Multipartite genomes are common among members of the α-proteobacteria [1]. Most symbiotic nitrogen-fixing bacteria belonging to the generaRhizobium,Sinorhizobium, Mesorhizobium andBradyrhizobium possess multipartite genomes organized as a single circular chromosome and a variable number of large plasmids [2]. In some species plasmids can represent, in terms of size, up to 40% of the total genome. InRhizobium andSinorhizobium species one plasmid (pSym) concentrates most of the genes required for nodulation and nitrogen fixation [3]. The complete genome sequences of different rhizobia have revealed that plasmids harbor mainly accessory genes and that most encode predicted transport systems and a variety of catabolic pathways that may contribute to the adaptation of rhizobia to the heterogeneous soil and nodule environments [2,4]. These genes are absent from closely related genomes, lack synteny and their G+C composition differs from that of the core genes. The core genes are mainly located on chromosomes, have essential functions in cell maintenance and have orthologs in related species [5,6]. In spite of this evidently biased distribution of core genes in the chromosome and accessory genes in plasmids, it is important to highlight the fact that there are interesting exceptions to this genomic rule: several typical core genes have been found encoded on rhizobia plasmids. Some are copies of genes located on chromosomes, with redundant functions that are totally dispensable for normal growth. Examples of these genes are the multiple copies of chaperonin-encoding genesgroEL/groES [7,8], twotpiA genes encoding putative triose phosphate isomerase, a key enzyme of central carbon metabolism [4,6,9], and two putativeS. meliloti asparagine synthetases (asnB andasnO), which may have a role in asparagine synthesis from aspartate by ATP-dependent amidation [10]. In contrast to these reiterated genes, a few single copy core genes have also been localized in plasmids. ThetRNA specific for the second most frequently used arginine codon, CCG, is located on pSymB inS. melioti [10]. Since this gene lies within a region of pSymB that could not be deleted [11], it is assumed to be essential for cell viability. The single copy of theminCDE genes, conceivably involved in proper cell division, have also been found in plasmids ofS. meliloti, R. leguminosarum andR. etli [4,6,10]. Studies inS. meliloti have demonstrated that even though these genes are expressed in free-living cells and within nodules they are nonessential for cell division, since their deletion did not produce the small chromosomeless minicells observed inE. coli andBacillu subtilis [12].

A recent bioinformatic study revealed that approximately ten percent of the 897 complete bacterial genomes available in 2009 carry some core genes on extrachromosomal replicons [13]. However, very few of these genes have been functionally characterized and so their real contribution to bacterial metabolism is still an open question.

The complete genome sequence ofR. etli CFN42 predicts that two putative "housekeeping" genes,panC andpanB, which may be involved in pantothenate biosynthesis, are clustered together on plasmid p42f. Pantothenate is an essential precursor of coenzyme A (CoA), a key molecule in many metabolic reactions including the synthesis of phospholipids, synthesis and degradation of fatty acids, and the operation of the tricarboxylic acid cycle [14]. TheR. etli panC gene is predicted to encode the sole pantoate-β-alanine ligase (PBAL), also known as pantothenate synthetase (PS) (EC 6.3.2.1), present in theR. etli genome. The function of this enzyme is the ATP-dependent condensation of D-pantoate with β-alanine to form pantothenate, the last step of the pantothenate biosynthesis pathway. ThepanB gene encodes the putative 3-methyl-2-oxobutanoate hydroxymethyltransferase (MOHMT) (EC 2.1.2.11), also known as ketopantoate hydroxymethyltransferase (KPHMT), the first enzyme of the pathway, responsible for the formation of α-ketopantoate by the transfer of a methyl group from 5,10-methylentetrahydrofolate to alpha-ketoisovalerate. The complete genome sequence ofR. etli CFN42 predicts that a second putative MOHMT enzyme (RHE_PE00443), similar to the product ofpanB, is encoded on plasmid p42e.

