Multilocus sequence analysis of bradyrhizobia isolated from Aeschynomene species in Senegal

ARTICLE IN PRESS

0723-2020/$ - se

doi:10.1016/j.sy

�CorrespondMediterraneennMontpellier II,

lier cedex 5, Fr

E-mail addr

Please cite thi

Systematic and Applied Microbiology ] (]]]]) ]]]–]]]

www.elsevier.de/syapm

Multilocus sequence analysis of bradyrhizobia isolated from

Aeschynomene species in Senegal

A. Nzouea, L. Michea, A. Klonowskaa, G. Laguerreb, P. de Lajudiea, L. Moulina,�

aIRD, UMR 113, Symbioses Tropicales et Mediterraneennes, F-34398 Montpellier, FrancebINRA, USC 1242, Symbioses Tropicales et Mediterraneennes, F-34398 Montpellier, France

Abstract

This study reports the multilocus sequence analysis (MLSA) of nine house-keeping gene fragments (atpD, dnaK,glnA, glnB, gltA, gyrB, recA, rpoB and thrC) on a collection of 38 Bradyrhizobium isolated from Aeschynomene speciesin Senegal, which had previously been characterised by several phenotypic and genotypic techniques, allowing acomparative analysis of MLSA resolution power for species delineation in this genus. The nifH locus was also studiedto compare house-keeping and symbiotic gene phylogenies and obtain insights into the unusual symbiotic properties ofthese Aeschynomene symbionts. Phylogenetic analyses (maximum likelihood, Bayesian) of concatenated nine lociproduced a well-resolved phylogeny of the strain collection separating photosynthetic bradyrhizobial strains (PB) fromnon-photosynthetic bradyrhizobial (NPB) ones. The PB clade was interpreted as the remains an expanding ancientspecies that presently shows high diversification, giving rise to potential new species. B. denitrificans LMG8443 andBTAi1 strains formed a sub-clade that was identified as recently emerging new species. Congruence analyses(by Shimodaira–Hasegawa (S–H) tests) identified three gene-fragments (dnaK, glnB and recA) that should be preferredfor MLSA analyses in Bradyrhizobium genus. The nine loci or nifH phylogenies were not correlated with the unusualsymbiotic properties of PB (nod-dependent/nod-independent). Advantages and drawbacks of MLSA for speciesdelineation in Bradyrhizobium are discussed.r 2009 Elsevier GmbH. All rights reserved.

Keywords: Bradyrhizobium; Aeschynomene; MLSA; Rhizobia; Nodulation; Phylogeny; Nod-independent nodulation

Introduction

Rhizobia are a functional group of soil bacteriacharacterised by their ability to form nitrogen-fixingsymbioses with leguminous plants. Rhizobia are poly-phyletic and do not correspond to a taxonomic unit[34]. They are currently divided into alpha and beta

e front matter r 2009 Elsevier GmbH. All rights reserved.

apm.2009.06.002

ing author at: Laboratoire des Symbioses Tropicales et

es, UMR 113, IRD, CIRAD, SupAgro, Universite de

USC INRA, Campus de Baillarguet, 34398 Montpel-

ance.

ess: [email protected] (L. Moulin).

s article as: A. Nzoue, et al., Mulmacmillanxyx, Syst. Appl.

sub-classes of Proteobacteria and the names alpha andbeta rhizobia have been proposed to distinguish them[25]. The alpha rhizobia include more than 50 bacterialspecies representing 10 genera Sinorhizobium,Mesorhizobium, Rhizobium, Methylobacterium, Devosia,Azorhizobium, Bradyrhizobium, Ochrobactrum, Bosea

and Phyllobacterium (website of the ISCP sub-committeeon the taxonomy of Rhizobium and Agrobacterium; http://edzna.ccg.unam.mx/rhizobial-taxonomy/).

The Bradyrhizobium genus occurs worldwide and isassociated with economically important legumes such assoybean, cowpea, peanut or acacias. Species delineationin this genus has proven difficult because of very low 16S

Microbiol. (2009), doi:10.1016/j.syapm.2009.06.002

http://edzna.ccg.unam.mx/rhizobial-taxonomy/

http://edzna.ccg.unam.mx/rhizobial-taxonomy/

www.elsevier.de/syapm

dx.doi.org/10.1016/j.syapm.2009.06.002

mailto:[email protected]

dx.doi.org/10.1016/j.syapm.2009.06.002

ARTICLE IN PRESSA. Nzoue et al. / Systematic and Applied Microbiology ] (]]]]) ]]]–]]]2

rRNA sequence divergence among heterogeneous bra-dyrhizobial strains and limited consensus betweentraditional taxonomic methods [51,53]. To date, theBradyrhizobium genus has been shown to harbour sixrecognised species: B. japonicum [15,17], B. elkanii [20],B. liaoningense [54], B. yuanmingense [55], B. betae [29]and B. canariense [47]. Correlations between AFLP,DNA/DNA hybridisation and 16S–23S internaltranscribed spacer (ITS) sequence analysis were foundthat led to the description of 11 genomic species inBradyrhizobium [53]. However, phenotypic data werenot congruent between or within genomic groups and noadditional species names were proposed.

The Bradyrhizobium sp. (Aeschynomene) group con-tains photosynthetic bradyrhizobial (PB) and non-photosynthetic bradyrhizobial (NPB) strains with dis-tinct host-ranges on Aeschynomene spp. [24]. Alazard [1]defined three cross-inoculation groups (CI) in Aeschy-

nomene: Cross-inoculation group 1 is root-nodulated byNPB strains only; CI group 3 is root- and stem-nodulated by PB strains only, while CI group 2 isnodulated by both PB and NPB, but with the distinctionthat PB strains nodulate both stem and roots while NPBnodulate only roots of CI group 2. Aeschynomene

symbionts have recently attracted particular attentionsince some of the PB strains (ORS278 or BTAi1) whichnodulate roots and stems of plants of CI group 3(A. indica, A. sensitiva) lack the canonical nodulationgenes [10]. Other PB strains carry both Nod factor-dependent and Nod factor-independent systems tonodulate Aeschynomene CI groups 2 and 3 species(A. afraspera/Nod-dependent; A. sensitiva/Nod-independent) [10].

PB strains belong to genospecies VI and VIIIthat form a homogenous clade in 16S rRNA geneand ITS phylogenies [24,51]. However this groupexhibits very low sequence divergence of 16SrRNA genes but high diversity in ITS sequences (withmultiple copies) compared to other bradyrhizobia,making this usual marker of Bradyrhizobium diversityless useful for genetic screening in the PB group[51,52].

To overcome limitations of conventional molecularmethods in rhizobial taxonomy, several authors havesuggested integrating phylogenetic analysis of protein-encoding genes, with a higher level of sequencedivergence than rRNA genes, but sufficient conservationto retain phylogenetic signal and to allow primer design(references are listed below). The ad hoc committee forredefinition of bacterial species concept recommendedthe use of five genes [38]. In rhizobia, several house-keeping genes such as glnA and glnB [43], atpD and recA

[9,49], dnaK [40], recA and glnB [39,47] combined withatpD and rpoB [48] or larger studies using atpD, dnaK,gap, glnA, gltA, gyrB, pnp, recA, rpoB and thrC formultilocus sequence analysis (MLSA) in Sinorhizobium

Please cite this article as: A. Nzoue, et al., Mulmacmillanxyx, Syst. Appl.

