The semi-aquatic legumes belonging to the genus Aeschynomene constitute a premium system for investigating the origin and evolution of unusual symbiotic features such as stem nodulation and the presence of a Nod-independent infection process. This latter apparently arose in a single Aeschynomene lineage. But how this unique Nod-independent group then radiated is not yet known.
We have investigated the role of polyploidy in Aeschynomene speciation via a case study of the pantropical A. indica and then extended the analysis to the other Nod-independent species. For this, we combined SSR genotyping, genome characterization through flow cytometry, chromosome counting, FISH and GISH experiments, molecular phylogenies using ITS and single nuclear gene sequences, and artificial hybridizations.
These analyses demonstrate the existence of an A. indica polyploid species complex comprising A. evenia (C. Wright) (2n = 2x= 20), A. indica L. s.s. (2n = 4x = 40) and a new hexaploid form (2n = 6x= 60). This latter contains the two genomes present in the tetraploid (A. evenia and A. scabra) and another unidentified genome. Two other species, A. pratensis and A. virginica, are also shown to be of allopolyploid origin.
This work reveals multiple hybridization/polyploidization events, thus highlighting a prominent role of allopolyploidy in the radiation of the Nod-independent Aeschynomene.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The pantropical legume genus Aeschynomene of the Dalbergioid clade has long retained the attention of scientists, due to the ability of 22 semi-aquatic Aeschynomene species to develop nitrogen-fixing nodules not only on their roots but also on stems (Alazard & Duhoux, 1987, 1988; Chaintreuil et al., 2013). Stem nodulation is an unusual behavior amongst legumes that is shared with very few hydrophyte species of the genera Sesbania, Neptunia and Discolobium (Boivin et al., 1997; James et al., 2001). It enables these species to fix nitrogen even in flooded conditions (Alazard & Duhoux, 1987). Stem nodules always develop in the vicinity of adventive root primordia (Boivin et al., 1997), but only those of the Aeschynomene and Discolobium species are considered as genuine ones owing to their connection to the stem vascular system (Loureiro et al., 1994, 1995). Another unusual feature of these Aeschynomene species is their capacity to be nodulated by photosynthetic Bradyrhizobium strains (Evans et al., 1990; Giraud & Fleischman, 2004). The photosynthetic activity of these bradyrhizobia has been shown to facilitate ex planta survival and infectivity, and to directly supply energy to the bacterium that can be used for biological nitrogen fixation during stem nodulation (Giraud et al., 2000).
The Aeschynomene species are also distinguished by their symbiotic infection process. This does not involve root hair curling nor infection thread formation, but instead occurs in an intercellular fashion and is followed by the endocytosis of the symbionts (Bonaldi et al., 2011). Such an infection process is regarded as primitive but it persists in c. 25% of the legumes (Sprent, 2007). However, the most unexpected symbiotic feature encountered in certain Aeschynomene species is that they are nodulated by bradyrhizobia lacking the canonical nodABC genes required for the synthesis of Nod factors (Giraud et al., 2007; Miché et al., 2010). These lipochitooligosaccharide signal molecules are produced by all the other characterized rhizobia and were thus considered as the obligate key to symbiosis initiation. This Nod-independent process, along with an intercellular symbiotic infection, is similar to what has been observed in some Lotus japonicus double mutant lines altered both in Nod factor perception and signaling (Madsen et al., 2010; Bonaldi et al., 2011). Such an infection process was suggested to correspond to the ground state of the rhizobial symbiosis that was then maintained during evolution in some legumes such as Aeschynomene spp.
A recent phylogenetic study of the genus Aeschynomene revealed that the stem-nodulating species are distributed in several clades, whereas all the species able to be nodulated with bradyrhizobia lacking the nodABC genes belong to a unique clade. This clade includes no species that compulsorily require Nod factors to interact efficiently with symbiotic bradyrhizobia (Chaintreuil et al., 2013). Whether the single occurrence of the Nod-independent nodulation in the Aeschynomene genus corresponds to an ancestral or a derived character remains to be clarified, and this does not exclude the coexistence of the Nod-dependent and Nod-independent processes as suggested earlier (Madsen et al., 2010; Okubo et al., 2012). Unraveling the mechanisms of this Nod-independent symbiotic process would therefore bring important insights into the evolution of nodulation in this group (Sprent & James, 2008). For this purpose, we recently screened the Nod-independent Aeschynomene species for their genomic and genetic properties. and this led to the proposal of A. evenia as a novel model legume species because it displays all the characteristics required for genetic and molecular analysis (Arrighi et al., 2012).
