A. caulinodans (azc) is a nitrogen fixing member of the Xanthobacteraceae family. It is primarily a stem nodulator of the African legume sesbania. It is thought to have originally been a non-nitrogen fixer that developed the ability to fix N₂ entirely through lateral gene transfer from another unknown species [7]. Unlike the other thirteen strains from this study, azc can reduce di-nitrogen both in symbiosis and in the free living stage, and as a result can be grown on nitrogen free medium, a key defining property. It has a relatively small genome—5.37 MB—in comparison to other rhizobia and it is the most taxonomically distant species in this study [7,8].

Three Bradyrhizobium genomes have been selected, B. japonicum USDA 110 (bja), Bradyrhizobium sp. BTAi1 (bbt), and Bradyrhizobium ORS278 (bra). Bradyrhizobium is primarily distinguished from the genera Mesorhizobium, Rhizobium and Sinorhizobium by the slower growth rate (doubling time) of at least 8 h.

The soybean root nodulator bja has the largest genome of this study, 9.1 MB, in a unique chromosome [9]. The main feature of the more recently obtained genomes of Bradyrhizobium strains BTAi1 (bbt) and ORS278 (bra) is that they are photosynthetic stem nodulators of tropical Aeschynomene species, and for the first time described as lacking common nodulation genes nodA, nodB and nodC. These genes are traditionally involved in the construction of the Nod factor backbone [10,11]. Both strains also have large genomes, similar to bja (Table 2).

As the suffix Meso suggests, species of this genus show growth rates intermediate to Bradyrhizobium (>8 h) and Rhizobium/Sinorhizobium (<6 h). Mesorhizobium loti MAFF303099 (mlo) is a root nodulator of Lotus japonicum, with one chromosome and two plasmids; as in other Mesorhizobium and Bradyrhizobium, mlo has a symbiotic island within the chromosome that contains all key genes for nodulation and nitrogen fixation [12].

Mesorhizobium sp. BNC1 (mes), formerly known as Agrobacterium sp. BNC1, and alternately named Chelativorans sp. BNC1, is the functional outlier of the fourteen species, being asymbiotic. It was isolated from a mixed-culture enriched from sewage using the chelating agent EDTA as the sole carbon and nitrogen source [13]. For this study, mes is utilized as an outlier for the basis of “symbiome” definition and investigation.

The genus Rhizobium, originally defining all bacterium with the ability to nodulate legumes [14], has undergone multiple redefinitions and now encompasses a variety of fast growing nitrogen fixers including some former Agrobacterium and Sinorhizobium (Ensifer) species and the genus Allorhizobium. In this study, five Rhizobium genomes are investigated: R. etli CFN 42 (ret), R. etli CIAT 652 (rec), R. leguminosarum bv. trifolii WSM1325 (rlg), R. leguminosarum bv. trifolii WSM2304 (rlt) and R. leguminosarum bv. viciae 3841 (rle). These five Rhizobium strains have different origins and hosts. The R. etli strains nodulate common bean (Phaselous vulgaris and were originally isolated in Central America) [15,16,17,18]. In contrast, the R. leguminosarum strains nodulate temperate legumes: bv. trifolii rlg and rlt are symbionts of clovers (Trifolium), although individually matched to annual (rlg) and perennial (rlt) species for N fixation [19,20], Bv. viciae rle has a broader host range, nodulating legumes within the tribe Viciae, including Pisum, Vicia, Lathyrus and Lens [21].

The genus Sinorhizobium is subject of some controversy. Members of this genus were originally classified as Rhizobium (e.g., Rhizobium meliloti) and the new genus was proposed in 1988 [22]. More recently, the genus has been reclassified as Ensifer [23,24], but in general the new nomenclature was not broadly accepted by rhizobiologists. Sinorhizobium and Rhizobium have similar morpho-physiological properties, and only the 16S rRNA taxonomy resolves the genus as a distinct clade [25].

Three Sinorhizobium were chosen for this study, S. meliloti 1021 (sme), S. medicae WSM419 (smd) and Sinorhizobium sp. strain NGR 234 (rhi), also denominated as S. fredii NGR 234 and Rhizobium sp. NGR 234 [26,27,28]. Both sme and smd nodulate temperate legumes of the genus Medicago. It has been established that smd has greater acid tolerance, and is a more effective nodulator of Medicago truncatula than sme [29,30]. It is also hypothesized that sme and smd evolved in association with hosts adapted to different edaphic conditions [31]. The tropical strain NGR 234 (rhi) is probably the most intriguing rhizobia isolated so far. Originally isolated in Papua New Guinea and labeled Rhizobium sp. NGR 234, it has since been found to be highly promiscuous, capable of nodulating at least 112 different legume genera, from most Vavilov centers of origin, in addition to the non-legume Parasponia [32,33,34].