[22] were used to study the evolutionary relationshipsamong strains and evaluate these data as an alternativeto DNA/DNA hybridisation methodology for bacterialclassification. Vinuesa et al. [47,48] were pioneers inapplication of multilocus sequence analysis usingphylogenetic methodologies to delineate species inBradyrhizobium. More recently, Rivas et al. [30] useda combination of five loci (atpD, recA, gyrB, rpoB,and dnaK) to resolve species delineation amonggenomic species of Bradyrhizobium by comparing withDNA/DNA hybridisation values. The authors foundthat hybridisation values were well reflected in the fivegene concatenate phylogeny. However, cut-off levels ofsequence similarities for species delineation could not beset for several markers (gyrB, rpoB and dnaK) as noclear-cut gap was found between intra- and interspecificsequences.

The aim of this study was to produce a well-revolvedphylogeny of photosynthetic bradyrhizobia and toevaluate the use of MLSA for species delineation inBradyrhizobium. Nine gene fragments (atpD, dnaK,glnA, glnB, gltA, gyrB, recA, rpoB and thrC) werechosen to perform MLSA on a collection of 38Bradyrhizobium spp. (Aeschynomene spp.) strains pre-viously characterized by ITS sequencing, DNA/DNAhybridisation and AFLP fingerprinting [52,53] allowinga comparison between MLSA and conventional taxo-nomic methodologies. In addition, one symbiotic gene(nifH) that reflects the bacterial symbiotic properties ofrhizobia [2,14,21,49] was studied for comparison withcore gene phylogenies and strain host range inAeschynomene–Bradyrhizobium symbiosis.

Materials and methods

Bacterial strains

All Bradyrhizobium strains are listed in Table 1.Bacteria were grown on yeast mannitol (YM) medium[46] at 37 1C, and conserved at �80 1C in thesame medium supplemented with glycerol (20% finalconcentration).

Molecular methods

Genomic DNA was extracted from 4-day culturesat 37 1C in 20ml of YM broth using standardprocedures [33].

PCR amplification was carried out as described in[26], and the primers used are listed in Table S2. Somelisted primers were redesigned for better PCR amplifica-tion and sequencing.

For each gene-fragment amplification, the follow-ing cycles were used: initial denaturation step


dx.doi.org/10.1016/j.syapm.2009.06.002

ARTICLE IN PRESS

Table 1. Bacterial strains used in this study.

Bradyrhizobium strainsa Host plant Specificity group Geno-speciesb AFLP group&

ORS277 A. sensitiva III VI 13

ORS278 A. sensitiva III Sep


ORS285 A. afraspera IIa 6

ORS287 A. afraspera IIa 32


ORS297 A. sensitiva III



ORS301 A. americana I 21

ORS303 A. afraspera IIa VI 2

ORS304 A. elaphroxylon I 17

ORS305 A. elaphroxylon I – Sep

ORS307 A. indica III – Sep

ORS309 A. uniflora I VII 34


ORS318 A. indica III VI 3

ORS324 A. afraspera IIa VI 6



ORS330 A. sensitiva III – Sep

ORS331 A. tambacoudensis III VI 8

ORS336 A. afraspera II VI 8


ORS354 A. afraspera IIb Sep

ORS356 A. afraspera IIa Sep

ORS358 A. nilotica IIb VII 34

ORS361 A. sensitiva III Sep




ORS377 A. elaphroxylon I 27


ORS390 A. indica III 11



ORS400 A. indica III Sep

BTAi1 A. indica III VIII Sep

ORS299 (Methylobacterium sp.) Not symbiotic Not symbiotic Sep

LMG8443T Blastobacter denitrificans A. indicac III VIII ND

aAll strains came from the ORS collection (ORSTOM-IRD), except BTAi1 [10] and Blastobacter denitrificans LMG8443 (LMG collection, Gand).bGenospecies as defined in [51,52] and AFLP groupings from [50].cNodulation tests from [45].

A. Nzoue et al. / Systematic and Applied Microbiology ] (]]]]) ]]]–]]] 3

(94 1C, 5min), 20 cycles of denaturation (94 1C, 30 s),annealing (TH1, see Table S2 with a decrease in0.5 1C for each cycle) and extension (72 1C, 1min 30 s),then 25 cycles of denaturation (94 1C, 30 s), annealing(TH2, see Table 2), extension (72 1C, 1min 30 s and finalextension (72 1C, 10min). PCR products were purifiedusing the QIAquick gel extraction kit (Qiagen) andcontracted for sequencing by the Macrogen ServiceCentre using Applied Biosystems ABI3730 machinesand ABI chemistry.


Sequence analysis

Sequences were corrected and assembled usingChromasPro v1.33 (Technelysium Pty Ltd.). Multiplenucleotide sequence alignments were generated usingMuscle3.6 [6] with default parameters and optimizedmanually using GeneDoc software [27]. A search fororthologs or paralogs of each marker was carried outusing the MAGE system [44] in the ‘‘Rhizoscopeproject’’ (http://www.genoscope.cns.fr/agc/mage).


http://www.genoscope.cns.fr/agc/mage

dx.doi.org/10.1016/j.syapm.2009.06.002

ARTICLE IN PRESS

Table 2. Multi-locus alignment statistics.

Gene

marker

Length

(bp)

Gap No seqa C sites

(All/Brady)

V sites

(All/Brady)

No PI

(All/Brady)

S

(All/Brady)

atpD 541 – 43 364/413 177/128 113/91 64/37

dnaK 659 6 43 443/499 216/154 125/115 87/39

glnA 382 – 43 274/309 108/73 62/53 46/20

glnB 600 – 43 389/436 211/164 139/130 72/34

gltA 496 43 303/366 193/130 125/106 68/24

gyrB 624 6 42 345/434 273/184 160/140 113/44

recA 522 – 43 324/374 198/148 128/114 70/34

rpoB 402 42 219/289 183/113 93/86 90/27

thrC 344 – 43 190/219 154/125 112/105 42/20

Nine loci 4570 12 43 2851/3339 1713/1219 1057/940 652/279

nifH 597 – 48b 333/400 258/191 218/172 40/19

Abbreviations: No: number, seq: sequences, Brady: statistics on Bradyrhizobium strains only, C: conserved sites, V: variable sites, PI: Parsimony

informative, S: singletons. Gap: number of gap columns in the alignment.aGene loci from five reference genomes were introduced (Bradyrhizobium sp. ORS278, BTAi1, B. japonicum USDA110, R. palustris CGA009, and

S. meliloti 1021), as well as B. elkanii USDA76T and B. japonicum USDA6T. rpoB from ORS279 was removed as it was related to Sphingopyxis

alaskensis (alpha-Proteobacteria). gyrB from ORS358 could not be amplified.bTwo copies of nifH are present in ORS278 and BTAi1, while no PCR amplification could be achieved in ORS368, ORS385 or ORS390. The nifH

copies of R. palustris CGA009, Azorhizobium caulinodans ORS571T, B. canarienseT, B. liaoningenseT, and B yuanmingenseT were included.

A. Nzoue et al. / Systematic and Applied Microbiology ] (]]]]) ]]]–]]]4

Recombination inference within alignment datasets wasestimated using the RDP2 software package [23].Distance score matrices of differences between strainsor groups of strains were generated with MEGA [42]using exact number of differences and the pairwisedeletion option to treat gaps or missing data. Averagenucleotide identity across groups of sequences wascalculated using DnaSP4 [32]. Sequences have beendeposited in the GenBank database under the followingaccession numbers: atpD FJ347201–FJ347236; dnaK

FJ347237–FJ347273; glnA FJ347274–FJ347309; glnB

FJ347310–FJ347346; gltA FJ347347–FJ347382; gyrB

FJ347383–FJ347417; recA FJ347452–FJ347487; rpoB

FJ347488–FJ347523; thrC FJ347165–FJ347200; nifH

FJ347418–FJ347451. Accessions for each strain areavailable in supplementary material Table S1.