During this screening, we also revealed an important variation in the genome sizes that suggested the ploidy levels may range from 2x to 6x for the Nod-independent Aeschynomene species (Arrighi et al., 2012). Among those for which the chromosome number is known, Aeschynomene indica was indeed shown to be tetraploid (Bairiganjan & Patnaik, 1989; Bielig, 1997). This raises the question of how polyploidy has contributed to radiation of the Nod-independent lineage. Polyploidy (whole-genome duplication) is recognized as an important process in plant evolution and is often invoked as a potential driver of angiosperm diversification (Soltis et al., 2009). It can act alone resulting in autopolyploids or in concert with hybridization producing allopolyploids. These former constitute an instantaneous mode of plant speciation due to the reproductive isolation from their progenitors (Soltis & Soltis, 2009).
Therefore, the Nod-independent Aeschynomene constitute a promising system to investigate the role of polyploidy in evolutionary radiation and to gain knowledge on the mechanisms of polyploid speciation. Here we focus on A. indica as a case study for such purposes. This is because this species is the most widespread in the tropics, it is frequently found as an adventive in rice culture, and it can be used as a forage legume, although toxicity to equines has been reported (Holm et al., 1997; http://www.fao.org). In addition, the tetraploid A. indica is closely related to the diploid model legume A. evenia and this latter has been well-characterized genetically (Arrighi et al., 2013; Chaintreuil et al., 2013). It appears to be morphologically similar to, and sometimes mistaken, for A. evenia (Rudd, 1955; http://www.fao.org). We also exploit genetic and genomic data on the Nod-independent Aeschynomene in order to understand the occurrence and mode of polyploid speciation in this group.
In the present study, we investigate the evolutionary relationships between A. evenia and A. indica and the genomic structure of A. indica. We have used the knowledge recently acquired on the intraspecific differentiation within A. evenia and characterized a set of 15 accessions designated as A. indica through molecular genotyping, genome size evaluation, molecular cytogenetic and phylogenic analyses. This work has revealed the existence of an A. evenia–A. indica species complex where three ploidy levels occur. Finally, we extended our investigations to the other Nod-independent Aeschynomene species to highlight the role of polyploidy in the evolution of this taxon.
Materials and Methods
All the accessions of Aeschynomene used in this study, their geographic origin and their sources are listed in Table 1. Seed germination, plant culture in glasshouse and hybridizations were performed as indicated in Arrighi et al. (2012).
Table 1. Accessions of Aeschynomene spp. used in this study, origin and characteristics
2C DNA content (pg)
Chromosome number (2n)
Ploidy level (x)
The data presented were obtained in this study unless specified: aArrighi et al. (2013); bArrighi et al. (2012); cIndex to Plant Chromosome Numbers (ICPN); Bairiganjan & Patnaik (1989); Bielig (1997).
This accession was provided by a Botanical Garden in the Netherlands but its origin is unknown.
Total genomic DNA was isolated from the leaves of plants using the CTAB extraction method (Doyle & Doyle, 1987). Taking advantage of an EST library available for the hexaploid A. indica line LSTM 19 (F. Cartieaux, unpublished data), already published SSR markers and new ones were tested and analyzed as performed in Arrighi et al. (2013) (Supporting Information Table S1).
Single nuclear genes were previously identified from the A. evenia IRFL6945 EST library (F. Cartieaux, unpublished data) and used for phylogenetic analyses in Arrighi et al. (2013): CYP1, eiF1α, Sucrose Synthase, SUI1 genes and a legume-specific gene homolog to Glyma07 g16420 and Glyma18 g37410 identified in Glycine max (Table S2). The sequence comprising the ITS-5.8S rDNA gene-ITS2 region was amplified using the primers listed in Table S2. PCR products with mixed sequences were cloned into pGEMT Easy (Promega) following the manufacturer's instructions. Individual transformed Escherichia coli colonies were used as template for subsequent PCR amplification and sequencing. The obtained sequences were deposited in Genbank (Table S3).
SSR marker genotyping data were processed to produce an UPGMA dendrogram and a similarity matrix calculated with the DICE coefficient as performed in Arrighi et al. (2013). For the phylogenetic analyses, the sequence of the five single nuclear genes was concatenated in order to obtain a well-resolved phylogeny. Concatenated single-gene sequences and ITS sequences were processed as previously described in Arrighi et al. (2013).
DNA content measurements
DNA measurements were done by flow cytometry as presented in Arrighi et al. (2012).
