from a random tree and lasting 2 x 105 generations. Evidence for convergence of the two searches was obtained by examining the correlation between the posterior probabilities of individual clades and by observing the distribution of the log likelihood values over generations. We sampled a tree every 100th generation after the 1.5 x 105th generation according to convergence criteria. The topologies obtained during the two searches were pooled to compute posterior probabilities of clades and thus to assess their robustness.
(iv) Comparison of divergence and selective patterns of nod and housekeeping genes at the interspecific level. To study the impact of selective pressures on the pattern of genetic divergence between S. medicae and S. meliloti bv. meliloti at nod genes, we used DNAsp 4.0 (39) to compute both the pairwise sequence divergence, i.e., D(x,y) (the number of substitutions per site, using JC69 correction), and the pairwise ratio of Ka (the number of nonsynonymous substitutions per nonsynonymous site) to Ks (the number of synonymous substitutions per synonymous site) between DNA sequences belonging to S. medicae and S. meliloti bv. meliloti. First, these measures were obtained for the symbiotic genes nodA and nodB but were also obtained from partial sequences of nodC and nodE genes. Computations were performed using sequences of strains STM 1604 and STM 1605, except for nodA analysis, which was performed on the S. meliloti 1021 sequence and the S. medicae A321 sequence in order to take into account the whole gene sequence. Second, the mean D(x,y) value and the mean Ka/Ks ratio were computed between sets of alleles belonging to S. meliloti bv. meliloti and to S. medicae, respectively, for the partial sequences of housekeeping genes. With this aim, we used the interspecific MLSA data set published by Van Berkum et al. (54), focusing on seven genes on the genome of S. meliloti bv. meliloti, namely, sucA, glnD, recA, asd, gap, zwf, and ordL2. The allelic richness of these loci ranges from 1 to 6 alleles for S. medicae and from 6 to 10 alleles for S. meliloti bv. meliloti. A comparison of the D(x,y) values and the Ka/Ks ratios obtained from the set of symbiotic genes and from the set of housekeeping genes was conducted using a Mann-Whitney U test.
MLSA of S. meliloti and S. medicae polymorphism. (i) Sequence alignment, phylogenetic inferences, and detection of recombination. Nucleotide sequence alignments were performed using Clustal X 1.83 and manually edited with BioEdit 5.0.9. We built a maximum likelihood phylogeny for each molecular marker, using PAUP* 4b10, applying models of sequence evolution determined by the likelihood ratio test implemented in Modeltest 3.7 and performing heuristic searches by the TBR algorithm, starting from the neighbor-joining tree. The four phylogenies were used in three ways. To assess the robustness of bipartitions for each data set, 100 bootstrap pseudoreplicates were assessed by maximum likelihood as indicated above.
First, pairwise distance matrices were computed for the four loci according to the obtained topologies with PAUP* 4b10. These matrices were used to compute the following parameters for each locus: 
d, the mean pairwise divergence between isolates that belong to two different groups among S. medicae, S. meliloti bv. meliloti, and S. meliloti bv. medicaginis; and s, the mean pairwise divergence within the two same groups.
Second, we investigated the congruency between sequence evolution at each locus and symbiotic abilities of bacterial strains. Using Shimodaira-Hasegawa tests (41), we compared for each locus the likelihood of the data given the best tree to the likelihood of the data based on assuming the best topology obtained by enforcing the monophyly of S. medicae, S. meliloti bv. meliloti, and S. meliloti bv. medicaginis. Constrained trees and congruence test significances were inferred using the relevant options implemented in PAUP* 4b10, based on the evolutionary models and search parameters described above.
Third, Shimodaira-Hasegawa tests were performed to investigate likelihood differences among sequence data sets according to the topologies inferred from each of the different markers. These analyses were performed to test the null hypothesis of congruence of the topology obtained for each marker and therefore to infer whether homologous recombination might have uncoupled the evolution of the markers we used. To reveal whether significant incongruence between topologies could be explained by recombination events involving both S. meliloti biovars, we concatenated sequences of the four loci to screen S. meliloti MLSA data sets, using methods implemented in RDP2, with the following ordering: IGSRKP, IGSEXO, IGSGAB, and IGSNOD. Assuming that all these IGS are inherited vertically, they would feature similar polymorphism patterns. Conversely, an event of homologous recombination would induce a significantly different polymorphism pattern for the transferred locus compared to those for the other markers. Pairs or triplets of concatenated sequences were thus scanned with MaxChi, Chimaera, and Geneconv, using the settings described above.