The phylogeny and taxonomy of nitrogen fixing Rhizobiales is in a state of flux. Confusion over the status of Sinorhizobium [22,23,35] and the split phylogeny of Agrobacterium within the genus Rhizobium [24,36] are just two cases in point. Therefore in this genomic comparison several methods of taxonomic and genomic comparison were utilized to highlight different aspects of the phylogeny of the selected strains.

The ‘classical’ relationship between Bradyrhizobium, Rhizobium, Sinorhizobium and Mesorhizobium is clearly shown in the 16S rRNA phylogeny (Figure 1). The four major groups are delineated from each other forming four major clusters, with Azorhizobium as a separate branch. All strains were clustered within the expected genera and as expected the Trifolium nodulators rlg and rlt were indistinguishable from each other using exclusively the 16S rRNA.

Further comparison of whole chromosomes adds valuable information to the 16S rRNA taxonomy results. A dotplot of the chromosome nucleotide sequences was compiled using an in-house application called FRECKLE (Figure 2) (http://code.google.com/p/freckle/). Two primary blocks of significant sequence similarity are those of the Bradyrhizobium species with another larger block consisting of the Rhizobium and Sinorhizobium strains, while Azorhizobium and Mesorhizobium are outliers. The sme chromosome shows a highly repetitive structure, more than any other chromosome. Interestingly, the dotplot also displayed a strong sequence similarity between Sinorhizobium NGR 234 and S. meliloti 1021, including a shared repetitive structure.

A dotplot of plasmids (Figure S1) shows markedly less homology. Greater symmetries were detected in Rhizobium non-symbiotic plasmids: plasmids p42e (R. etli CFN 42), pRL11 (R. leguminosarum bv. viciae 3841), pRLG202 (R. leguminosarum bv. trifolii WSM2304) and pR132502 (R. leguminosarum bv. trifolii WSM132502) are especially homologous, suggesting higher stability. In addition, pSMED01 (S. medicae WSM419) and pSymB (S. meliloti 1021) also show symmetry. This observation has been previously explored, resulting in the proposal that these plasmids—with little or no direct involvement in the symbiosis—represent secondary chromosomes [37], or alternatively ‘chromids’ [38], as they contain essential metabolic genes, as well as tRNA and/or rRNA.

In contrast, little homology was detected in the analysis of the symbiotic plasmids (Figure S1). Symbiotically important plasmids have similar nod, nif and, in some cases, fix clusters. It is hypothesized that multiple lateral transfer events, which originally enabled the symbiotic and nitrogen fixing ability of many of these bacteria, would have also caused the overall lack of homology amongst the symbiotic plasmids. Our study has pointed out that the rhi plasmids pNGR234a and pNGR234b display lesser homology with the other plasmids, suggesting a greater number of lateral transfer events. Such ‘elastic’ genomes could partly explain the adaptation to a wide host range of rhi.

The contrast between the 16S rRNA taxonomy and the two dotplots (chromosomal and plasmids) relies on the disparate relationships between the plasmids. In fact, Young et al. 2006 concluded that R. leguminosarum bv. viciae (rle) had a core genome, mostly chromosomal with a high G+C content and shared with other organisms, and an accessory genome with a lower G+C content, on plasmids and chromosomal genomic islands [21]. The broader whole genome taxonomy in this study suggests that this two-component genome model may prevail in other rhizobial species. The R. etli and R. leguminosarum bv. trifolii genomes species share much of their chromosome with R. leguminosarum bv. viciae, further expanding the “core—genome” definition of Young et al. 2006, and, in turn, shrinking the “accessory genome”.

Most important, the low homology observed in this study in the comparison of the genomes of both closely related strains within the same species and also between species showing reasonable similarity of the 16S rRNA, and clearly showing several events of lateral gene transfer reinforces the discussion about the validity of prokaryote species definition [39,40].

The next stage of our genomic study was to compare KEGG genes and pathways in each species. The summary of KEGG orthologs is shown in Table 3, and the full table of KEGG pathways is included as Table S1.