In situ hybridizations
Chromosomes were prepared from root tips and fluorescent in situ hybridization (FISH) was performed as in Arrighi et al. (2012). For genomic in situ hybridization (GISH), genomic DNA of A. evenia ssp. evenia (CIAT8232) was labeled with digoxigenin-11-dUTP and detected with Texas Red (red), while genomic DNA of A. scabra (PI 296044) was labeled with biotin-14-dUTP and detected with FITC (green). DNA labeling and chromosome hybridization were done using the protocol described in Guimaraes et al. (2008). For both the FISH and GISH experiments, fluorescent images were captured separately using a cooled high-resolution black and white CCD camera (Orca, Hamamatsu, Tokyo, Japan) and a Leica DMRAX2 fluorescence microscope. The camera was interfaced to a PC running the Volocity software (Perkin-Elmer, Tokyo, Japan). The same software was used for the deconvolution treatment.
Evidence for 3 main groups in A. indica
In order to identify the different genotypes present in A. indica, we analyzed fifteen accessions procured from different seed banks and originating from various regions of the world (Table 1a). We also included in this analysis four accessions of the closely related diploid A. evenia: two belonging to A. evenia ssp. evenia and two members of A. evenia ssp. serrulata that were previously characterized in Arrighi et al. (2013) (Table 1b).
This set of accessions was genotyped with 42 Simple Sequence Repeats (AiSSR) markers obtained from EST sequences available for A. indica LSTM19 (F. Cartieaux, unpublished data) (Table S1). Cluster analysis of the SSR data was then performed to produce an UPGMA dendrogram (Fig. 1). The A. indica accessions were split into three distinct groups designated as A. indica-1, -2 and -3; A. indica-1 was a separate group, whereas A. indica-2 and 3 formed sister groups. Notably, the A. evenia ssp. evenia accessions fell into the same cluster as the A. indica-1 group, whereas the two A. evenia ssp. serrulata accessions formed an outgroup that was well separated from the three others. The similarity, calculated using the DICE coefficient, ranged from 75% to 100% within each A. indica group indicating they are genetically homogeneous (Table S4).
During the accession genotyping, we noticed four relevant profiles that were obtained with some AiSSR markers (Fig. 1b, summary in Fig. 4b). Two profiles corresponded to SSRs with different band numbers, one (e.g. AiSSR23) with a single band for A. evenia and A. indica-1, and multiple bands for A. indica-2 and -3, the other (e.g. AiSSR15) with a single band for A. evenia and A. indica-1 and -2, and several bands for A. indica-3. Two other profiles occurred with SSR markers that displayed specific amplification. For one, SSRs (e.g. AiSSR17) showed a band only for A. indica-2 and -3, and, for the other, SSRs (e.g. AiSSR14) amplified uniquely in A. indica-3. Specific amplifications suggested that the different A. indica groups were either genetically well differentiated, as previously observed for the two A. evenia subspecies (Arrighi et al., 2013) or contained genomes of different origins. Multiband SSRs could be interpreted as different allelic versions or as indicative of the presence of several genomes.
Notably, A. indica-1, which clusters the A. evenia ssp. evenia accessions, always displayed a genotype of typical diploid species, that is, single-banded SSRs. Conversely, the concordant profiles for specific amplifications and multibandings favored the hypothesis that A. indica-2 and -3 are polyploids.
A. indica sp. is composed of three cytotypes sharing the same A. evenia ITS signature
In order to test this hypothesis, the ploidy status of the A. indica accessions was first investigated by cytometry.
The nuclear 2C DNA amounts obtained are presented in Table 1a,b. Accessions of the same group had similar 2C DNA quantities, whereas significant differences were evident between the three A. indica groups. Hence accessions of A. indica-1 displayed an average of 0.87 ± 0.04 pg 2C DNA, those of A. indica-2 an average of 1.67 ± 0.05 pg 2C DNA, and those of A. indica-3 an average of 2.63 ± 0.03 pg 2C DNA. The A. indica-1 group exhibited the same 2C DNA content as the accessions of the diploid A. evenia ssp. evenia (0.83 ± 0.01 pg 2C DNA). This suggested that the three A. indica groups correspond to a series of 2x, 4x and 6x forms. To firmly link the genome sizes to ploidy levels, one accession per group was used for chromosome counting (Table 1a, Fig. S1a,b,c). In accordance with the DNA content, A. indica-1 was found to be diploid (2n = 2x= 20), A. indica-2 was tetraploid (2n = 4x= 40) and A. indica-3 was hexaploid (2n = 6x= 60).