Management of multiple copy genes and conflicting

phylogenetic signals

The BTAi1 genome harbours two copies of atpD

gene, one sharing 66% amino acid identity with itsclosest relative Burkholderia pseudomallei, the otherbeing a ‘‘classical’’ bradyrhizobial sequence (only thislatter one was included in our datasets). Some conflict-ing sequences were found among the set of ninesequences for each strain: the rpoB sequence fromORS279 was found related to Sphingopyxis alaskensis

and atpD from ORS297 was related to B. japonicum andnot to PB strains. No recombination between genes ofvarious datasets could be detected using software fromthe RDP2 package [23]. All detected conflicting


sequences were removed from alignments and treatedas missing data in concatenated datasets. StrainORS299, which grouped separately from other Bradyr-

hizobium spp. in ARDRA and AFLP experiments[24,51], showed sequences close to Methylobacterium

radiotolerans (98% nucleotide identity with dnaK, gltA,gyrB, rpoB and thrC sequences) and was considered as acontaminant and thus removed from the analyses.

The nifH gene is present in two copies in photosyn-thetic Bradyrhizobium ORS278 and BTAi1, while asingle nifH copy is present in B. japonicum USDA110.Sequencing of nifH fragment in several Bradyrhizobium

(Aeschynomene) sp. strains revealed at least two closelyrelated sequences in the chromatograms of the nifH

PCR products (2–19 over a 700 bp length), suggestingthe presence of two or more nifH copies with very closesequences. These differences were considered as missingdata in the subsequent phylogenies. An exception wasORS309 from which the nifH PCR product was clonedand copies were sequenced separately, and they werenamed ORS309-nifH1 and ORS309-nifH2.

Molecular phylogeny

Phylogenies of each gene marker were inferred bymaximum likelihood (ML) using the PAUP4 [41] andPHYML [12] software as well as by a Bayesian (BY)framework using MrBayes3.1.2 [31]. Best evolutionmodels for each marker were inferred by ModelTest3.6[28] using both hierarchical likelihood ratio tests (hLRT)and Akaike information criterion (AIC) as well as asite-specific substitution model following the codon


dx.doi.org/10.1016/j.syapm.2009.06.002

ARTICLE IN PRESSA. Nzoue et al. / Systematic and Applied Microbiology ] (]]]]) ]]]–]]] 5

structure of protein-coding DNA. The codon position isshown to be an important parameter omitted inModelTest but that best fits in many cases withprotein-coding sequences [35]. These different modelswere used as input parameters for ML or as priors forBY phylogenies. Mixed models incorporating bothpartitions ‘‘by gene’’ and ‘‘by codon position’’ werealso used in MrBayes searches on multiple genepartitions. Bayesian search parameters: 2 runs, n

chains ¼ 4, heat parameter fixed to 0.2, n ¼ 2. 10e6

generations, sample freq ¼ 200; outgroup ¼ S. meliloti

1021. A 50%majority rule consensus tree was built fromthe T file produced by MrBayes using the ‘sumt’ option,and a �ln L score was assessed from this consensus treewith PAUP4. Posterior probabilities from BY phylo-genies were compiled from the two independent runs offour chains, sampling trees after the chains had reachedapparent stationary state (by comparing marginal �ln Lagainst generations). To evaluate node robustness inML trees, 100 bootstrapping replicates were performedusing PHYML on-line [13].

Incongruence length difference tests (ILD) [7] im-plemented in PAUP4 were used to determine if genemarkers had congruent signals under parsimony criter-ion. Tree topology congruence was assessed using theShimodaira–Hasegawa likelihood-based test (S–H test[36]) implemented in PAUP4 using 1000 bootstrapreplicates (by the RELL method).

Results

Characteristics of each locus and individual

gene phylogenies

In order to build a clear-cut phylogeny of ourBradyrhizobium strain collection, nine house-keepinggene fragments of 38 strains previously characterised bypolyphasic taxonomy were sequenced and compared.The nine loci are presented in Table 2 together with thestatistics of each alignment generated. The size offragments and number of variable sites for eachbradyrhizobial locus varied from 344 to 659 pb andfrom 154 to 273 bp, respectively. Some conflictingsignals were found in the sequences generated, and themanagement of misleading sequences is explained inMaterials and Methods.

The phylogeny of each gene fragment was assessed bymaximum likelihood and Bayesian inferences. The bestfitting models selected by ModelTest using both hLRTand Akaike criterion are indicated in Table S4. Anucleotide substitution model (site-specific, SS), whichincorporated the codon position of each nucleotide, wasalso used as suggested by Shapiro et al. [35]. Substitu-tion rates estimated (after 100 rearrangements with


PAUP4) for each position in the codon are indicated inTable S3. Table S4 indicates the likelihood (�ln L)scores of the best and consensus trees found by ML andMrBayes, respectively using the various parametersdescribed above.

Shimodaira–Hasegawa tests of congruence betweenML trees and Bayesian trees (S–H test using RELLbootstrap, 1000 replicates) were performed for eachMLSA gene locus individually, and congruence wasfound between trees (of the same locus), whatever theinference method or parameters used (Table S4).Individual gene phylogenies built with best performingmodels are available as supplementary material(Fig. S1). As shown in Fig. S1, some discrepancies inthe individual gene phylogenies were found, especiallystrain placement within the three main branches(PB-Photosynthetic Bradyrhizobium, BJ-B. japonicum

and BE-B. elkanii lineages).

Building a robust phylogeny of photosynthetic

Bradyrhizobium by MLSA of house-keeping genes

A major advantage of Bayesian phylogenetic infer-ences is the possibility to apply mixed evolution modelsto a data partition [16,31]. A 4570 bp long (includinggaps) partition of nine house-keeping gene fragmentswas built. Bayesian inferences were conducted usingpartitions of the data by gene (using best-performingML models shown in Table S4), codon position (sitespecific) or by both. Likelihood scores (�LnL) for eachBayesian inference are indicated in Table 3 together withS–H congruence test P values (assessing the topologycongruence between trees). Fig. 1 shows the Bayesian50% consensus-derived tree using mixed models foreach partition. Such tree topology was also found to becongruent with the best ML tree found by PAUP using aGTR+I+G model (S–H test P value ¼ 0.509).

Fig. 1 is thus considered as the most representative ofBradyrhizobium species tree. In this tree, all photosyn-thetic strains grouped in a single clade highly supportedby both bootstrap values (from PHYML) and posteriorprobabilities from the Bayesian analysis. An exceptionwas strain ORS377 that, despite its non-photosyntheticstatus, grouped within the photosynthetic clade. TheBayesian and ML trees supported the monophyly ofBradyrhizobium genus and the following evolutionaryscenario: photosynthetic bradyrhizobia representing thefirst speciation event in Bradyrhizobium genus, followedby the appearance of two main lineages (B. elkanii andB. japonicum ancestors) with subsequent diversificationin each branch. The Rhodopseudomonas palustris straingrouped outside Bradyrhizobium genus, confirmingprevious house-keeping gene phylogenies based on fewermarkers [49] and in contrast to previous 16S rRNAgene (or single gene-based) phylogenies that placed


dx.doi.org/10.1016/j.syapm.2009.06.002

ARTICLE IN PRESS

Table 3. Likelihood scores and S–H tests across partitions of congruent datasets.