In this work we describe the isolation and use ofpanC andpanB mutants to analyze the involvement of these plasmid-encoded genes in pantothenate biosynthesis. A survey of the localization ofpanCB genes among members of theRhizobiales with multipartite genomes allowed us to infer apanCB phylogeny and to establish the probable chromosomal origin of these plasmid-borne genes. We also report that thepanCB genes could not totally restore the growth in minimal medium (MM) of a strain cured of plasmid p42f, suggesting that other functions essential for growth in MM are encoded in this plasmid.

Functional characterization of plasmid p42f encodedpanCBgenes

The predicted function of the product ofpanC (RHE_PF00001) annotated as PBAL, is the catalysis of the last step of pantothenate synthesis. This PBAL (298 amino acids) showed 43% identity and 62% similarity over 279 amino acids with the functionally characterized PBAL ofE. coli K12 (284 amino acids). A search for conserved domains (CD-search) at NCBI-CDD revealed the presence of a typical pantoate-binding site. ThepanB gene (RHE_PF00002) is located immediately downstream ofpanC. The four nucleotide overlap between thepanC TGA codon andpanB ATG codon suggest that these genes might be transcribed as an operon. ThepanB gene encodes a putative MOHMT, the first enzyme of the pantothenate pathway. A BlastP comparison between the functionally characterized MOHMT ofE. coli K12 (264 amino acids) and the putative MOHMT encoded on plasmid p42f ofR. etli CFN42 (273 amino acids) showed 37% identity and 56% similarity over a length of 240 amino acids. A CD-search indicated that in the putative MOHMT ofR. etli CFN42 the magnesium binding and active site domains are conserved. Additionally, Paralog Search (KEGG SSDB) and pathway tools predicted a second probable MOHMT, encoded on plasmid p42e (locus tag RHE_PE00443). Both proteins are similar in length (273 and 270 aa for the products encoded bypanB and RHE_PE00443, respectively). However, a BlastP comparison of these sequences showed only 36% identity and 56% similarity over a tract of 140 amino acids. A CD-search revealed that only 5 of 12 of the invariable residues present in the active site domain are conserved in RHE_PE00443. The metal binding domain could not be detected by the CD-search. To determine whether thepanC andpanB genes located on plasmid p42f are required for pantothenate synthesis, mutations in these genes were generated by site-directed vector integration mutagenesis via a single cross-over recombination (see details in Material and Methods and Table1). Mutants ReTV1 (panC^-) and ReTV2 (panB^- ) were unable to grow in minimal medium (MM) lacking calcium pantothenate (Figure1a). Supplementation of MM with 1 μM calcium pantothenate allowed thepanC andpanB mutants to recover their wild-type growth rate (Figure1b). The pantothenate auxotrophy displayed by thepanB mutant ReTV2 allowed us to discard a functional role of the putative MOHMT encoded by RHE_PE00443 in pantothenate biosynthesis. Moreover, a pBBRMCS3 clone constitutively expressing RHE_PE00443 (pTV7) was unable to complement the pantothenate auxotrophy of thepanB mutant (data not shown).

Pantothenate auxotrophy ofR. etliCFN42panCandpanBmutants. Growth of theR. etli CFN42 wild-type strain and its derivativepanC (ReTV1) andpanB (ReTV2) mutants in: (a) minimal medium, (b) minimal medium supplemented with 1 μM calcium pantothenate. Values represent the means of three independent experiments; error bars show standard deviations.

Full size image

Plasmid pTV4, harboring thepanC andpanB genes, as well as plasmids pTV5 and pTV6, carrying onlypanC orpanB respectively, were introduced into mutant strains ReTV1 and ReTV2 and the growth phenotype of each construction was evaluated in MM. ThepanC mutant ReTV1 complemented with thepanCB genes (ReTV1-4) recovered wild type growth in MM. In contrast, when complemented only withpanC (strain ReTV1-5) no growth occurred in the absence of pantothenate. These results strongly suggest that thepanCB genes form a single transcriptional unit. As expected, wild type growth ofpanB mutant ReTV2 was recovered by complementation with thepanCB genes or with thepanB gene (strains ReTV2-4 and ReTV2-6 respectively).