The A. indica species was initially described from an Indian sample (Rudd, 1955) and several cytogenetic studies have indicated that Indian accessions are tetraploid (Bairiganjan & Patnaik, 1989; Bielig, 1997). The same results were obtained here with the accessions belonging to the A. indica-2 group (Table 1a). Thus, the diploid cytotype appears to be closely linked to A. evenia, the tetraploid cytotype defines A. indica s.s. and the hexaploid cytotype corresponds to a form that has not been described thus far. To get a better insight into the genome structure of this new hexaploid form, we performed FISH experiments using both 5S and 45S probes and we compared the results with the diploid A. evenia IRFL6945 line. A. evenia showed one pair of 5S rDNA and two pairs of 45S rDNA loci with contrasted signal intensities that were both located in secondary constrictions of satellite chromosomes (Fig. 2a). For the hexaploid A. indica, six 5S rDNA sites and four 45S rDNA sites (two major and two minor, all located at secondary constrictions) were detected (Fig. 2b). Thus, the hexaploid form exhibited the same 45S rDNA signature as the diploid A. evenia and displayed additivity for the 5S rDNA loci that could be related to the ploidy level.
ITS analysis and phylogeny
In order to ascertain the phylogenetic relationship of the three A. indica forms and to resolve their relationship with A. evenia, the ITS1-5.8S rDNA gene-ITS2 sequence for one representative accession of each form was analyzed. There was evidence for intra-individual variations in ITS sequences in both diploid and hexaploid accessions as their chromatograms displayed several double peaks. The PCR products were then cloned and sequenced separately (Tables S2, S3). In each case, two sequences were obtained that were closely related, with divergence of just 1% over 704 bp. This situation was thus similar to what had been previously observed in some A. evenia accessions (Arrighi et al., 2013). The ITS sequences of the different A. indica groups (labeled ‘a’ and ‘b’ in the case of double sequences) were used and compared to the other Aeschynomene species to generate a phylogram (Fig. 3). They all fell in the same clade and strongly clustered with that of A. evenia ssp. evenia (node with bootstrap score of 86%). The ITS sequence of A. evenia ssp. serrulata was more distant and formed a distinct branch.
As a result, the three A. indica groups exhibited the A. evenia ssp. evenia ITS signature. Along with the same SSR profiles and the shared diploid status, this demonstrated that accessions of the A. indica group 1 corresponded in fact to A. evenia ssp. evenia. Moreover, this indicated that A. evenia ssp. evenia is at the origin of the two other A. indica groups that are polyploid.
Searching for the other genome donors in the polyploid A. indica forms
In order to determine the genome constitution of the polyploid A. indica (2n = 40 and 60), we searched for potential diploid progenitor species, other than A. evenia ssp. evenia, using three complementary approaches.
Cross-amplification tests with AiSSRs
In a first approach, we selected the AiSSR markers developed for the hexaploid A. indica LSTM19 line that displayed group-specific amplification patterns (Fig. 1b). They were used to perform cross-amplification tests on the other Aeschynomene species. Most of the markers were amplified in most species. This could be explained by their genetic proximity because they were previously shown to belong to the same phylogenetic clade (Chaintreuil et al., 2013). However, two groups of AiSSRs exhibited a very restricted amplification profile. The first one corresponded to four AiSSRs amplifying in both the tetraploid and hexaploid A. indica forms as well as in A. scabra (and in A. virginica; see the section on ‘Molecular analysis for A. virginica’) (Fig. 4a, markers listed Fig. 4b). The second group gathered five AiSSRs that amplified only in the hexaploid A. indica form and among the A. evenia ssp. serrulata/A. denticulata/A. ciliata species group (Fig. 4a, markers listed Fig. 4b). This showed that the A. indica tetraploid and hexaploid genomes share specific loci with diploid Aeschynomene species other than A. evenia ssp. evenia.