All fourteen genomes have a large number of nitrogen, methane, sulfur, amino acid, vitamins and cofactors metabolism KEGG orthologs. This diverse suite of metabolism pathways is indicative of the ability to live in the complex rhizosphere environments, as well as to adapt to the nodule environment. However, the fact that the non-symbiotic genome of mes shows the same wide range of KEGG metabolism pathway orthologs as the symbiotic mlo, indicates the role of these genes in determining saprophytic capacity in a broad range of soil conditions. Noteworthy was the wide range of nitrogen metabolism protein orthologs detected (Table S2), showing a variety of nitrogen metabolism capabilities. The largest number of orthologs in Bradyrhizobium and Sinorhizobium indicate the capacity to metabolize nitriles, nitrates, formamide, nitroalkanes, as well as related amino acids glutamine and asparagine.

The genomes of Mesorhizobium and Rhizobium leguminosarum lack norC, norD and norE orthologs, protein subunits of the nitric oxide reductase complex. Interestingly, within the same species R. etli, ret but not rec contains asparagine synthetase (asnB) and nor orthologs. R. etli CFN 42 (ret) is unable to use nitrate for respiration and lacks nitrate reductase activity, as well as the nap and nar genes encoding respiratory nitrate reductase; however, the strain carries proteins closely related to denitrification enzymes, norCBQD. The functionality of nor genes in CFN 42 has been recently demonstrated, allowing the reduction of nitrite and nitric oxide [41]. It is worth mentioning that differences between strains within the same species made visible only by comparative genomics can help to explain metabolomic advantages, e.g., in this case of ret over rec.

A comprehensive method of comparative genome analysis is to cluster protein families with the BlastlineMCL algorithm [42], representing a Markov clustering of all orthologous protein groups across species. This could be considered a more complete clustering of orthologous proteins than from KEGG, as it includes all proteins annotated across all the genomes, not just the more restricted set of established KEGG orthologs. The use of this tool has been shown to be crucial in the comparison of rhizobial species, as many symbiotic and nitrogen fixing related gene pathways and families are not represented in KEGG.

Cluster information resolved from BlastlineMCL analysis is available online (see experimental section), where users can compare any combination of genome protein family homology to search for orthologous protein groups of interest. Raw sequence data is available in FASTA format and CLUSTALW aligned format (MSF). This resource represents a vast repository of comparative genomics information on the fourteen genomes, of which only a few are discussed in this paper.

A nitrogen-fixing Rhizobiales “pan-genome” of 1,126 clusters and 28,110 proteins was resolved amongst the fourteen genomes (Table 4). The clusters within this group include ABC transporters, some transposases, ribosomal RNA synthetases, DNA polymerases and other core proteins. The ABC transporters are the second most abundant family of non-hypothetical protein encoding genes found in all sequenced prokaryotic, eukaryotic, viral genomes as well as in metagenomic sequences. ABC transporters represent one of the largest superfamilies of active membrane transport proteins (MTPs), with a highly conserved ATPase domain that binds and hydrolyzes ATP, supplying energy for the uptake of a variety of nutrients and for the extrusion of drugs and metabolic wastes [43]. It is therefore not surprising that by far the biggest cluster found in the BlastlineMCL analysis is that of ABC type transporters, with 1,128 separate proteins. Together with the previous KEGG findings, this is indicative of ability to use a very wide range of substrates, a key property for adaption in the variety of environments, such as the numerous geographical rhizospheres represented (such as China, Papua New Guinea, Uruguay and Mexico) and the diverse hosts they nodulate.

The number of clusters shared amongst all thirteen nitrogen fixers was of 105, containing a total of 2,038 proteins. Many of these clusters are of nif proteins, the components of the nitrogenase complex. Nine protein clusters were shared by the eleven strains containing the nodABC genes, with a total of 113 proteins. These clusters include other nod, nol and noe gene families, protein families crucial to nodulation.

Table 4 lists many other cluster sharing combinations. For example, 857 clusters are exclusively shared by the Bradyrhizobium species, confirming the differences between this genus and the others. In addition, the 242 clusters containing 767 proteins—60% in chromosomes—found only in rhi, smd, sme confirm rhi as a Sinorhizobium species.

The number of singleton clusters (clusters containing proteins only from one species) approximately correlated with previously resolved taxonomic divisions, as well as genome size (Table 4). The largest number of singletons was found in the biggest genome, bja (1,760 clusters containing 1,839 proteins). The number of singletons amongst the various Rhizobium and Sinorhizobium was much smaller, ranging from 758 clusters containing 797 proteins in rhi to 333 clusters containing 344 proteins in rec. The most common protein family found amongst these singletons is composed by hypothetical proteins.