(ii) Population structure. Sequences that differed from others by at least one nucleotide position were considered different alleles. The diversity of each bacterial group (i.e., S. medicae, S. meliloti bv. meliloti, and S. meliloti bv. medicaginis) was assessed using the summary statistics allelic richness (R) and Nei diversity (H) (33) indices, which were computed for each of the four MLSA data sets. In order to cope with differences in sample size, these statistics were obtained for 1,000 pseudoreplicates of the original data set by bootstrapping Nmin bacterial genotypes, where Nmin is the minimum population size observed. Expected allelic richness for each genetic group and 95% confidence intervals were estimated for the Nmin observed among groups. The contribution of each symbiotic group to the allelic richness of pairs of ecotypes at each locus was also computed. These measures were obtained using a program written in Turbo-Pascal. Differentiation among bacterial populations/taxa was analyzed using Wright's FST values among populations according to the Weir and Cockerham procedure (59), and exact tests of genotypic differentiation (36) were performed using Genepop 3.3, updated from the work of Raymond and Rousset (37).
Nucleotide sequence accession numbers. All sequences have been deposited in the GenBank database under the following accession numbers: IGSRKP, DQ405642 to DQ405767; IGSNOD, DQ405467 to DQ405592; IGSGAB, DQ406393 to DQ406518; IGSEXO, DQ406104 to DQ406229; and nodABC-nodEG, DQ406568 to DQ406583.
|
RESULTS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
|
We first investigated the patterns of differentiation on the interspecific scale. No allele at any locus was shared between S. medicae and either S. meliloti biovar. We therefore observed high FST values between species, and all differentiation tests between these groups rejected the null hypothesis of no differentiation (P < 0.001). Moreover, large genetic divergences were estimated between the two species for the markers we used (Fig. 3A).
For IGSNOD, while the genetic divergence between S. meliloti bv. medicaginis and S. medicae was almost the same as the divergence between the two S. meliloti biovars, it was 8.79 times higher than the divergence between S. meliloti bv. meliloti and S. medicae. Thus, for IGSNOD, the divergence between biovars within a species was much larger than the divergence between one of the two biovars and S. medicae. This intriguing pattern was consistent with the particular phylogenetic relationship among the three groups inferred from the IGSNOD maker. IGSNOD alleles of both S. medicae and S. meliloti bv. meliloti indeed grouped in a monophyletic cluster, whereas S. meliloti bv. medicaginis alleles formed a sister clade. This is also consistent with the results we obtained based on the nodABC-nodEG phylogeny (Fig. 2).
Conversely, the divergence between the two S. meliloti biovars was 5 to 16 times lower than the divergence between S. meliloti and S. medicae at IGSEXO (16.06 times), IGSRKP (16.56 times), and IGSGAB (5.22 times) (Fig. 4). Therefore, for these three loci, the divergence between biovars within a species was much lower than the divergence between any biovar and the other species.
|
d/s ratio, i.e., the mean pairwise divergence between isolates that belong to two different biovars of S. meliloti over the mean pairwise divergence within these groups, was much lower for IGSNOD than for the other loci (Fig. 3B). At this point, it is noteworthy that IGSNOD is the only marker supporting a topology congruent with the delineation of the two S. meliloti biovars and, therefore, the evolution toward two specificity groups within this species. For IGSRKP, IGSEXO, and IGSGAB, Shimodaira-Hasegawa tests comparing the likelihoods of data sets based on constrained (i.e., enforcing each S. meliloti biovar from a monophyletic group) and unconstrained topologies revealed significant differences (P < 0.05). (iii) Recombination evidence. Since we detected obvious differences in the population structures revealed by the four MLSA loci on the intraspecific scale, we looked for traces of recombination events within our data set. Since significant differentiation had been detected among S. meliloti biovars, we used a phylogenetic approach rather than population genetic tools such as linkage disequilibrium to study the influence of sex on S. meliloti diversification. The last measures indeed depend on both the sexual behavior and the genetic structure of the population under study. Furthermore, all of the Shimodaira-Hasegawa tests which aimed at comparing the likelihoods of the data based on the four phylogenies obtained from MLSA markers rejected the null hypothesis, suggesting that IGSGAB, IGSEXO, and IGSRKP also evolved independently (P < 0.05). Finally, based on the scan of the concatenated sequence alignment of the four loci with MaxChi, Chimaera, and Geneconv, we consistently inferred recombination breakpoints (P < 0.05) located between each locus by using sequence sets, i.e., pairs or triplets, which included genotypes belonging to either one or two S. meliloti biovars.