Molecular phylogeny using single nuclear genes
We took advantage of the available EST libraries for both the hexaploid A. indica LSTM19 line and the diploid A. evenia IRFL6945 (F. Cartieaux, unpublished data) to perform gene comparisons. In silico analysis revealed three homologous versions for many genes of A. indica 6x when only one version was encountered in A. evenia (Table S2). Five genes were selected for further analysis: cyclophilin 1 (CYP1), eukaryotic translation initiation factor 1A (eiF1α), sucrose synthase, translation factor SUI1 genes and a legume-specific gene homolog to Glyma07 g16420 and Glyma18 g37410 identified in Glycine max (Table S2). For each gene, two sequences were obtained after cloning for the tetraploid A. indica PI196206 line and only one in the other diploid Aeschynomene species. The five genes treated separately gave similar results when subjected to maximum likelihood Bayesian analysis (not shown). In order to produce a well-resolved phylogeny, their coding sequences were concatenated into a unique 2070-bp sequence. In the resulting tree, sequences of the A. indica polyploids were distributed in three lineages (A, B, C) (Fig. 5). Lineage A includes the sequences isolated in the tetraploid and hexaploid forms. They strongly cluster with those of A. evenia ssp. evenia (node with a bootstrap of 99%). Lineage B also groups sequences obtained with the 4x and 6x forms and they appear very similar to those of A. scabra (node with a bootstrap of 79%) and A. virginica (see the section on ‘Molecular analysis for A. virginica’). Lineage C corresponds to the sequences isolated only in the hexaploid cytotype. It formed an isolated branch, rising between those of A. denticulata and A. ciliata. These results strongly supported an allotetraploid origin for A. indica 4x that would derive from the diploid A. evenia ssp. evenia and A. scabra. The A. indica 6x most probably has an allohexaploid origin involving the two genomes present in the 4x form and another unidentified third diploid progenitor, related to A. denticulata and A. ciliata.
To confirm the results obtained with the molecular data, GISH was carried out on chromosome preparations of A. indica 4x and 6x using labeled genomic DNA of both putative progenitors. The biotin-labeled probe of total DNA from A. evenia ssp. evenia was detected with Texas Red (red), the digoxigenin-labeled probe of total DNA from A. scabra was detected with FITC (green), and the hybridized chromosomes were counterstained with DAPI. The results showed that, for both the A. indica 4x and 6x forms, the two genomic DNAs hybridized mainly to the centromeric regions as well as the secondary constrictions of four satellited chromosomes corresponding to 45S rDNA sites. Almost no hybridization could be observed along the chromosome arms (Fig. 6a,b). Most of the chromosomes were labeled in both red (biotin) and green (digoxigenin) indicating important sequence homology with the two DNA probes and thus close relationship with and between the putative progenitors (Fig. 6a,b). However, notable differences in fluorescence intensity were distinguishable among most of the chromosomes and the overlay of the Texas Red with FITC signals allowed for both the tetraploid and hexaploid A. indica to distinguish two groups of chromosomes (Fig. 6a,b). Half of these chromosomes showed a predominantly red fluorescence, suggesting a higher affinity for the genomic DNA probe from A. evenia ssp. evenia, whereas the other half was rather green indicating higher affinity for the genomic probe from A. scabra (Fig. 6a,b). In the case of the hexaploid A. indica, nonlabeled chromosomes, which might have corresponded to the third genome, are not observed. This probably reflected the low genome divergence of the progenitor species.
Identification of other allopolyploidization events in the Nod-independent Aeschynomene
Ploidy analysis in other Aeschynomene species
Comparison of 2C DNA values with known ploidy levels in the Nod-independent Aeschynomene species revealed that, in addition to A. indica, three others displayed a genomic DNA content (> 1.5 pg/2C) compatible with polyploid genomes. These species were A. deamii (1.93 pg/2C), A. pratensis (3.09–3.19 pg/2C) and A. virginica (2.08 pg/2C) (Table 1b). For these species, the ploidy status was ascertained by direct chromosome counting. In mitotic root cells of A. deamii 2n = 20 chromosomes were counted, revealing this species to be diploid (Table 1b; Fig. S1d). Interestingly, the average size of the metaphasic chromosome for A. deamii was found to be 2.31 μm ± 0.46, whereas those of A. evenia ssp. evenia were 1.26 μm ± 0.15 long, accounting for the difference in genome size between these two diploid species (1.93 pg/2C and 0.85 pg/2C respectively). In contrast to A. deamii, both A. pratensis and A. virginica were found to be tetraploid with 2n = 40 chromosomes (Table 1b; Fig. S1e,f). The nature of their polyploidy was therefore investigated.