Induction of nodulation genes leads to the production and secretion of return signals, the Nod Factors, which are lipochitooligosaccharides of variable structure. Nod factors are essential for the Rhizobiales to trigger root hair curling, to induce the formation of nodule primordia, and to enter the root via infection threads. Purified Nod factors are sufficient to induce root hair deformations, cortical cell divisions, and on some host plants, fully grown nodule-like structures. NodD is the core signaling protein, reacting to plant flavonoids then binding to nod boxes, binding sites upstream of nod genes, typically nodA and/or nodB, triggering the expression of a nod gene cascade and thus the construction of the Nod Factor (e.g., [46]).

The results of BlastlineMCL clustering were utilized to the extraction of all nod protein ortholog clusters (Table S3). In total, twenty five different Nod protein ortholog clusters amongst the fourteen species were found. All fourteen species contained nodD, nodE, nodG, nodI, nodM, nodP, nodQ, nodV and nodW. Unsurprisingly, the non-symbiotic mes had the least number of nod genes, with ten. In a previous study, nodD and nodM have also been indicated as common genes of bacteria of the order Rhizobiales, but including both symbiotic and pathogenic stains [5].

One of the Nod orthologs found in all species is NodG, which is in the second largest MCL cluster. It is a cation protein exporter, involved in the secretion of the finished Nod factor into the environment. It is noteworthy that in R. tropici strain PRF 81 nodG can be promptly transcribed—after only 5 min—in the presence of host specific flavonoids [47].

Two nod genes were found exclusively in one strain: nodR in R. leguminosarum bv. trifolii WSM1325 and nodO in R. leguminosarum bv. viciae 3804. The function of nodR is unknown; however, it is suspected that NodR might contribute to the superior host nodulation efficiency of WSM1325, compared to WSM2304 [20]. In relation to NodO, it catalyzes the addition of carbamoyl to the Nod factor backbone [48], and it is also a calcium-binding protein that promotes infection thread development in root hairs [49]. Interestingly, in Rhizobium sp. BR 816 (reclassified as Sinorhizobium) the transfer of nodO can extend its host range [50], therefore it might be involved in the ability of rle to nodulate a wide range of species within the Viciae tribe.

While all species have many core and non-core nod genes, an investigation of nod operon/gene structure in the thirteen symbiotic genomes displays a wide variety of arrangements. The Sesbania nodulator azc has a large nod operon nodABCSUIJZnoeCHOP with nodD1 located downstream [7]. The soybean nodulator bja also has a large nod operon nodD1-YABCSUIJ [9]. Surprisingly, the other two Bradyrhizobium, bbt and bra, have no nodA, nodB or nodC genes [11], and the signaling mechanisms with the host plant are still under investigation [10].

The three R. leguminosarum and the two Sinorhizobium genomes studied have similar initial nod operon structures, nodDABCIJ. In R etli, the arrangement differs with nodD, nodIJ, a 200 bp gap and then nodCB; the gene nodA is 16 kb downstream. Finally, in mlo, there is a separate nodAC and nodIJ operon, with nodD and nodB in disparate regions of the 600 kb symbiotic island.

The most intriguing organization is found in rhi, apparently a composite arrangement. It has a nodABC operon, a gap of 152 bp, following a nodIJ operon. nodD1 is ~200 kb downstream from nodAB [34]. This arrangement of key nod genes in rhi seems to be a composite of R etli, R. leguminosarum and Sinorhizobium nod gene arrangements.

The taxonomy of nod genes of the fourteen genomes was further explored by measuring the similarity of the NodD proteins within the designated MCL cluster. This was achieved through a neighbor joining (NJ) gene tree based on BLOSUM-62 matrix alignment of the proteins in that cluster (Figure 3). The cluster itself is a broad collection of LysR transcription regulator-like proteins, and it is divided onto two broad branches representing proteins annotated as NodD (colored according to cluster table color scheme) and those annotated only as LysR or putatively LysR (black nodes). This initial division shows that bra and bbt nodD1 are possibly not true NodD1 proteins but more general LysR, probably regulating the transcription of other gene systems, e.g., secretion, as both strains have no nodABC genes.

Further exploration of the NodD tree shows that the azc is the most distant in terms of amino acid domain composition with its closest neighbor being bja NodD2. The rest of the proteins in the cluster annotated as Nod largely cluster according to previously determined taxonomic genus relationship, except for rhi NodD1, which more closely aligns with bja NodD1, rather than with its fellow Sinorhizobium species. The largest cluster consists of the NodD1 and NodD2 of sme and smd with the NodD1 of rle, rlg and rlg and NodD2 of rlt. (Figure 3).