Gene partition Evolution model �ln L Diff �ln L S–H test P value

Nine loci ML GTR+I+G 31230.11088 (best)

ML GTR+I+SS 31247.61931 17.50843 0.932

BY by_gene 31249.34042 19.22954 0.909

BY by_codon 31247.61931 17.50843 0.932

BY gene+codon 31258.83288 28.72200 0.861

dnaKglnBrecA ML GTR+I+G 31531.66601 301.55513 0.071

ML GTR+I+SS 31517.72945 287.61857 0.078

BY gene+codon 31546.01793 315.90705 0.054

atpDgltAgyrB ML GTR+I+G 31783.62639 553.51551 0.002�

ML GTR+I+SS 31708.23592 478.12504 0.004�

BY gene+codon 31863.95617 633.84529 0.000�

rpoBthrC ML GTR+I+G 33267.15379 2037.04291 0.000�

ML GTR+I+SS 33252.19786 2022.08699 0.000�

BY gene+codon 32857.84350 1627.73262 0.000�

�Po0.05. ML/BY: Maximum likelihood or Bayesian inference. SH test using RELL bootstrap (one-tailed test); Number of bootstrap

replicates ¼ 1000. P value indicates the significance of the difference in �ln L scores (Diff �ln L) between each locus partition-derived tree. Input

dataset in PAUP for all S–H tests: nine loci concatenated alignment with the GTR+I+G ML model. The glnB tree is found with the best �LnL

score; dnaK and recA loci exhibit congruent tree topologies with that of glnB.

A. Nzoue et al. / Systematic and Applied Microbiology ] (]]]]) ]]]–]]]6

Rhodopseudomonas among Bradyrhizobium species [40].The PB strains were distributed in several sub-branches,some being strongly supported by bootstraps andposterior probabilities, but no clear statistically sup-ported structuring was found.

Comparison of MLSA with previous genotypic

fingerprinting methods and bacterial host range

In order to evaluate the contribution of MLSA tobacterial taxonomy, genospecies defined in [53] andAFLP groupings from [50] were plotted on the BY treein Fig. 1. MLSA using nine loci appeared highlyinformative for resolving the species tree of theBradyrhizobium genus, and was highly congruent withthe AFLP data (except for ORS377, ORS310 andORS287) and could resolve strain placement unsolvedby AFLP studies (as ORS310, ORS361 and ORS356).The host group of each Bradyrhizobium strain, asdefined by Molouba et al. [24], was also plotted inorder to detect correspondence between the host rangeand the taxonomic classification but no correlationcould be found. It is noticeable that PB strains from hostgroup 3 (nodulating CI group 3, such as A. indica and A.

Fig. 1. Bayesian phylogeny of nine concatenated gene fragments (atp

resolved with significant posterior probabilities (40.90) are indicate

same node (only bootstrap 450%; *if o50%). Strain name is fol

photosynthetic strains, cross inoculation (CI) group 1; HG2a: pho

(A. afraspera) and group 3 (A. indica) of Aeschynomene; HG2b: only

and root nodulation only on group 3 (A. indica). The scale bar in

Bradyrhizobium, BJ: B. japonicum lineage, BE: B. elkanii lineage, AN

clade. Accession numbers are available in Table S1.


sensitiva via a common Nod gene-independent pathway)[10] are intertwined with PB strains from host group 2(nodulating a wider number of Aeschynomene and usingboth Nod- and Nod-independent infection processes) inthe BY tree, indicative of lateral transfers of symbioticdeterminants between the photosynthetic bradyrhizobia.

Phylogenetic congruence evaluation between single

and combined gene partitions

In order to infer phylogenetic congruence betweeneach house-keeping gene locus, incongruence lengthdifference (ILD, [7]) and Shimodaira–Hasegawa treetopology test of congruence [36] were performed todetermine if the datasets had congruent phylogeneticsignals (ILD test) or if ML trees had significantlydifferent scores (by S–Htests). ILD tests performedunder parsimony criterion showed no congruencebetween the various datasets (all P values were o0.05;not shown). Following Dowton and Austin [4], suchresults may be explained by the fact that ILD tests areunlikely to be an effective measure of congruence whentwo datasets differ markedly in size (our datasets variedfrom 344 to 659 bp in alignments).

D, dnaK, glnA, glnB, gltA, gyrB, recA, rpoB and thrC). Clades

d followed by the bootstrap value obtained by PHYML on the

lowed by its host group (HG) as defined by [24]: HG1: non-

tosynthetic, stem and root nodulation on both CI groups 2

root nodulation on CI group 2, not photosynthetic; HG3: stem

dicates number of substitutions per site. PB: photosynthetic

I: the average nucleotide identity among all loci within the PB


dx.doi.org/10.1016/j.syapm.2009.06.002


The results of various S–H test P values aresummarized in Table S5. Using the concatenation ofnine house-keeping gene loci as the input in PAUP withthe best-fit model defined by ModelTest (GTR+I+G),

ORS301 (A. americana, HG1)

Rhodopseudomonas palustris CGA009

ORS279 (A. sensitiva, HG3)


ORS400 (A. indica, HG3)



ORS377 (A. elaph


ORS393 (A. indica, HG3




ORS303 (A. afraspera, HG2a)





ORS331 (A. tambacoudensis, HG3)


ORS324 (A. afraspera, HG2a

ORS385 (A. indica , HG3)

ORS300 (A. sensitiva , HG3)


ORS287 (A. afraspera, HG2a


ORS327 (A. indica, HG3)A. indica, H





BTAI1 (A. indica, HG3)

ORS297 (A. sensitiva , HG3)

ORS358 (A. nilotica, HG2b)

ORS309 (A. uniflora, HG1)

ORS304 (A. elaphroxylon, HG1)

ORS336 (A. afraspera, HG2b)

B. elkanii USDA76T (soybea

ORS354 (A. afraspera, HG2b)

ORS305 (A. schimperi, HG1)

B. japonicum USDA110 (soybean)

B. japonicum USDA6T (soybean)

0.05

1.00; 96

1.00; 93

1.00; 77

1.00; 62

1.00; 77

1.00; 82

1.00; 100

1.00; 98

1.00;92

0.61;26

1.00; 99

1.00; 100

1.00; 53

1.00; 97

1.00; 1000.94; 52

1.00; 98

1.00; 100

1.00; 93

1.00; 88

1.00; *

1.00; *

0.99;*

1.00; *

0.99;*

1.00; 100

0.99; 31

0.77;*

0.99; 84

1.00 ;100

0.99; *

1; 100

1; 100

0.82; 820.81; *

0.77; *

0.99; *

1.00 ; 82


the best �ln L tree score of all markers was that of glnB

(�ln L ¼ 32021, Table S5A), and S–H tests showed thatdnaK- and recA-derived trees were congruent with glnB

tree. Congruence between each locus (one by one) was

Sep

Sep

Sep

SepSep

27 (Non-photosynthetic)

11

Sep

Sep

Sep

Sep

17

Genospecies

6

32

Sep

Sep

21Non-Photosynthetic(NPB)

PhotosyntheticBradyrhizobium(PB)ANI:96.1%

S. meliloti 1021

roxylon, HG1)

)

))

)

G3)

n)

VI 13

VI 13

VI 11

VI 3

VI 3

VI 3

VI 2

VI 9

VI 9VI 8

VI 6

VI 6

VI 7

VI 6

VI 10

VI 10

VI 5

VI 8II 32

I 15

I 12

VIII 28

VII 34

VII 34

AFLP


dx.doi.org/10.1016/j.syapm.2009.06.002


also tested to detect the congruent signals among otherloci that would indicate alternative evolutionaryschemes other than that of glnB, dnaK or recA.Phylogenetic congruence was found between (dnaK,glnB and recA), as expected, but also between (atpD,gltA and gyrB) and (rpoB and thrC) (Table S5B).