The occurrence ofpanCB genes in plasmids is highly conserved amongR. etli andR. leguminosarum strains but not in other members of theRhizobialeswith multipartite genomes

To investigate whether the presence of thepanCB genes in plasmids is a common characteristic of theRhizobiales, we examined the location ofpanCB genes in 22 members of theRhizobiales having fully sequenced multipartite genomes (Table2). To date, the genomes of sevenR. etli strains, in addition to CFN42, have been totally sequenced [15]. However, with the exception of strain CIAT 652, the genomes were released as draft assemblies, precludingpanCB localization. We experimentally determined the localization ofpanCB genes in the genome of four of theseR. etli strains (CIAT 894, Kim5, 8C-3, and IE4771) by hybridization of their plasmid profiles with [³²P]dCTP-labelledpanC andpanB genes from CFN42 under high stringency conditions. Both probes produced intense hybridization signals on the same plasmid of each strain, indicating that thepanCB genes are also plasmid-borne in theseR. etli strains (Table2). Coincidentally, in the threeR. leguminosarum strains with fully sequenced genomes reported in the NCBI database, thepanCB genes are assigned to plasmids. In contrast, in other species ofRhizobiales with multipartite genomes, thepanCB genes are always confined to the chromosome, or to chromosome I in those species harboring two chromosomes, with exception ofAgrobacterium tumefaciens C58 which carriespanCB on the linear chromosome II andMethylobacterium nodulans ORS2060 that carriespanC on their single chromosome andpanB on plasmid pMNOD02 (Table2).

Phylogenetic analysis of rhizobialpanCBgenes indicates a common origin of chromosomal and plasmid-borne sequences

Two possible hypotheses were considered to explain the presence ofpanCB genes in plasmids ofR. etli andR. leguminosarum strains: (1) an intragenomic rearrangement ofpanCB genes from chromosome to plasmid, which must have occurred in the last common ancestor of both species; (2) by xenologous gene displacement, that is, a horizontal transfer event in which a gene is displaced by a horizontally transferred ortholog acquired from another lineage [16]. In the latter hypothesis we assume that the presence of these xenolog genes in plasmids conferred a selective advantage that may have eventually led to the loss of the chromosome-locatedpanCB genes. To test these hypotheses the phylogeny of 16 rhizobial species inferred from ten orthologous single copy housekeeping genes (fusA, guaA, ileS, infB, recA, rplB, rpoB, rpoC, secY and valS) located on primary chromosomes, was compared with the phylogeny of the same rhizobial species inferred from thepanCB genes located on plasmids and chromosomes. The rationale for this comparison was that if the plasmid-bornepanCB phylogeny agrees with the current phylogeny of theRhizobiales, inferred from the housekeeping genes, it would support the hypothesis of intragenomic transfer of thepanCB genes. On the other hand, if both phylogenies are incongruent, it would favor the hypothesis of horizontal transfer of thepanCB genes. Concatenated nucleic acids multiple alignments were used to infer both phylogenies with the maximum likelihood method described in materials and methods. The resulting phylogenetic trees are shown in Figure2. The housekeeping genes inferred tree (Figure2a) was consistent with the recently reported phylogeny of 19Rhizobiales performed on a data set of 507 homologous proteins from the primary chromosome [17]. Both trees are in close agreement with the phylogeny inferred from thepanCB genes (Figure2b). Thus the phylogeny ofR. etli andR. leguminosarum inferred from plasmid-encodedpanCB genes is consistent with the phylogeny deduced from their housekeeping genes supporting the hypothesis of a chromosomal origin for the plasmid-encodedpanCB genes.

Comparison of phylogenetic trees constructed from core andpanCBgenes. Maximum-likelihood phylogenetic trees of 16Rhizobiales constructed using the concatenated nucleic acid sequences of 10 housekeeping genes (a) orpanC andpanB concatenated genes (b). Bootstrap values are shown over each branch (based on 100 pseudo-replicates).

