PLANT MICROBE INTERACTIONS

Genetic Diversity Patterns and Functional Traitsof Bradyrhizobium Strains Associated with Pterocarpus officinalisJacq. in Caribbean Islands and Amazonian Forest(French Guiana)

Christine Le Roux & Félix Muller & Jean-Marc Bouvet &Bernard Dreyfus & Gilles Béna & Antoine Galiana &

Amadou M. Bâ

Received: 7 May 2013 /Accepted: 10 February 2014# Springer Science+Business Media New York 2014

Abstract Pterocarpus officinalis Jacq. is a legume tree nativeto the Caribbean islands and South America growing as adominant species in swamp forests. To analyze (i) the geneticdiversity and (ii) the symbiotic properties of its associatednitrogen-fixing soil bacteria, root nodules were collected fromP. officinalis distributed in 16 forest sites of the Caribbeanislands and French Guiana. The sequencing of the 16S-23Sribosomal RNA intergenic spacer region (ITS) showed that allbacteria belonged to the Bradyrhizobium genus. Bacteria iso-lated from insular zones showed very close sequence homol-ogies with Bradyrhizobium genospecies V belonging to theBradyrhizobium japonicum super-clade. By contrast, bacteriaisolated from continental region displayed a larger geneticdiversity and belonged to B. elkanii super-clade. Two strains

from Puerto Rico and one from French Guiana were notrelated to any known sequence and could be defined as anew genospecies. Inoculation experiments did not show anyhost specificity of the Bradyrhizobium strains tested in termsof infectivity. However, homologous Bradyrhizobium sp.strain-P. officinalis provenance associations were more effi-cient in terms of nodule production, N acquisition, and growththan heterologous ones. The dominant status ofP. officinalis inthe islandsmay explain the lower bacterial diversity comparedto that found in the continent where P. officinalis is associatedwith other leguminous tree species. The specificity in efficien-cy found between Bradyrhizobium strains and host tree prov-enances could be due to a coevolution process between bothpartners and needs to be taken in consideration in the frame-work of rehabilitation plantation programs.

Introduction

Pterocarpus officinalis Jacq. (Leguminosae) is the dominantwetland tree species of the seasonally flooded swamp forestsin the Caribbean basin [3, 26, 27]. It covers large areas of thecoastal floodplain, individual trees, and small stands occurringalong rivers and in mountains [8]. During the last two centu-ries, the distribution of P. officinalis was greatly reduced whencoastal plain forests were cut and drained for humanbuildings and agriculture. Nowadays, most populations ofP. officinalis are restricted to areas adjacent to mangroves,where changes in hydrology and salinity scale could affecttheir survival [32].

As most legumes, P. officinalis is able to establish mutual-istic relationships with nitrogen-fixing bacteria and arbuscularmycorrhiza [2, 13, 34]. Since freshwater swamp forests

Christine Le Roux and Félix Muller contributed equally to this paper.

C. Le Roux : F. Muller :A. Galiana (*)“Laboratoire des Symbioses Tropicales et Méditerranéennes,” UMRLSTM, CIRAD, Campus international de Baillarguet, TA A-82/J,34398 Montpellier cedex 5, Francee-mail: antoine.galiana@cirad.fr

F. Muller :A. M. BâLaboratoire de Biologie et Physiologie Végétales, UFR des SciencesExactes et Naturelles, Université des Antilles et de la Guyane, BP592, 97159 Pointe-à-Pitre, Guadeloupe, France

F. Muller : J.<M. Bouvet“Genetic Diversity and Breeding of Forest Tree Species,” UMRAGAP, CIRAD, TA A-108/01, 34398 Montpellier cedex, 5, France

B. Dreyfus :G. Béna :A. M. Bâ“Laboratoire des Symbioses Tropicales et Méditerranéennes,” UMRLSTM, IRD, Campus international de Baillarguet, TA A-82/J,34398 Montpellier cedex 5, France