As three distinct phylogenetic signals were detected byS–H tests, the congruence was tested between partitionsof dnaK–glnB–recA, atpD–gltA–gyrB and rpoB–thrC

and the partition of all genes. The results are presentedin Table 3. As shown, only the dnaK–glnB–recA

partition-derived tree was congruent with that of thenine loci partition (P value 0.052). A comparison of thebest ML trees found for each partition is presented insupplementary material Fig. S2. Choice of gene markeris undoubtedly crucial for resolving relationshipsbetween the strains, and they must be carefully chosenbefore combination. Here, dnaK, glnB and recA seemedthe most appropriate of the nine markers used to resolvean accurate and congruent phylogeny in Bradyrhizo-

bium. Incongruence between loci could be explained bydiscrepancies found in individual gene phylogenies(see Fig. S1).

Phylogeny of Bradyrhizobium based on congruent

(dnaK–glnB–recA) loci

As congruence tests showed that three markers mimicthe overall phylogeny of the combined nine markers,dnaK, glnII and recA genes were used for all Bradyrhi-

zobium spp. type strains to assess the phylogeneticplacement of Aeschynomene symbionts within theBradyrhizobium genus. All Bradyrhizobium spp. typestrains for which three MLSA markers were available inGenBank were included, and three sequences fromBlastobacter denitrificans LMG8443T were added toassess the suggestion of van Berkum et al. [45] totransfer Blastobacter denitrificans to Bradyrhizobium.An ML inference was applied using a GTR+I+Gmodel (Table S4) and the ML tree obtained is shown inFig. 2. We also plotted on Fig. 2 the percentage ofaverage nucleotide identity (ANI) calculated on all threeloci inside and between the highly supported clades,which was also plotted on Fig. 2, using 95% ANI as aspecies cut-off value as suggested by Konstantinidis etal. [18] and Goris et al. [11] for conserved genes (valueestimated from bacterial whole genome analyses oforthologs). Nucleotide identities varied from 94% to99.4% (average 96.3%) within the photosyntheticBradyrhizobium lineage. If BTAi1 was excluded fromthe PB cluster (BTAi1 belongs to genospecies VIII whileother PB strains belong to genospecies VI), then theidentity values varied from 95.1% to 99.4%. Nucleotideidentity between BTAi1 and Blastobacter denitrificans

reached 99.5% on all three loci corroborating the results


of Willems et al. [51] that placed both strains in the samegenospecies VIII. Average nucleotide identity was lowerthan 95% between PB, BJ and BE (91.8% between PBand BJ, 91.5% between PB and BE and 92.1% betweenBJ and BE). Two sub-branches within the BE clade(ORS304, ORS358 and ORS309) and (ORS336,ORS301, and B. elkanii USDA76) shared 92.1% ANIand could represent two species. On the other hand,within the BJ lineage more than 95% ANI was foundbetween types strains of B. japonicum and B. canariense,and between B. liaoningense and B. yuanmingense

species or ORS305 (Fig. 2).

nifH gene phylogeny

A nifH phylogeny was produced for almost all strains(some strains could not be amplified; see Materials andMethods, and Table S1) and it was compared to thespecies tree. The best ML model found by ModelTestfor the nifH dataset was GTR+I+G (Table S4), andthe ML-derived tree is shown in Fig. S3 includingbootstrap values obtained by PHYML (100 replicates).

All bradyrhizobial strains included in the nifH

phylogeny formed a highly supported monophyleticgroup. The sequences from Aeschynomene bradyrhizo-bia were distributed into three main well-resolvedclusters. One cluster included most PB strains plus theA. caulinodans type strain as previously found [49].Interestingly, A. caulinodans strain ORS571 shared anunique property with PB strains to induce stem nodules(on Sesbania rostrata). Almost all NPB strains con-stituted a second cluster that also encompassed soybeanbradyrhizobia belonging to B. japonicum, B. liaoningense

and B. elkanii, and the type strain of B. yuanmingense

isolated from a Lespedeza cuneata plant [55]. ThreeAeschynomene bradyrhizobial strains constituted a thirdlineage including two PB strains and NPB strainORS377 that was placed in the PB lineage on thespecies tree. Sequence divergence within the PB nifH

cluster strains was low (ANI at 98.9%) compared tooverall average nucleotide identity between all bradyr-hizobial strains included (ANI ¼ 92.1), and betweenNPB Aeschynomene bradyrhizobial strains(ANI ¼ 88.3%). Resolution in the PB sub-brancheswas not achieved, probably due to short branch lengthand low divergence between PB strains. Although PBstrains mainly clustered together, there were severalinconsistencies with the species tree: the group of threestrains belonging to the PB lineage, which formed aseparate group in the nifH tree; the PB strain ORS393that grouped with bradyrhizobial strains belonging toother Bradyrhizobium species and ORS301, a BE strainin the species tree grouping with PB strains in the nifH

tree. Additionally, PB strains belonging to different sub-groups in the species tree and different cross inoculation


dx.doi.org/10.1016/j.syapm.2009.06.002

ARTICLE IN PRESS

S. meliloti 1021 Rhodopseudomonas palustris CGA009

ORS298 ORS294

ORS279 ORS277

ORS400 ORS377 ORS390

ORS391 ORS393

ORS318 ORS310

ORS372 ORS278 ORS285 ORS287 ORS300

ORS328 ORS385

ORS324 ORS368

ORS297 ORS356

ORS361 ORS344

ORS330 ORS375 ORS327 ORS307

ORS303 ORS331

BTAi1 Bl. denitrificans LMG8443T

ORS358 ORS309

ORS304 ORS336

ORS301 B. elkanii USDA76T

B. betae LMG21987T

B. japonicum USDA110 B. japonicum USDA 6T

B. canariense BTA-1T

ORS354B. liaoningense LMG18230T

ORS305B. yuanmingense CCBAU10071T

100100

10099

99

100

100

100

100

91

100

10087

71

100

98

82

83

10054

100

7881

70

68

56

96

100

100

100

10096

100

100100

10051

51

100

100

100

0.02

PBANI:96.3%*

BEANI%:95.3

BJANI%:95.4ANI

96.6% ANI 96.3%

ANI95.8%

ANI99. 5%

ANI 96.6%

ANI 97.9% ANI

92.1%

NPB

Fig. 2. Maximum likelihood phylogeny of dnaK–glnB–recA partition (GTR+I+G). Bootstrap percentages from 100 replicates are

shown at the tree nodes (only if 450%). Average nucleotide identities (ANI) are shown inside and between strains or groups of

strains. The scale bar indicates number of substitutions per site. Abbreviations are the same for Fig. 1. Accession numbers are

indicated in the supplementary material Table S1.

A. Nzoue et al. / Systematic and Applied Microbiology ] (]]]]) ]]]–]]] 9

Please cite this article as: A. Nzoue, et al., Mulmacmillanxyx, Syst. Appl. Microbiol. (2009), doi:10.1016/j.syapm.2009.06.002

dx.doi.org/10.1016/j.syapm.2009.06.002


groups showed identical nifH sequences (e.g. ORS303and ORS298).

Discussion

In this study, we evaluated nine house-keeping genefragments to resolve phylogenetic relationships in acollection of Bradyrhizobium isolated from Aeschyno-

mene species. We chose to work on a collection of strainsthat had been previously studied by AFLP, ITSphylogeny and partially in DNA/DNA hybridisationsstudies [24,53,54] for comparison of the levels oftaxonomic resolution of these methodologies.