|
DISCUSSION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
These clustering patterns are remarkably related to those inferred from the phylogeny of legumes (61). Galega species, which interact with R. galegae (35), form a sister group to the vicoid clade. This clade includes the following three main lineages: (i) the Vicieae tribe (i.e., Vicia, Lens, Lathyrus, and Pisum spp.), whose members interact with R. leguminosarum bv. viciae (32); (ii) the genus Trifolium, which is associated with R. leguminosarum bv. trifolii (46); and (iii) other members of the Trifolieae tribe (i.e., Ononis, Trigonella, Melilotus, and Medicago spp.), which mostly interact with S. meliloti and S. medicae, with the only obvious exceptions to this symbiotic specificity being the genus Ononis, which interacts with a much wider range of strains (46), and the species Medicago ruthenica, which interacts with R. mongolense. Nevertheless, it is noteworthy that M. ruthenica appears to be the first divergent plant species in the strict consensus most-parsimonious molecular phylogeny of the genus Medicago (13). In accordance with phylogenetic evidence, Galega species and vicoid species are known to share the peculiar ability to interact with rhizobia that produce Nod factors (i.e., the products of the nod gene pathway) with alpha-beta-unsaturated fatty acids, such as R. galegae, R. leguminosarum bv. viciae, R. leguminosarum bv. trifolii, or S. meliloti (12). These data suggest that coevolution played a fundamental role in the diversification of both the symbiotic specificity of vicoid species and the nod genes of their specific symbionts, even though a strict cocladogenesis was not observed.
Conversely, phylogenies based on chromosomal genes do not support the monophyly of the bacterial lineages which interact with Medicago spp., with the topologies being consistent with the monophyly of the Sinorhizobium and Rhizobium genera (e.g., see reference 52). The incongruence between symbiotic and housekeeping gene phylogenies suggests that HGT of nod genes occurred between these two genera. In this context, the analyses performed up to now suggest that Sinorhizobium species acquired the ability to nodulate Medicago species through HGT of nod genes from the ancestor of a Rhizobium species, most probably R. mongolense. Such HGT across different taxonomic groups of bacteria is usually invoked to explain the spread of nod genes, both between alpha- and betaproteobacteria (7) and within alphaproteobacteria (60). This hypothesis was validated, for instance, with the report of transfer of nodulation and N2 fixation genes by acquisition of a symbiotic island that integrated into a Phe-tRNA that occurred in natura between Mesorhizobium strains (48).
Speciation and specialization of S. meliloti and S. medicae. During the evolution of Medicago symbionts, HGT of nod genes might have occurred either towards the common ancestor of the two Sinorhizobium species or towards one Sinorhizobium species followed by a spread to the other. All the topologies we obtained based on molecular markers located in the nod gene region showed that the sequences of S. meliloti bv. meliloti and S. medicae cluster in a common clade, while the whole S. meliloti species forms a paraphyletic group. Moreover, S. meliloti bv. meliloti also forms a paraphyletic group on the nodABC-nodEG supertree due to the position of the S. medicae clade. This branching pattern is explained by the incomplete lineage sorting of S. meliloti bv. meliloti and S. medicae at the nodA and nodB genes and is in agreement with topologies previously obtained based on nodD1 or nodB sequences (3, 42).
The close relationship between the nod genes of S. meliloti bv. meliloti and S. medicae is consistent with the fact that these two groups are characterized by overlapping host ranges (3). Conversely, the differentiation and divergence we observed for most MLSA loci between S. medicae and S. meliloti illustrate a substantial process of sexual isolation, consistent with previous results illustrating the differentiation between S. medicae and S. meliloti bv. meliloti (1, 14, 31, 54). Moreover, phylogenies based on a sample of housekeeping genes suggest that S. meliloti and S. medicae are not sister species (28). Most of the topologies obtained indeed suggest that S. medicae might be the first emerging taxon within a clade including S. medicae and S. meliloti but also another species, Sinorhizobium arboris (28), suggesting a rather ancient speciation event leading to the first two species.