In summary, there is a significant difference within this protein cluster between nodulation specific LysR type proteins (NodD proteins) and the other LysR transcription regulation orthologs, indicative of their biochemical roles. Also, NodD orthologs show a slightly different taxonomic relationship compared to 16S rRNA and whole genome approaches.

The complexity of Nod factor synthesis and regulation is represented in the current study by two other important gene families, nol and noe. Among important proteins within this class we may cite NoeI, that catalyses the addition of -O-Me fucose to the Nod Factor backbone; NoeE, 3-O-SO3– fucose; NoeL; 3-or 4-O-acetyl fucose and NoeC, involved in the addition of arabinose to the Nod factor [48]. nol gene family is largely predominant in bja, but also abundant in both strains of R. etli and in S. fredii NGR 234 (Table S5), and thus possibly important for the synthesis of specific Nod factors in these species.

Many differences were found in noe clusters (Table S4). Again, the greatest number of genes in this category was found in bja. Sinorhizobium species have only noeA and noeB, and these genes were thought to be specific for Medicago nodulation [50]; however, a noeA ortholog is found in the Trifolium nodulator rlg as well as the non-symbiont mes. While the functionality of these specific orthologs is not known, this suggests that noeA may not be specific to Medicago.

Protein NolG is ubiquitous, found in cluster 23 and present in all strains from this study. However, it is only annotated as such in sme, with nolG-like genes annotated in smd and rec. Similar to the observations in relation to NodG, the NolG cluster contains many proteins, 130, all belonging to the very broad COG ortholog “COG0841, AcrB, cation/multidrug efflux pump”. Therefore it is possible that several orthologs represent only non-nodulation cation pumps.

After the initial signaling between the host plant and the rhizobia, the next stage controls the progression of the bacteria into the plant and their maturation into nitrogen fixing bacteroids, and many gene families are related to these steps; among the most crucial genes in these stages are the protein secretion systems. Bacterial secretion systems have been classified into seven main groups, Types I through VII, with the fimbrial chaperone-usher pathway forming an additional group. Four of these systems, Types III, IV, V, and VI, assemble surface structures that contact target cells and deliver DNA and/or protein effectors.

The secretion systems found in Rhizobiales range from Type I to Type VI, in addition to the twin-arginine transfer system (Tat). Noteworthy is that different nitrogen fixing Rhizobiales with differing host relationships contain different subsets of these systems [34,51,52,53,54]. Previously, it has been reported that only rhi had a suite of secretion systems spanning from Type I–V, which may strongly affect its extraordinary host range [34].

Using BlastlineMCL protein clustering and the KEGG pathway orthology, the number of Type I, II, II, IV, V and VI secretion genes, as well as of exopolysaccharide gene families exo and pss were investigated (Table S7–S14). Interesting, the non-symbiotic mes contains many of these secretion systems, including the virulence related Type IV genes. This could be construed as additional evidence that this strain, or a recent ancestor, was once a symbiont, but lost the symbiotic ability due to the availability of alternate energy sources.

Exopolysaccharide production has shown to be critical in host symbiont relationships, e.g., for the initial bacterial invasion that leads to the formation of indeterminate type of nodules on legumes [61]. The examination of exopolysaccharide synthesis and its role in legume infection has been mainly focused on the S. meliloti—Medicago truncatula model, with sme producing two exopolysaccharides, EPS I (succinoglycan) and EPS II (galactoglucan) [46]. In our comparison of genomes two exopolysaccharide synthesis families were surveyed, related to exo and pss genes; these were originally described in S. meliloti and R. leguminosarurm, respectively, and produce alternatively structured EPS I.

In a few cases, the ontology of exopolysaccharide synthesis proteins in nitrogen fixing Rhizobiales seems to be overlapping with four exo and pss terms used interchangeably [21]. For example, exoM is clustered with pssC in the R. etli species, both coding for a glycosyltransferase. This could be either an ontology clash, that the two proteins are in the same cluster due to similar amino acid structure, or that the same single protein is involved in the two systems.