Advantages and drawbacks of MLSA for

species delineation

It was found that the meta-alignment derived from allloci produced a well-resolved phylogeny of the bacterialstrains, but phylogenetic signals between loci were notcongruent by ILD tests, and only three of the markers(dnaK, glnB and recA) gave best �ln L scores andcongruent tree topologies in S–H tests when comparedto the nine loci tree. Alternative congruent treetopologies (by S–H tests) were found between atpD–gl-

tA–gyrB and rpoB–thrC, but trees derived from thesepartitions were not congruent with the nine loci tree. Itis important to note that despite being incongruent,those latter tree topologies were very close to thednaK–glnB–recA and the ‘‘species’’ trees, and differencesbetween the three main lineages (PB, BJ and BE) couldbe attributed to a few strains that were placed differentlyaccording to individual gene phylogenies (the case ofORS297, ORS391, ORS377 and ORS328; see Fig. S1)probably due to recombination events. Discordancebetween loci can also be accredited to variationsobserved in the sub-branches of the PB branch due toshort internal branch length that caused poor statisticalsupport. Since the choice of congruent markers is crucialin obtaining tree topologies as close as possible to thespecies tree, it can be concluded from the analysis thatdnaK, glnB and recA are the best performing ones toachieve this aim as they exhibit the most congruentsignals with few recombination events detected in ourdatasets. However, we cannot exclude that a differentset of genes would have come out best if all genomicspecies of Bradyrhizobium or additional gene loci hadbeen included. Nevertheless, our results are in accor-dance with Vinuesa et al. [48] who found congruencebetween glnB and recA phylogenies by studying a largecollection of B. canariense and B. japonicum strains.Rivas et al. [30] studied the phylogenies of atpD, recA,gyrB, rpoB and dnaK on three representative strains ofeach genomic species in Bradyrhizobium and found the


five concatenated sequence phylogeny in agreement withDNA–DNA hybridisation data. We reached the sameconclusions with our 9-gene (and 3-gene) phylogeny onthe PB and the NPB clades using more representativestrains but in a restricted number of genomic groups.

As a conclusion of an MLSA study on Ensifer species,Martens et al. [22] reported that among ten gene loci,recA, gyrB, gltA and thrC were good phylogeneticmarkers of rhizobia (glnB not being included in theirstudy). In our study, Bradyrhizobium ML trees builtfrom gyrB, gltA and thrC sequence data gave closetopologies to the most probable species tree (i.e. the nineloci-derived tree), but they were not statisticallycongruent due to different bipartitions in the trees.These results thus suggest that best gene-markerselection for bacterial classification may differ accordingto the bacterial genus under study.

Recombination event detection is critical in phylo-geny in order to avoid misleading taxonomic character-ization, and such events have proven to be widespreadfor the 16S rRNA gene [5]. Here, the use of combinationof several loci minimizes the impact of recombination orthe presence of anomalous gene trees among multiplemarkers [3], and Figs. 1 and 2 represent the bestavailable pictures of evolutionary relationships for thephotosynthetic Bradyrhizobium species.

Phylogeny of photosynthetic Bradyrhizobium:

insights into species delineation

Our study shows the whole PB clade clearly separatedfrom other Bradyrhizobium species with a phylogeneticdistance long enough to even consider this clade as aseparate genus. This idea has already been proposed bySo et al. [37] and Fleischman and Kramer [8] based onthe phenotypic data, but the PB group was stillconsidered as a sub-generic group of Bradyrhizobium

following AFLP and DNA/DNA hybridisation studies[51,53]. The proposal of a single species for the whole PBclade would not match with the 70% DNA/DNAhybridisation cut-off since Willems et al. [52] reportedvariations inside this group from 43% to 100%hybridisation values. The PB clade thus appears torepresent the remains of an expanding ancient species.Most of the PB clade sub-groups have not evolvedenough to represent clearly distinguishable species.Exceptions are strains BTAi1 and LMG8443(B. denitrificans) that form a sub-clade that has evolvedquite importantly from the other sub-groups (theyexhibit less than 95% ANI with other PB strains, anddiverged enough to represent a separate genospecies –VIII – as previously reported [52]) and could represent amodern species derived from the ancient PB species. Ourresults are in accordance with the proposal of vanBerkum et al. [45] to transfer Blastobacter denitrificans


dx.doi.org/10.1016/j.syapm.2009.06.002


to Bradyrhizobium denitrificans species, but we suggestincluding only BTAi1 (and other genospecies VIIIstrains) within B. denitrificans.

The nifH phylogeny produced in this work mirroredthe nine loci phylogeny for the PB clade with somediscrepancies symptomatic of the lateral transfer be-tween Bradyrhizobium lineages. Both nifH and thespecies trees reflected the division between NPB strains(host group 1 and 2b, nodulating roots of Aeschynomene

species of cross inoculation groups 1 and 2) and the PBstrains of host group 2a and 3 (nodulating roots andstems of CI group 2 and 3), these latter groups hostingthe Nod-independent pathways contrary to NPBstrains. However, neither nifH nor the MLSA treesdiscriminated PB strains on their host range or Nod-dependent/independent nodulation abilities. It seemslikely that the Nod-independent pathway is a mono-phyletic character in the PB clade, allowing strains tonodulate Aeschynomene from cross inoculation group 3,and that laterally transferred nodulation genes [26]appeared in some PB strains (such as ORS285 [10])conferring the ability to nodulate also the crossinoculation group 2 Aeschynomene species.

As observed in previous studies on several bacterialgroups [18], there is a high variability in the phylogeneticperformance of single genes, and accurate phylogenies atthe inter- and intraspecies level can be built using only afew of the best performing genes. Goris et al. [11]correlated the ANI value of 95% of conserved genes tothe 70% DNA/DNA hybridisation cut-off value forspecies boundary in bacteria. On the basis of 95% ANIretained as a cut-off value for species delineation in theBradyrhizobium genus, then all photosynthetic Bradyr-

hizobium strains studied would belong to a single speciesincluding Blastobacter denitrificans. This result corro-borates previous studies based on the analysis of16S–23S ITS and rRNA gene sequences [45,53]. TheB. elkanii species delineation fits well with the 95% ANIspecies cut-off, as its type strain shares 90.1% and 92%ANI with B. japonicum and the photosynthetic Bradyr-

hizobium, respectively. However, the 95% ANI speciescut-off for three concatenated gene phylogeny does notmatch the species delineation in the BJ lineage: morethan 95% nucleotide identity was found betweenB. japonicum and B. canariense, and betweenB. liaoningense and B. yuanmingense. Thus, the 95%ANI cut-off value for conserved genes seems unsuited tothe bradyrhizobial species delineation. However, suchobservation was expected since a universal cut-off valuemight not be relevant when studying the evolutionaryhistory of microorganisms as well as bacterial groupsevolving at various rates and being at different states ofspeciation.

Konstantinidis et al. [19] concluded that existingtaxonomic designations, including the species level,correspond frequently to a continuum of genetic


diversity as opposed to natural groupings. This seemsparticularly true for the Bradyrhizobium genus where itis possible to identify clearly three lineages in Bradyrhi-

zobium (BJ, BE and BP) by MLSA with isolates withineach lineage exhibiting a continuum of sequencediversity that seriously complicates species boundarydelineation. A solution might be to increase significantlythe number of isolates and habitats to describephylogeographic patterns for rhizobial species as re-commended by Vinuesa et al. [49].

Acknowledgements

This work was funded by the Bureau des RessourcesGenetiques (BRG, France). AN was funded by a Ph.D.grant from Institut de Recherche pour le Developpe-ment (IRD, France). The authors thank two anonymousreviewers for helpful comments.

Appendix. Supplementary materials

The online version of this article contains additionalsupplementary data. Please visit doi:10.1016/j.syapm.2009.06.002.

References

[1] D. Alazard, Stem and root nodulation in Aeschynomene

spp., Appl. Environ. Microbiol. 50 (2000) 732–734.