Full size image

ThepanCB genes do not fully complement the growth deficiency of aR. etliCFN42 p42f cured derivative in MM

It was reported previously thatR. etli CFNX186, a p42f-cured derivative ofR. etli CFN42, is unable to grow in MM [18]. To assess if the growth deficiency of strain CFNX186 in MM was due to the absence of thepanC andpanB genes, plasmid pTV4 (panCB) was introduced into strain CFNX186. The growth of the transconjugant (CFNX186-4) after 15 hours of culture in MM was only 50% that of the WT strain grown under the same conditions (Figure3a). The growth of CFNX186-4 did not improve even after 72 h in culture (data not shown). Interestingly, strain CFNX186-4 had the same growth rate as strain CFNX186 cultured in MM supplemented with 1 μM calcium pantothenate (Figure3b). This shows that the growth deficiency of CFNX186 is only partly due to the absence of thepanCB genes and indicates that other functions encoded in plasmid p42f are required for growth in MM.

panCBgenes do not fully restore the growth deficiency of CFNX186. Growth ofR. etli CFN42 wild-type strain, its p42f-cured derivative CFNX186, CFNX186 complemented with thepanCB genes (CFNX186-4) and CFNX186 complemented with a 20 kbEcoRI fragment of plasmid p42f containing thepanC,panB,oxyR andkatG genes (CFNX186-24) in: (a) minimal medium, (b) minimal medium supplemented with 1 μM pantothenate. Growth curves are the mean of at least three independent experiments; error bars represent standard deviations.

Full size image

Previous studies have demonstrated that thekatG gene, which encodes the sole catalase-peroxidase expressed in free-living growth conditions, is located on plasmid p42f ofR. etli CFN42. These studies also revealed that the growth rate of akatG mutant in MM was significantly reduced in comparison with that of the wild-type parental strain [19]. On plasmid p42fkatG, as well as its putative transcriptional regulator protein encoded byoxyR, are located 80 bp downstream of thepanCB genes. We speculated that introduction of thepanCB genes together with thekatG andoxyR genes might improve the growth of CFNX186 in MM. To test this hypothesis, we used pCos24, which contains a 20 kb fragment of p42f carryingpanCB,katG andoxyR (see Material and Methods). pCos24 was introduced into CFNX186 and the resulting transconjugant (CFNX186-24) grown in MM. Figure3 shows that after 15 hours of culture there was no significant difference between the growth rate of CFNX186 complemented only withpanCB (CFNX183-4), and CFNX186 complemented with cosmid pCos24 (CFNX186-24). Furthermore, the growth of CFNX186-24 did not increase even after 72 h of culture (data not shown) indicating thatkatG andoxyR did not improve the growth rate ofpanCB complemented CFNX186 in MM. We also tested the possibility that arginine might improve the growth of strain CFNX186-24 due to the presence of a putativeN-acetylornithinase (EC 3.5.1.16) encoded in the plasmid p42f. In theEnterobactericeae this enzyme catalyzes the conversion ofN-acetylornithine to ornithine, a key step in the arginine biosynthesis pathway [20]. However, the growth deficiency of strain CFN186-24 in MM was not corrected by the addition of 1, 5, 10 or 15 mM arginine (data not shown). Furthemore, we constructed anargE mutant strain (ReTV3, Table1) that was able to grow in MM without exogenous arginine at the same rate as parental strain CFN42 (data not shown), confirming that this gene is not essential for arginine synthesis.

Seminal studies on the phenotypic characterisation of plasmid-cured strains ofR. leguminosarum andR. etli revealed that the absence of several plasmids cause a growth deficiency in rich and minimal medium [18,21]. These findings suggested that undefined metabolic traits are present on rhizobial plasmids. The bioinformatic analysis of 897 bacterial genomes performed by Harrisonet al [13] revealed the presence of extrachromosomal core genes in 82 genomes mainly belonging to the Proteobacteria. In contrast with thesein silico data, there is little experimental information on the contribution of these core genes to bacterial metabolism or cellular process. The few genes that have been functionally characterized encode redundant functions and are totally dispensable for the cell [7–9,12]. Our study provides experimental evidence that the enzymes MOHMT (EC 2.1.2.11) and PBAL (EC 6.3.2.1) encoded on plasmid p42f are indispensable for the synthesis of pantothenate. Moreover, our results showed that the cluster ofpanCB, katG andoxyR genes was insufficient to restore full growth capacity to the p42f cured derivative CFNX186, implying that in addition to pantothenate synthesis, there are more functions encoded on plasmid p42f required for growth in MM. Obvious candidates for these functions could not be identifieda priori among the 567 proteins encoded in p42f even though their predicted functions were recently updated with KAAS (KEGG Automatic Annotation Server and Pathway Reconstruction Server). We discarded arginine limitation as the cause for the growth deficiency of strain CFNX186-24. The arginine prototrophy displayed by a mutation in the p42f encodedargE suggests that inR. etli the conversion ofN-acetylornithine to ornithine is catalyzed by the chromosome-encoded ArgJ, an ornithine acetyltransferase (OATase, EC 2.3.1.35), which transfers the acetyl group ofN-acetylornithine to glutamate to produce ornithine andN-acetylglutamate. Functional OATases have been found in the majority of bacteria [20].