Two different scenarios might explain the contrast between the polymorphism patterns observed for nod genes and other markers. First, horizontal transfer of nod genes might have occurred between S. meliloti and S. medicae after their speciation. According to this hypothesis, the close genetic relationship between nod genes of S. medicae and S. meliloti bv. meliloti should be due to their recent divergence. An alternative explanation which has previously been proposed is that S. medicae emerged from an ancestral population of S. meliloti bv. meliloti (6, 42). According to this scenario, the monophyly of the nod sequences of S. medicae and S. meliloti bv. meliloti would be due to historical factors, and after speciation, selective pressures imposed on the nod gene cluster by the host plants shared by the two groups could have prevented their divergence. While the divergence between S. medicae and S. meliloti bv. meliloti was significantly less for nod genes than for a sample of housekeeping genes, we did not detect a significant difference in the Ka/Ks ratios for these two groups of genes, suggesting that selection influenced sequence divergence in a similar way for the two groups of genes. Whereas the two groups of genes evolved under purifying selective pressures, as revealed by Ka/Ks ratios of <1, the similarity of Ka/Ks ratios of nod genes and housekeeping genes suggests that purifying selection driven by host plants could not explain the low level of divergence between S. medicae and S. meliloti bv. meliloti in the nod gene cluster.
Altogether, these results suggest that HGT of nod genes between S. medicae and S. meliloti bv. meliloti is the most likely explanation for the pattern we observed. Moreover, the discrepancy between the topologies supported by nodD1-nodAB and nodEG raises questions about the number of transfers that occurred between the two groups. Further insights about these issues will likely be obtained from comparison of the genome sequences of S. meliloti bv. meliloti strain 1021 and S. medicae strain WSM419 when the genome of the latter strain becomes available.
Recombination within S. meliloti and ecological specialization. Previous studies have demonstrated that the genetic structure of S. meliloti bv. meliloti is strongly influenced by homologous recombination (1, 31, 49, 54). Nevertheless, this gives us no direct information about the impact of recombination on the evolution of bacterial populations in the context of ecological specialization. Based on current knowledge of the bacterial speciation process, the impact of disruptive selection on the pattern of bacterial diversification remains an open question (15). In particular, some models predict that global selective sweeps affecting all conspecific ecotypes are required to limit the genetic divergence of the niche-specific lineages (9). Here we analyzed the genetic structure of S. meliloti biovars adapted to different host plants in order to know whether homologous recombination could prevent conspecific bacterial ecotypes from undergoing speciation. With this aim, we summarize the evidence of disruptive selection within the nod gene cluster of S. meliloti bv. meliloti and S. meliloti bv. medicaginis and then focus on the recombination mechanism that would affect this set of genes. Next, the impact of recombination and disruptive selection on the diversification of other genomic areas is considered by comparing the observed patterns of polymorphism with the expected patterns provided by a theoretical model which was previously published. Finally, the ecological conditions required to allow recombination among biovars are discussed.
S. meliloti bv. meliloti and S. meliloti bv. medicaginis are completely differentiated at IGSNOD (i.e., they have no shared allele). Moreover, IGSNOD was the only marker supporting the split between the two S. meliloti biovars, based on Shimodaira-Hasegawa tests. According to the model proposed by Majewski et al. (26), the s/d ratio, i.e., the sequence divergence within ecotypes over the sequence divergence between ecotypes, for IGSNOD (i.e., 5.85 x 10–2) would be obtained in cases where recombination is not strong enough to prevent the sorting of ecotypes. We therefore believe that a disruptive selection pressure due to the host plant has influenced the variability of some S. meliloti nod genes and might have further modified the polymorphism pattern of the entire nod gene region by hitchhiking. This hypothesis is consistent with a previous study which revealed that the efficiency of the interaction of S. meliloti bv. medicaginis strains with M. laciniata required a specific nodC allele (2). Furthermore, the high level of divergence between S. meliloti ecotypes at IGSNOD raises questions about the mechanism of recombination that affects the nod gene cluster. According to previous studies showing that the efficiency of homologous recombination depends on sequence divergence in a log-linear way (27, 58), the recombination rate among nod gene clusters of the different biovars of S. meliloti would actually be quite low.