[2] C. Appunu, A. N’Zoue, G. Laguerre, Genetic diversity of

native bradyrhizobia isolated from soybeans (Glycine

max L.) in different agricultural–ecological–climatic

regions of India, Appl. Environ. Microbiol. 74 (2008)

5991–5996.

[3] J.H. Degnan, N.A. Rosenberg, Discordance of species

trees with their most likely gene trees, PLoS Genet. 2

(2006) e68.

[4] M. Dowton, A.D. Austin, Increased congruence does not

necessarily indicate increased phylogenetic accuracy—the

behavior of the incongruence length difference test in

mixed-model analyses, Syst. Biol. 51 (2002) 19–31.

[5] B.D. Eardly, S.M. Nour, P. van Berkum, R.K. Selander,

Rhizobial 16S rRNA and dnaK genes: mosaicism and the

uncertain phylogenetic placement of Rhizobium galegae,

Appl. Environ. Microbiol. 71 (2005) 1328–1335.

[6] R.C. Edgar, MUSCLE: multiple sequence alignment with

high accuracy and high throughput, Nucleic Acids Res. 32

(2004) 1792–1797.

[7] J.S. Farris, M. Kallersjo, A.G. Kluge, C. Bult, Testing

significance of incongruence, Cladistics 10 (1994)

315–319.

[8] D. Fleischman, D. Kramer, Photosynthetic rhizobia,

Biochim. Biophys. Acta. 1364 (1998) 17–36.

[9] M.W. Gaunt, S.L. Turner, L. Rigottier-Gois, S.A. Lloyd-

Macgilp, J.P. Young, Phylogenies of atpD and recA


dx.doi.org/10.1016/j.syapm.2009.06.002

dx.doi.org/10.1016/j.syapm.2009.06.002

dx.doi.org/10.1016/j.syapm.2009.06.002


support the small subunit rRNA-based classification

of rhizobia, Int. J. Syst. Evol. Microbiol. 51 (2001)

2037–2048.

[10] E. Giraud, L. Moulin, D. Vallenet, V. Barbe, E. Cytryn,

J.C. Avarre, M. Jaubert, D. Simon, F. Cartieaux, Y. Prin,

G. Bena, L. Hannibal, J. Fardoux, M. Kojadinovic, L.

Vuillet, A. Lajus, S. Cruveiller, Z. Rouy, S. Mangenot, B.

Segurens, C. Dossat, W.L. Franck, W.S. Chang, E.

Saunders, D. Bruce, P. Richardson, P. Normand, B.

Dreyfus, D. Pignol, G. Stacey, D. Emerich, A. Vermeglio,

C. Medigue, M. Sadowsky, Legumes symbioses: absence

of Nod genes in photosynthetic bradyrhizobia, Science

316 (2007) 1307–1312.

[11] J. Goris, K.T. Konstantinidis, J.A. Klappenbach, T.

Coenye, P. Vandamme, J.M. Tiedje, DNA–DNA hybri-

dization values and their relationship to whole-genome

sequence similarities, Int. J. Syst. Evol. Microbiol. 57

(2007) 81–91.

[12] S. Guindon, O. Gascuel, A simple, fast, and accurate

algorithm to estimate large phylogenies by maximum

likelihood, Syst. Biol. 52 (2003) 696–704.

[13] S. Guindon, F. Lethiec, P. Duroux, O. Gascuel, PHYML

Online—a web server for fast maximum likelihood-based

phylogenetic inference, Nucleic Acids Res. 33 (2005)

W557–W559.

[14] K. Haukka, K. Lindstrom, J.P. Young, Three phyloge-

netic groups of nodA and nifH genes in Sinorhizobium and

Mesorhizobium isolates from leguminous trees growing in

Africa and Latin America, Appl. Environ. Microbiol. 64

(1998) 419–426.

[15] A.B. Hollis, W.E. Kloos, G.H. Elkan, DNA:DNA

hybridization studies of Rhizobium japonicum and

related Rhizobiaceae, J. Gen. Appl. Microbiol. 123

(1981) 215–222.

[16] J.P. Huelsenbeck, F. Ronquist, R. Nielsen, J.P. Bollback,

Bayesian inference of phylogeny and its impact on

evolutionary biology, Science 294 (2001) 2310–2314.

[17] D.C. Jordan, Transfer of Rhizobium japonicum Buchanan

1980 to Bradyrhizobium gen. nov., a genus of slow

growing root nodule bacteria from leguminous plants,

Int. J. Syst. Bacteriol. 32 (1982) 136–139.

[18] K.T. Konstantinidis, A. Ramette, J.M. Tiedje, Toward a

more robust assessment of intraspecies diversity, using

fewer genetic markers, Appl. Environ. Microbiol. 72

(2006) 7286–7293.

[19] K.T. Konstantinidis, J.M. Tiedje, Prokaryotic taxonomy

and phylogeny in the genomic era: advancements and

challenges ahead, Curr. Opin. Microbiol. 10 (2007)

504–509.

[20] L.M. Kuykendall, B. Saxena, T.E. Devine, S.E. Udell,

Genetic diversity in Bradyrhizobium japonicum Jordan

1982 and a proposal for Bradyrhizobium elkanii sp. nov.,

Can. J. Microbiol. 38 (1992) 501–505.

[21] G. Laguerre., S.M. Nour, V. Macheret, J. Sanjuan, P.

Drouin, N. Amarger, Classification of rhizobia based on

nodC and nifH gene analysis reveals a close phylogenetic

relationship among Phaseolus vulgaris symbionts, Micro-

biology 147 (2001) 981–993.

[22] M. Martens, P. Dawyndt, R. Coopman, M. Gillis, P. de

Vos, A. Willems, Advantages of multilocus sequence


analysis for taxonomic studies: a case study using 10

housekeeping genes in the genus Ensifer (including former

Sinorhizobium), Int. J. Syst. Evol. Microbiol. 58 (2008)

200–214.

[23] D.P. Martin, C. Williamson, D. Posada, RDP2: recombi-

nation detection and analysis from sequence alignments,

Bioinformatics 21 (2005) 260–262.

[24] F. Molouba, J. Lorquin, A. Willems, B. Hoste, E. Giraud,

B. Dreyfus, M. Gillis, P. de Lajudie, C. Masson-Boivin,

Photosynthetic bradyrhizobia from Aeschynomene spp.

are specific to stem-nodulated species and form a separate

16S ribosomal DNA restriction fragment length poly-

morphism group, Appl. Environ. Microbiol. 65 (1999)

3084–3094.

[25] L. Moulin, A. Munive, B. Dreyfus, C. Boivin-Masson,

Nodulation of legumes by members of the beta-subclass

of Proteobacteria, Nature 411 (2001) 948–950.

[26] L. Moulin, G. Bena, C. Boivin-Masson, T. Stepkowski,

Phylogenetic analyses of symbiotic nodulation genes

support vertical and lateral gene co-transfer within the

Bradyrhizobium genus, Mol. Phylogenet. Evol. 30 (2004)

720–732.

[27] K.B. Nicholas, B.J. Nicholas, Genedoc: a tool for editing

and annotating multiple sequence alignments, distributed

by the author, 1997.

[28] D. Posada, K.A. Crandall, MODELTEST: testing the

model of DNA substitution, Bioinformatics 14 (1998)

817–818.

[29] R. Rivas, A. Willems, J.L. Palomo, P. Garcia-Benavides,

P.F. Mateos, E. Martinez-Molina, M. Gillis, E. Velaz-

quez, Bradyrhizobium betae sp. nov., isolated from

roots of Beta vulgaris affected by tumour-like

deformations, Int. J. Syst. Evol. Microbiol. 54 (2004)

1271–1275.