Also, we have demonstrated that plasmid-localization ofpanCB inR. etli CFN42 is not unique to this strain. A screening of the location ofpanCB genes among members of theRhizobiales, showed that the occurrence of these genes in plasmids is a highly conserved trait amongR. etli andR. leguminosarum strains. Furthermore, the synteny of thepanCB,oxyR,katG genes inR. etli CFN42 is conserved inR. etli CIAT652 and inR. leguminosarum strains 3841, WSM1325 and WSM2304. In contrast, genomes ofRhizobium sp., Sinorhizobium, Bradyrhizobium andMesorhizobium species carried chromosomalpanCB genes. Only inA. tumefaciens C58 thepanCB genes are localized in the linear chromosome, whereas in all otherRhizobiales harboring secondary chromosomes thepanCB genes were located in chromosome I. A bioinformatic analysis with MicrobesOnline operon predictions [22] indicates thatpanCB genes are organized as possible operons in most of theRhizobiales examined in this work: all these predicted operons conserve the four nucleotide overlap between thepanC TGA codon and thepanB ATG codon observed inR. etli CFN42 (data not shown). In the genomes ofBradyrhizobium sp. BTAi1,Nitrobacter hamburgensis X14,Methylobacterium extorquens AM1,Methylobacterium radiotolerans JCM2831 andXantobacter autotrophicus Ry2,panC andpanB are encoded in separate chromosomal loci, whereas inMethylobacterium nodulans ORS2060panC is located in the chromosome andpanB in plasmid pMNOD02.

TheRhizobiales phylogeny inferred from concatenatedpanC andpanB genes was consistent with the phylogeny deduced from 10 concatenated housekeeping genes. The low bootstrap values obtained for some nodes of thepanCB phylogeny might be due to the small number of informative characters in the alignments of only two genes (1 977 nucleotides). This is consistent with previous reports that state that trees from longer alignments obtained by the concatenation of genes encoding multiple-protein families have higher bootstrap support than trees inferred from genes encoding single proteins [23]. The phylogenetic relationships amongRhizobium species carryingpanCB genes in plasmids with their closest relatives,Agrobacterium andSinorhizobium species, harboringpanCB genes in the chromosome was also observed in neighbor-joining trees inferred from singlepanC andpanB genes (data not shown). These data agree with the hypothesis that plasmid-encodedpanCB genes are orthologs of thepanCB genes located in chromosome. From these results, we propose that the presence of thepanCB genes in plasmids inR. etli andR. leguminosarum species may be due to an intragenomic transfer event from chromosome to plasmid. The mechanism leading to the transfer of core genes from chromosome to plasmids could involve cointegration and excision events between the replicons, similar to rearrangements that have been visualized inS. meliloti [24]. The translocation of genes from chromosome to plasmids may be part of the complex evolution of multipartite genomes. A study based on the analysis of clusters of syntenic genes shared among plasmids and secondary chromosomes of bacteria with multipartite genomes suggested that secondary chromosomes may have originated from an ancestral plasmid to which genes had been transferred from a primary chromosome [17].

Our pioneering work on plasmid-encoded functions inR. etli CFN42 established that a functional relationship among different replicons is required for symbiotic and free-living functions [18,25]. More recently, a functional connectivity among most of the proteins encoded in the replicons ofR. etli CFN42 was predictedin silico [6]. Our results demonstrated that the putative MOHMT encoded by RHE_PE00443 is not functional under the conditions studied and provides evidence of functional cooperation between p42f and chromosomally encoded proteins for pantothenate biosynthesis.