Molecular analysis for A. pratensis
Preliminary analysis of A. pratensis accessions with RAPD markers had shown two genetic profiles with discrete differences (data not shown). One accession for each RAPD profile was further characterized by sequencing the ITS region (Table 1b). The ITS analysis revealed that the accession CIAT 8190 contained two ITS sequences: one corresponded to the A. pratensis ITS sequence found in the accession IRRI 13 006 whereas the second was highly similar to the A. sensitiva ITS sequence (Fig. 3). Amplification of the five nuclear genes used above revealed single sequences for the CYP1 and the legume-specific gene, and two sequences for the eiF1α, Sucrose Synthase and SUI1 genes. No significant difference was observed between the two A. pratensis accessions analyzed, so only the IRRI 13006 accession data were kept for the molecular phylogeny. Each gene gave similar results when subjected separately to maximum likelihood Bayesian analysis. Therefore, A and B versions could be identified and they were combined into two sequences with missing expected sequences indicated as missing data in the data matrix analyzed. In the resulting molecular phylogeny, both the A and B concatenated sequences fell in the same clade as A. sensitiva and A. selloi. Although the A concatenated sequences formed a single branch, the B concatenated sequences closely clustered with those of A. sensitiva (Fig. 5). The concordance of the ITS and single nuclear genes data suggests that the tetraploid A. pratensis originated from a cross between A. sensitiva and another related genome donor that remains to be identified.
Molecular analysis for A. virginica
As previously indicated, we identified some AiSSR markers amplifying specifically in the tetraploid and hexaploid A. indica as well as in A. scabra, but also in A. virginica (Fig. 4a,b). This was in apparent contradiction with the phylogenetic analysis based on the ITS sequence that showed A. virginica to be phylogenetically closer to A. rudis than to A. scabra (Fig. 3). Analysis of the five nuclear genes, the CYP1, eiF1α, Sucrose Synthase and SUI1 genes, and the legume-specific gene, revealed two sequences for each. The A versions closely clustered with A. scabra and the B versions with A. rudis. They were combined to be included in the molecular phylogeny (Fig. 5). This suggested the tetraploid A. virginica was most certainly the result of hybridization between the diploid A. scabra and A. rudis.
Interspecific hybridization experiments
In order to test the extent to which interspecific crossings can occur among the Nod-independent species, we performed artificial hybridizations with the putative progenitors of the tetraploid A. indica and A. virginica; that is, between A. evenia ssp. evenia and A. scabra and between A. scabra and A. rudis, respectively. Fertilizations were found to be successful as indicated by the development of pods. However, most of them aborted and fell before reaching their full size (Fig. 7a). One pod resulting from an A. evenia ssp. evenia × A. scabra cross reached maturation, but it produced wrinkled and nonviable seeds (Fig. 7a). By contrast, two fully developed pods were obtained after A. scabra × A. rudis crosses, containing healthy seeds. They readily germinated and produced vigorous plants that abundantly flourished, but without producing seeds (Fig. 7a). The hybrid status of these F1 plant was further checked using polymorphic AiSSR markers (Fig. 7b). These results indicate that despite the genetic divergence between Aeschynomene species belonging to the Nod-independent clade, rare interspecific hybridizations can occur. However, concurrent genome duplication may be necessary to overcome the reproductive block observed and allow the subsequent establishment of newly formed allopolyploids.
The Nod-independent lineage within the semi-aquatic Aeschynomene represents a group of special interest by which to understand the evolution of the nitrogen-fixing symbiosis. Despite such interest, its evolutionary origin and radiation were obscure. Here, we addressed how this group diversified by studying the role of polyploidy in species formation. Special emphasis was laid on the pantropical A. indica that is closely related to the well-characterized diploid A. evenia and for which abundant genetic resources are available. The characteristics of the newly unraveled A. evenia–A. indica polyploid species complex are discussed in line with those of other polyploid Aeschynomene species.
A. evenia and A. indica form an allopolyploid species complex
The A. indica accessions we screened were shown to represent three forms genetically well-differentiated and corresponding to different cytotypes (2n = 20, 40 and 60) that display the same A. evenia ssp. evenia signature. These data confirmed the genetic relationship linking A. evenia ssp. evenia and A. indica and further revealed they are part of a polyploid species complex where the 2x form corresponds to A. evenia ssp. evenia, the 4x form defines A. indica s.s., and the 6x form represents a so far undescribed form. This can explain why these species share a high similarity for morphological traits and are then sometimes misidentified (Rudd, 1955; http://www.fao.org). The difficulty in distinguishing species in polyploid complexes due to little or overlooked morphological differentiation is a recurrent problem as already reported notably for Stylosanthes scabra (Chandra & Kaushal, 2009), Hordeum murinum (Ourari et al., 2011) and Mercurialis annua (Obbard et al., 2006).