Nevertheless, Shimodaira-Hasegawa tests show that recombination uncoupled the evolution of IGSNOD and IGSGAB, both on pSymA. This agrees with results from a previous study in which bacterial sampling was restricted to a single soil sample and a single biovar, where no significant linkage disequilibrium was found between these two molecular markers (1). Conversely, Sun et al. (50) found some significant linkage disequilibrium among the markers they studied on pSymA (49). Yet nod gene polymorphism was not taken into account in their study. According to the evidence of recombination between IGSNOD and IGSGAB, the lack of significant evidence of recombination within the nodABC-nodEG gene cluster, and Sun et al.'s results (i.e., linkage disequilibrium among pSymA loci that are not located within the nod gene cluster), it can be hypothesized that the whole nod gene region is frequently transferred among S. meliloti isolates and behaves as a hot spot of homologous recombination within the megaplasmid pSymA of S. meliloti. In agreement with this hypothesis, a study based on comparative genomic hybridization of S. meliloti strains revealed that pSymA is a hot spot of genomic polymorphism within the S. meliloti genome and that a gain or loss of genes tends to be clustered spatially in this replication unit, especially near nod genes (18). Alternatively, it could be that the linkage disequilibria found by Sun et al. (49) were essentially due to population structure, as the strains they studied originated from all over the world.
Furthermore, the significant incongruence among IGSRKP, IGSEXO, IGSGAB, and IGSNOD phylogenies confirmed that homologous recombination has influenced the evolution of S. meliloti for the different markers we used. Recombination analyses allowed us to infer homologous recombination events between S. meliloti bv. meliloti and S. meliloti bv. medicaginis, showing that they are not sexually isolated. Lastly, we observed for IGSRKP, IGSEXO, and IGSGAB that s and d are almost the same, with the ratio of the two ranging from 7.32 x 10–1 to 8.64 x 10–1. According to Majewski and Cohan's theoretical model on the effect of ecological specialization on bacterial diversification (26), such a pattern is expected when the rate of recombination on the whole-species scale is the main factor that limits the divergence between ecotypes. In spite of disruptive selective pressure, sex makes the two groups poorly distinguishable for molecular markers not involved in ecological specialization. This suggests that divergence at nod loci between biovars neither prevents the transfer of this genomic area nor leads to an overall decrease of homologous recombination between S. meliloti genomes. It also suggests that no strong coevolution between genes involved in symbiotic specificity, in particular nod genes, and the core genome constrains gene exchanges among ecotypes. Within S. meliloti, the efficiency of nod gene variants would not be affected by the genetic background in which they are located.
Finally, to allow homologous recombination to occur, both S. meliloti biovars have to mate under suitable ecological conditions. Pretorius-Guth et al. (34) suggested that plant nodules create the most favorable natural conditions for recombination among S. meliloti isolates. Under this hypothesis, the maintenance of a common genetic background might thus involve coinfection of both S. meliloti biovars within mixed nodules, eluding the barrier to gene flow due to host plant specificity (17). Such events could indeed occur in natura since the coinoculation of nodulation mutants (Nod–) and infection mutants deficient in exopolysaccharide production (Inf– EPS–) during ex situ experiments results in mixed infected nodules (22). For such coinfection to occur, it is necessary that the host plant species of both biovars of S. meliloti cooccur locally, which is true for M. truncatula and M. laciniata (50). Alternatively, frequent disturbance events that would homogenize the rhizosphere could also facilitate such coinoculation events, and thus gene flow among biovars.
|
ACKNOWLEDGMENTS |
|---|
This research was funded by the Bureau des Ressources Génétiques, Programme Recherches Méthodologiques pour l'Amélioration des Processus de Gestion et de Conservation des Ressources Génétiques Animales, Végétales et Microbiennes, to I.O. G. B. X. Bailly was supported by a Ph.D. fellowship from the French Ministry of Education and Research and by a Jean and Marie-Louise Dufrenoy fellowship from the French Academy of Agriculture.
This is publication ISEM 2007.039 of the Institut des Sciences de l'Evolution, Montpellier, France.
|
FOOTNOTES |
|---|
Published ahead of print on 11 May 2007.
|
REFERENCES |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