[30] R. Rivas, M. Martens, P. de Lajudie, A. Willems,

Multilocus sequence analysis of the genus Bradyrhizo-

bium, Syst. Appl. Microbiol. 32 (2009) 101–110.

[31] F. Ronquist, J.P. Huelsenbeck, MrBayes 3: Bayesian

phylogenetic inference under mixed models, Bioinfor-

matics 19 (2003) 1572–1574.

[32] J. Rozas, J.C. Sanchez-DelBarrio, X. Messeguer, R.

Rozas, DnaSP, DNA polymorphism analyses by the

coalescent, other methods, Bioinformatics 19 (2003)

2496–2497.

[33] J. Sambrook, D.W. Russel, Molecular Cloning, a

Laboratory Manual, third ed., C.S.H.L. Press, Cold

Spring Harbor, New York, 2001.

[34] H. Sawada, L.D. Kuykendall, J.M. Young, Changing

concepts in the systematics of bacterial nitrogen-fixing

legume symbionts, J. Gen. Appl. Microbiol. 49 (2003)

155–179.

[35] B. Shapiro, A. Rambaut, A.J. Drummond, Choosing

appropriate substitution models for the phylogenetic

analysis of protein-coding sequences, Mol. Biol. Evol.

23 (2006) 7–9.

[36] H. Shimodaira, M. Hasegawa, Multiple comparisons of

log-likelihoods with applications to phylogenetic infer-

ence, Mol. Biol. Evol. 16 (1999) 1114–1116.

[37] R.B. So, J.K. Ladha, J.P. Young, Photosynthetic

symbionts of Aeschynomene spp. form a cluster with


dx.doi.org/10.1016/j.syapm.2009.06.002


bradyrhizobia on the basis of fatty acid and rRNA

analyses, Int. J. Syst. Bacteriol. 44 (1994) 392–403.

[38] E. Stackebrandt, W. Frederiksen, G.M. Garrity, P.A.

Grimont, P. Kampfer, M.C. Maiden, X. Nesme, R.

Rossello-Mora, J. Swings, H.G. Truper, L. Vauterin,

A.C. Ward, W.B. Whitman, Report of the ad hoc

committee for the re-evaluation of the species definition

in bacteriology, Int. J. Syst. Evol. Microbiol. 52 (2002)

1043–1047.

[39] E.T. Steenkamp, T. Stepkowski, A. Przymusiak, W.J.

Botha, I.J. Law, Cowpea and peanut in southern Africa

are nodulated by diverse Bradyrhizobium strains harbor-

ing nodulation genes that belong to the large pantropical

clade common in Africa, Mol. Phylogenet. Evol. 48 (2008)

1131–1144.

[40] T. Stepkowski, M. Czaplinska, K. Miedzinska, L.

Moulin, The variable part of the dnaK gene as an

alternative marker for phylogenetic studies of rhizobia

and related alpha Proteobacteria, Syst. Appl. Microbiol.

26 (2003) 483–494.

[41] D.L. Swofford, PAUP*. Phylogenetic Analysis Using

Parsimony (*and Other Methods), Version 4, Sinauer

Associates, Sunderland, 1998.

[42] K. Tamura, J. Dudley, M. Nei, S. Kumar, MEGA4:

molecular evolutionary genetics analysis (MEGA) soft-

ware, version 4.0, Mol. Biol. Evol. 24 (2007) 1596–1599.

[43] S.L. Turner, J.P. Young, The glutamine synthetases of

rhizobia: phylogenetics and evolutionary implications,

Mol. Biol. Evol. 17 (2000) 309–319.

[44] D. Vallenet, L. Labarre, Z. Rouy, V. Barbe, S. Bocs, S.

Cruveiller, A. Lajus, G. Pascal, C. Scarpelli, C. Medigue,MaGe: a microbial genome annotation system supported

by synteny results, Nucleic Acids Res. 34 (2006) 53–65.

[45] P. van Berkum, J.M. Leibold, B.D. Eardly, Proposal for

combining Bradyrhizobium spp. (Aeschynomene indica)

with Blastobacter denitrificans and to transfer Blastobac-

ter denitrificans (Hirsch and Muller, 1985) to the genus

Bradyrhizobium as Bradyrhizobium denitrificans (comb.

nov.), Syst. Appl. Microbiol. 29 (2006) 207–215.

[46] J.M. Vincent, A Manual for the Practical Study of Root-

Nodule Bacteria. I.B.P. Handbook, vol. 15, Blackwell

Scientific Publications, Ltd., Oxford, 1970.

[47] P. Vinuesa, M. Leon-Barrios, C. Silva, A. Willems, A.

Jarabo-Lorenzo, R. Perez-Galdona, D. Werner, E.

Martinez-Romero, Bradyrhizobium canariense sp. nov.,


an acid-tolerant endosymbiont that nodulates endemic

genistoid legumes (Papilionoideae: Genisteae) from the

Canary Islands, along with Bradyrhizobium japonicum bv.

genistearum, Bradyrhizobium genospecies alpha and Bra-

dyrhizobium genospecies beta, Int. J. Syst. Evol. Micro-

biol. 55 (2005) 569–575.

[48] P. Vinuesa, C. Silva, D. Werner, E. Martinez-Romero,

Population genetics and phylogenetic inference in

bacterial molecular systematics: the roles of migration

and recombination in Bradyrhizobium species cohesion

and delineation, Mol. Phylogenet. Evol. 34 (2005)

29–54.

[49] P. Vinuesa, K. Rojas-Jimenez, B. Contreras-Moreira,

S.K. Mahna, B.N. Prasad, H. Moe, S.B. Selvaraju, H.

Thierfelder, D. Werner, Biogeography and evolutionary

genetics of four Bradyrhizobium species that nodulate

soybeans on the Asiatic continent, assessed by multilocus

sequence analysis, Appl. Environ. Microbiol. 74 (2008)

6987–6996.

[50] A. Willems, F. Doignon-Bourcier, R. Coopman, B.

Hoste, P. de Lajudie, M. Gillis, AFLP fingerprint analysis

of Bradyrhizobium strains isolated from Faidherbia albida

and Aeschynomene species, Syst. Appl. Microbiol. 23

(2000) 137–147.

[51] A. Willems., R. Coopman, M. Gillis, Comparison of

sequence analysis of 16S–23S rDNA spacer regions,

AFLP analysis and DNA–DNA hybridizations in

Bradyrhizobium, Int. J. Syst. Evol. Microbiol. 51 (2001)

623–632.

[52] A. Willems, R. Coopman, M. Gillis, Phylogenetic and

DNA–DNA hybridization analyses of Bradyrhizobium

species, Int. J. Syst. Evol. Microbiol. 51 (2001) 111–117.

[53] A. Willems, A. Munive, P. de Lajudie, M. Gillis, In most

Bradyrhizobium groups sequence comparison of 16S–23S

rDNA internal transcribed spacer regions corroborates

DNA–DNA hybridizations, Syst. Appl. Microbiol. 26

(2003) 203–210.

[54] L.M. Xu, C. Ge, Z. Cui, J. Li, H. Fan, Bradyrhizobium

liaoningense sp. nov., isolated from the root nodules of

soybeans, Int. J. Syst. Bacteriol. 45 (1995) 706–711.

[55] Z.Y. Yao., F.L. Kan, E.T. Wang, G.H. Wei, W.X. Chen,

Characterization of rhizobia that nodulate legume species

of the genus Lespedeza and description of Bradyrhizobium

yuanmingense sp. nov, Int. J. Syst. Evol. Microbiol. 52

(2002) 2219–2230.


dx.doi.org/10.1016/j.syapm.2009.06.002

Documents

Multilocus sequence analysis of bradyrhizobia isolated from Aeschynomene species in Senegal