Our study shows that the presence of the corepanCB genes in a plasmid is a characteristic conserved inR. etli andR. leguminosarum strains but not in otherRhizobiales. The phylogenetic approach used in this study suggests that the unusual presence ofpanCB in plasmids may be due to an intragenomic transfer event from chromosome to plasmid rather than a xenologous gene displacement. UsingR. etli CFN42 as a model, we showed that the plasmid-encoded corepanCB genes were indispensable for the synthesis of pantothenate. ThepanCB genes could not totally restore growth of a strain cured of plasmid p42f in minimal medium, suggesting that other functions essential for growth in this medium are encoded in this plasmid. Our results support the hypothesis of functional cooperation among different replicons for basic cellular functions in multipartite rhizobial genomes.

Bacterial strains, media and growth conditions

The bacterial strains and plasmids used are listed in Table1.Rhizobium strains were grown at 30°C in three different media: a) PY rich medium [26], b) Minimal medium (MM) [27] and c) Minimal medium plus 1 μM calcium pantothenate (MMP). MM was prepared as follows: a solution containing 10 mM succinate as carbon source, 10 mM NH₄Cl as nitrogen source, 1.26 mM K₂HPO₄, 0.83 mM MgSO₄, was adjusted to pH 6.8 and sterilized. After sterilization the following components were added to the final concentration indicated: 0.0184 mM FeCl₃ 6H₂O (filter sterilized), 1.49 mM CaCl₂ 2H₂O (autoclaved separately), 10 μg ml^-1 biotin and 10 μg ml^-1 thiamine (both filter sterilized). MMP contains the same components plus 1 μM calcium pantothenate. To determine growth rates on MM or MMP,Rhizobium strains were grown to saturation in PY medium, the cells were harvested by centrifugation, washed twice with sterile deionized water and diluted to an initial optical density of 0.05 at 600 nm (OD₆₀₀) when added to 30 ml of MM. These cultures were grown for 24 h in 125 ml Erlenmeyer flasks to deplete any endogenous pantothenate. Cells were then harvested and washed as described above and added to fresh MM or MMP in the same manner as for the first inoculation and cultured for 15 hours. Bacterial growth was quantified by measuring optical density at 600 nm (OD₆₀₀) every 3 hours. Antibiotics were used at the following concentrations (in μg ml^-1): chloramphenicol (Cm), 30; tetracycline (Tc), 10; kanamycin (Km), 30; gentamicin (Gm), 30; spectinomycin (Sp), 100; nalidixic acid (Nal), 20.E. coli transformants harboring recombinant plasmids (β-galactosidase-positive) were identified by growth on LB plates with 30 μg ml^-1 5-bromo-4-chloro-3-indolyl-β-D-galactoside (X-Gal).

DNA manipulations

Standard techniques described by Sambrooket al. [28] were used for plasmid and total DNA isolation, restriction, cloning, transformations, and agarose gel electrophoresis. Plasmid mobilization fromE. coli toRhizobium was done by conjugation performed on PY plates at 30°C by using overnight cultures grown to stationary phase. Donors (E. coli strain S17-1) and recipients (R. etli CFN42 wild type and mutant strains) were mixed at a 1:2 ratio, and suitable markers were used for transconjugant selection.

Mutagenesis of thepanC andpanBgenes and genetic complementation of mutant strains