Overall the clustering of exo (Table S15), and pss (Table S16) was similar to the phylogenetic clustering The African nodulator azc has only seven exo and three pss orthologs, compared to the twenty-three and six orthologs, respectively, found in sme. In addition, Bradyrhizobium has also few exo and pps genes, while Mesorhizobium has few pps genes. This interesting comparison could be indicative of a limited EPS I production in these microsymbionts, or that they use different strategies or sets of genes. It cannot be related to the tropical origin or the formation of determinate nodules, as other rhizobia, e.g., the tropical R. etli has several EPS genes and forms determinate nodules in common beans. Still considering the pps family, both ret and rle have twenty pss orthologs. Furthermore, EPS may also play other important roles, such as providing environmental protection, as pointed out in proteomic and transcriptomic studies with B. japonicum [62,63]. Continuing, the Trifolium nodulators have eighteen pss orthologs, missing pssH and pssI. Finally, rec has sixteen pps, missing ppsF, pssH, pssI, pssJ and pssK. In the other genomes the number of pps orthologs range from only three in azc to six in Sinorhizobium.

The role and importance of EPS in the symbiosis has been long discussed and studied. However, the comparison of the genomes has shown that the complexity of the genes regulating EPS is far from being understood. Different species have adopted different sets of genes that seem more related to evolution of the core genome than of the symbiotic genes. Therefore, it can be concluded that the production of exopolysaccharide is not necessarily a feature of a ‘universal’ symbiome.

The fixation of nitrogen via nitrogenase requires an anaerobic or micro-aerobic bacteroid environment and the fix gene family is involved in the regulation and metabolism of oxygen in this circumstance. The fix gene family is commonly found in three core operon structures: fixABCX, fixGHIS and fixNOPQ. The first operon is involved in the regulation of gene transcription under low oxygen concentrations. Metabolism of oxygen by the bacteroid occurs via the cytochrome cbb3 oxidase complex, a membrane complex encoded by the fixNOPQ operon, that mediates electron exchange and synthesis of ATP via oxidation on the outer surface of the bacteroid membrane. Studies in bja have shown that the second operon, fixGHIS, is required for the initial construction of the cbb3 complex, and like fixNOPQ, is only expressed under micro-aerobic or anaerobic conditions.

BlastlineMCL clustering of the Fix proteins shows both important species distinctions as well as a possible limitation to the algorithm (Table S17). All thirteen symbiotic species have the complete set of genes of the three operons; in the non-symbiont mes, fixABCX is entirely missing, as well as fixQ.

Each different species, except for azc, has a unique fixS ortholog cluster, and mlo has two. However, as the FixS protein is only fifty five amino acids in size, this could also be related to a limitation of the algorithm used by BlastineMCL, adjusted to larger proteins. Another stem nodulator, bbt, has no fixQ ortholog, signifying different cbb3 apparatuses from the structural models already resolved. This diversity of both FixQ and FixS proteins may suggest a diverse range of cbb3 complexes for each species that could result from differences in host nodulation strategies.

The nif family codes for the MoFe-dependant nitrogenase complex, the enzyme required for the catalysis for the nitrogen fixing process [64]. The thirteen symbionts contained, with a higher or lower similarity, the nifBHDKENX operon, which in turn was absent in the non-fixer mes (Table S18). Therefore, along with nod and fix, this species has also lost nif genes.

While a lack of nifX has been previous reported in rle [65], the current analysis has shown that both nifX and nifZ are also absent in R. leguminosarum rlt and rlg. This suggests the MoFe nitrogenase in these leguminosarum species may have a distinct structure from the other Rhizobiales. Also confirmed is the lack of nifQ, required for Mo-incorporation into the complex, in R. leguminosarum, S. meliloti and S. medicae, as well as in the non-symbiotic mes. Consequently, these species should use other means to incorporate Mo into the nitrogenise complex [65,66].

A Nif ortholog cluster was chosen for more in depth analysis of protein alignment (Figure 4). In this example, the NifN/K cluster was examined via neighbor joining (NJ) tree based on BLOSUM62 matrix. Both NifN and NifK (considered structural homologs in regards to their role in nitrogenase complex) split into two clear clades. Unlike much of the NodD cluster, the NifN clade is in agreement with the 16S rRNA phylogeny. Interestingly, the Bradyrhizobium and Azorhizobium NifN orthologs are genetically very distant from the other species, suggesting a different nitrogenase structure. A lower agreement with the 16S rRNA was obtained in the analysis of NifK; more specifically, NifK homologs of rhi show low similarity with other Sinorhizobium NifK homologs, showing higher similarity with mlo and with the R. leguminosarum orthologs. This is another suggestion of an important event of lateral transfer.