For polyploid species the use of genetic markers as well as molecular phylogenies based on the sequence of nuclear single-copy genes have been valuable tools to investigate their genomic structure and to identify among diploid species those representing potential progenitors (Coffea arabica: Lashermes et al., 1999; Arachis hypogaea: Jung et al., 2003; de Carvalho Moretzsohn et al., 2004; Ma et al., 2004; Stylosanthes scabra: Chandra & Kaushal, 2009; Hordeum murinum: Jacob & Blattner, 2010; Mercurialis annua: Korbecka et al., 2010). In the case of the polyploid A. indica forms, these two approaches unambiguously indicated that in addition to A. evenia ssp. evenia, A. scabra was involved in the formation of both A. indica 4x and 6x. The third genome donor of the hexaploid form could not be identified as it did not match to any tested Aeschynomene taxon in this study, but it was found to be closely related to A. denticulata and A. ciliata. Whether the putative third genome donor is nowadays extinct or remains to be identified is an open question. Prospecting missions and testing more accessions could help in resolving this issue.
The genome organization of A. indica 4x and A. indica 6x was confirmed by GISH using genomic DNA from A. evenia ssp. evenia and A. scabra as probes. For each, the two subgenomes could be differentiated with hybridization signals found mainly on centromeres and rDNA sites. Such hybridization patterns were very similar to those observed for Coffea arabica and Musa spp. (Lashermes et al., 1999; D'Hont et al., 2000). This is in contrast with what was observed in other allopolyploid species like A. hypogaea for which GISH experiments gave uniform chromosomal hybridization patterns (Seijo et al., 2007; Guimaraes et al., 2008). This phenomenon has been explained by D'Hont (2005): considering that GISH is essentially based on the detection of repeated sequence and that in small genomes these repeated sequences are mainly concentrated in peri- and centromeric regions, difficulties in obtaining an even GISH labeling may be encountered when the average chromosome size is c. 0.06 pg/chromosome or less. A. indica 4x and 6x have chromosomes with an average size of 0.045 pg and therefore they are below this critical value.
Another notable genomic feature revealed by cytogenetic analysis was that the hexaploid A. indica exhibited locus number additivity for the 5S rDNA loci (6 signals), according to the number observed in the related diploid A. evenia (2 signals). On the contrary, the number and characteristics of the 45S rDNA signals were the same in both species (2 strong signals, 2 weaker signals). GISH analysis of the tetraploid A. indica also revealed four satellites corresponding to the 45S rDNA loci. This suggests that structural changes occurred (i.e. gene loss) following polyploidization to maintain the number of 45S rDNA loci constant. Similar cases of rDNA locus additivity or deletions have been reported in other polyploid species. For example, Hordeum murinum is characterized by 5S and 45S locus loss (Ourari et al., 2011) ,whereas Arachis hypogaea displays additivity of both 5S and 45S loci (Robledo et al., 2009; Robledo & Seijo, 2010). The loss of 45S rDNA loci in A. indica is linked with a homogenization of the rDNA sequences because both the tetraploid and the hexaploid forms displayed only ITS sequence of the progenitor A. evenia ssp. evenia. Such a process may be mediated by unequal crossing over leading to locus loss and interlocus gene conversion, mechanisms frequently found in polyploids and collectively referred to as concerted evolution (Kovarik et al., 2005).
Overall, these data show that A. evenia, A. indica 4x and 6x are part a species complex that arose by allopolyploidy (Fig. 8). Results are consistent with a scenario where A. evenia ssp. evenia and A. scabra are involved in the formation of the tetraploid A. indica (s.s.). In turn, the tetraploid A. indica most likely hybridized with the third genome donor to produce the hexaploid A. indica. But, although this evolutive model appears to be the simplest, we cannot rule out the existence of other Aeschynomene lineages that could have provided the different genomes amongst the current living A. indica polyploids.
Allopolyploidy is a recurrent speciation process in the Nod-independent Aeschynomene
The Nod-independent nodulation process is shared to date by only 14 analyzed Aeschynomene species that were shown to belong to the same lineage (this study; Chaintreuil et al., 2013). To investigate whether, in addition to the A. evenia–A.indica species complex, other polyploidization events occurred and thus contributed to the species diversification of the Nod-independent lineage, we screened Aeschynomene species for their ploidy levels and identified two other tetraploid species: A. pratensis and A. virginica.