Mutants were generated by site-directed vector integration mutagenesis. Internal 400 bp DNA fragments ofpanC andpanB were amplified by PCR with primers A and B; C and D, respectively (Table3). PCR fragments ofpanC andpanB were cloned in vector pBC as 400 bpBamHI-XbaI fragments, generating pBC1 and pBC2 respectively, and then subcloned asKpnI-XbaI fragments into suicide vector pK18mob [29] to form plasmids pTV1 and pTV2, respectively. These plasmids were mobilized intoR. etli CFN42 by conjugation and single crossover recombinants selected on PY plates containing Km and Nal. The disruption of thepanC andpanB genes was confirmed by Southern blot analysis using a 400-bp PCR internal fragment of each gene as a probe (data not shown). The resultant mutants were named ReTV1 and ReTV2 respectively. To complement the phenotype of thepanC andpanB mutants, plasmids pTV4, pTV5, pTV6 and pTV7 were constructed as follows: a 3.1 kbEcoRI fragment from cosmid vector pCos24, isolated from a genomic library ofR. etli CFN42 [30] and containing thepanC andpanB genes, was subcloned in broad-host-range vector pRK7813, generating plasmid pTV4. To construct plasmid pTV5, a 1.2 kb fragment containing onlypanC (894 bp) was amplified by PCR with primers E and F and cloned in theKpnI-XbaI sites in the broad-host-range vector pBBRMSC3 so that the gene would be constitutively expressed from the vector'slacZ promoter. Primers G and H (Table3) were used to amplify a 1 kb PCR fragment containing only thepanB gene (822 bp). This DNA fragment was cloned in plasmid pBBRMSC3 in theKpnI-XbaI restriction sites, generating plasmid pTV6. Plasmid pTV7 contains the secondpanB gene (RHE_PE00443), encoded onR. etli plasmid p42e, this gene was amplified with primers I and J. The resultant 1 kb PCR fragment was cloned in theKpnI-XbaI sites of plasmid pBBRMSC3. To complement the growth deficiency of strain CFNX186, a derivative ofR. etli CFN42 cured of plasmid p42f, plasmid pTV4 and cosmid vector pCos24 were introduced by conjugation. The complemented strains obtained were named CFNX186-4 and CFNX186-24 respectively. TheargE gene was disrupted as described above. Briefly, an internal 400 bp PCR fragment ofargE amplified with primers K and L was cloned directly in pK18mob using theKpnI andXbaI sites to give pTV3 (Table1). This recombinant suicide plasmid was mobilized intoR. etli CFN42 and the resultant mutant named ReTV3 (Table1).

Filter blots hybridization and plasmid visualization

For Southern-type hybridizations, genomic DNA was digested with appropriate restriction enzymes, electrophoresed in 1% (w/v) agarose gels, blotted onto nylon membranes, and hybridized under stringent conditions, as previously reported by [31], using Rapid-hyb buffer. To use thepanC andpanB genes as probes, both genes were amplified by PCR, separated on a 1% agarose and purified by a PCR purification kit (QIAquick). They were labeled with [α-32P]dCTP using a Rediprime DNA labeling system. Plasmid profiles were visualized by the Eckhardt technique as modified by [21], and hybridized in a similar manner.

Identification of orthologous proteins, multiple sequence alignments and phylogenetic analysis

All genomic sequences analyzed in this study were obtained from the Integrated Microbial Genomes System of the DOE Joint Genome Institutehttp://img.jgi.doe.gov/). We obtained protein and gene sequences ofpanB, panC and 10 chromosomal housekeeping genes (fusA, guaA, ileS, infB, recA, rplB, rpoB, rpoC, secY and valS) from 16 rhizobial species. Accession numbers for these sequences and the species list are shown in Table S1 (see Additional file1). An orthologous data set for each gene was constructed using Blast [32] and the bidirectional best hit method applying the criteria reported by Poggioet al [33]. Multiple alignments of putative orthologous proteins were performed using the MUSCLE program [34] with default settings. After removing poorly conserved regions two concatenated protein alignments were obtained, one for the 10 chromosomal housekeeping genes (8469 amino acids) and the other forpanB andpanC (659 amino acids). Both concatenated protein multiple alignments were used to generate nucleic acids multiple alignments of their respective genes with the Tranalign program of the EMBOSS suithttp://emboss.sourceforge.net/. Nucleic acids multiple alignments were used to obtain two phylogenies with the maximum likelihood method implemented in PHYML [35] with HKY as substitution model [36]. The phylogenetic reconstruction was carried out with a nonparametric bootstrap analysis of 100 replicates for each alignment. TreeDyn program [37] was used to visualize and edit both phylogenies.