Aeschynomene pratensis is a perennial species present in the Caribbean area as well as in South America (Rudd, 1955). In the present study, we characterized accessions of the commonly found A. pratensis var. caribaea. Nuclear gene sequence analysis revealed the existence of A and B versions, with the latter being highly similar to A. sensitiva. This suggested A. sensitiva to be one of the progenitors of A. pratensis (Fig. 8) and thus corrobated the observations of Rudd (1955) who noticed very close morphological relatedness between of A. pratensis var. caribaea and A. sensitiva. This proximity was further supported by the observation that the Colombian A. pratensis accessions harbored two ITS sequences: one corresponding to A. pratensis and the other to A. sensitiva. Whether this situation corresponds to a concerted evolution that has not proceeded to completion as for the A. indica polyploids, or betrays a possible introgression of the diploid A. sensitiva in A. pratensis, remains an open question. Molecular and cytogenetic analysis of A. pratensis var. pratensis, a rare and threatened plant endemic to South Florida, might help unraveling the genomic story of A. pratensis.
Aeschynomene virginica is a rare annual species, native to the East Coast of the United States. It occurs in freshwater tidal marshes with populations scattered in small stands. To ensure its maintenance, A. virginica has been the subject of several studies and recovery plans (Griffith & Forseth, 2002, 2003, 2005, 2006). Our gene sequence analysis has unraveled the genomic structure of A. virginica, identifying A. scabra and A. rudis as its two progenitors (Fig. 8). The discovery of its allotetraploid origin should help in further analysis of its population dynamics and genetics. Along with A. indica 4x for which the two parental species are also identified, A. virginica shows good genome additivity (genome reduction < 0.1% for A. virginica and < 10% for A. indica 4x), coherent with little genome rearrangement from the parental species. This is in accordance with the low gene sequence divergence between the constitutive genomes of these two tetraploid species and those of their progenitor species, suggesting that the polyploid speciations probably took place relatively recently.
This prompted us to perform artificial hybridizations between the parental species of A. indica 4x and A. virginica. Manual hybridizations usually led to aborted pods. Only crossings with the progenitors of A. virginica produced viable seeds that gave healthy but completely sterile hybrids. This situation is reminiscent of what was previously observed when crossing A. evenia ssp. evenia with A. evenia ssp. serrulata (Arrighi et al., 2013). Similarly, in the Trifolium genus, the production of interspecific hybrids was shown to be possible, but it always required embryo rescue and subsequent chromosome doubling to restore fertility (Williams et al., 2012). These results indicate that the Aeschynomene species have undergone genome differentiation leading to reproductive barriers and genomic incompatibilities. This idea is strengthened by the observation that the parental species of the Aeschynomene polyploids always belong to the same subgroups within the Nod-independent lineage, suggesting that hybridization is possible only between closely related species.
However, it is worth noting the success of interspecific hybridizations can be genotype dependent as suggested by Weiss-Schneeweiss et al. (2011) from their observations on the allopolyploid speciations in the American daisies. In the Aeschynomene genus, A. scabra appears to be involved in the formation of both A. indica and A. virginica, indicating that this species may be more prone to participation in allopolyploid formation. Indeed, when performing interspecific hybridizations, our crosses were successful only when A. scabra was used as female and A. evenia or A. rudis as male (JF Arrighi, personal observations). This suggests that A. scabra may have the propensity to ‘tolerate’ exogenous pollen, predisposing to allopolyploid formation. Although Aeschynomene species are autogamous – that is, they reproduce by self-pollination – several species are known to be pollinated by insects. In particular, A. virginica flowers are visited by Least Skipper (Ancylozypha numitor) and leaf-cutter bees (family Megachilidae) (www.centerforplantconservation.org). Such pollinating insects may therefore mediate outcrossings by providing pollen from one Aeschynomene species to another sympatric one.
This study sheds light on the mechanisms of diversification in the Nod-independent lineage within the Aeschynomene genus by showing that hybridization and polyploidization have played a significant role in species formation (summarized in Fig. 8). In particular, it unravels an A. evenia–A. indica polyploid complex with three ploidy levels. Within the Aeschynomene genus, this is the only group to be present in all tropical and subtropical areas. An initial assessment of the origins of the accessions tested in this study suggests that the distributional range of the 2x, 4x and 6x forms are different. This opens the way for a study into whether a combination of polyploidy and geographical differentiation/speciation has driven cladogenesis in this group. This will need further genetic studies to provide insight into the mechanisms of its large-scale radiation.
We are grateful to L. G. Santos Meléndes (CIAT, Columbia) and to S. Norton (AuspGRIS, Australia) for provision of the new A. indica accessions used in this study. We thank M. Lartaud (UMR AGAP, CIRAD, Montpellier, France) for the deconvolution of the confocal images and M. Bourge (Imagif Platform, CNRS, Gif-sur-Yvette, France) for his help in cytometry.