Microb EcolDOI 10.1007/s00248-014-0392-7

present slow mineralization rates and denitrification [5],legume trees growing in such environment have to rely onnitrogen-fixing symbiosis to fulfill their high N demand[23]. Nodulated legume trees have been reported as morecommon in flooded than in terra firme forest in FrenchGuiana [9, 20], as well as in the Orinoco basin in Venezuela[5, 36]. Severe waterlogging is known to affect nodulationin a majority of legumes but does not seem to restrainnodulation and nitrogen fixation in P. officinalis, indicatingits good adaptation to swamp forests [13, 20]. Furthermore,flooding of P. officinalis seedlings induces several morpho-logical and physiological adaptation mechanisms like forma-tion of hypertrophied lenticels, aerenchyma tissue, adventi-tious roots, and nodulation on submerged portions of thestem [13]. Stem and root nodulation appeared to be efficientin terms of nitrogen fixation and therefore contributed toplant growth and nutrition in flooding conditions. Despitethis original adaptation, the diversity of rhizobia associatedwith P. officinalis has not been studied so far. The diversityof this symbiosis is more especially interesting to study asnatural stands of P. officinalis, though threatened, still existin both continental and insular zones. Several studies onrhizobia-legume associations have compared the genetic di-versity and structure of both symbiotic partners [1, 28],suggesting that coevolution may have occurred between hostplants and their bacterial partners. However, none of themhas been conducted in an insular context so far, wherenumerous specificities exist such as their isolated statusand the small size of populations and number of colo-nizers [17].

In previous studies aiming at characterizing the geneticdiversity of P. officinalis, we found that the insular populationswere less diversified than the continental ones [26, 27]. Con-comitantly to this study, nodules were collected from roots ofthese different P. officinalis populations in order (i) to isolateand investigate the diversity and distribution of the associatednodulating bacteria originating from French Guiana and dif-ferent Caribbean islands and (ii) to determine the host speci-ficity and functional traits of these associated bacteria tohighlight a possible coevolution of both symbiotic partners.

Materials and Methods

Sampling Strategy and Isolation of Rhizobial Strains

Nodules were sampled from individuals of 16 populations ofP. officinalis in the Caribbean islands of Guadeloupe (10populations), Marie-Galante (1 population), Puerto Rico (1population), Martinique (1 population), Dominica (1 popula-tion), and on the continent in French Guiana (2 populations)along rivers, in mountains, and in wetland areas adjacent tomangroves. The number of populations and individuals varied

from one island to another. Populations were generally smalland scattered in each island, except in Guadeloupe where ahigh density of individuals is found on large areas. Therefore,sampling was unbalanced in terms of number of individualsper island varying from 13 (Puerto Rico) to 33 (Belle-Plaine,Guadeloupe) [27]. The minimum sampling intensity was ap-plied to 20 individuals per population and 10 nodules wereharvested from each P. officinalis tree sampled, in parallel withthe collection of seeds as described by Muller et al. [27].Nodules were immediately dried in silica gel tubes for furtherbacterial isolation.

Before isolation, the dry nodules were transferred intosterile water for 30 min for rehydration. Two to five noduleswere sampled per individual for isolation of bacteria. Noduleswere surface sterilized by immersion in 30 % H2O2 for 7 minand rinsed four times with sterile distilled water. The noduleswere individually crushed in drops of sterile yeast-mannitol(YM) broth. The nodule suspensions were then plated on YMagar medium [37]. Plates were incubated at 28 °C, and purecultures of bradyrhizobial strains were obtained after severalsubculture steps.

16S-23S rRNA Gene ITS Sequencing and PhylogeneticAnalysis

A loopful of rhizobial cells was suspended in 20 μl of sterilewater and boiled for 5 min. Cell debris were removed bycentrifugation at 13,000 rpm for 1 min at room temperature,and 2 μl of the supernatant was used as a template for PCR.Twenty-five microliters of reaction mixture, containing200 μM of each deoxynucleoside triphosphate, 0.8 μM ofeach primer, 1.5 mM of MgCl2, 1.25 U of Taq DNA poly-merase (Promega, Charbonières, France), and the buffer sup-plied with the enzyme, was used for PCR amplification aPerkin Elmer model 2400 thermocycler (Perkin Elmer Ap-plied Biosystems, Norwalk, USA).

The partial internal transcribed spacer (ITS) of the 16S and23S ribosomal RNA (rRNA) genes was amplified usingprimers BR5 CTTGTAGCTCAGTTGGTTAG [40] andFGPL132′ CCGGGTTTCCCCATTCGG [29]. The PCRwas done as follows: initial denaturation at 96 °C for 3 minfollowed by 35 cycles consisting of a 30-s denaturation at95 °C, 30-s at an annealing temperature of 55 °C, followed bya 40-s primer extension at 72 °C. Then, the PCR productswere run on a 1 % agarose gel (Sigma, France) in TAE bufferwith a DNA size standard (Eurogentec Smartladder). Ampli-fied fragments were purified with a QIAquick Gel ExtractionKit (Qiagen, France). Purified PCR products were sequencedon both strands using the same primers BR5 and FGPL132′.Sequencing was performed using a BigDye Terminator CycleSequencing kit (Perkin Elmer Applied Biosystems, FosterCity, USA), and reactions were analyzed on an automated

C. Le Roux et al.

DNA sequencer (Applied Biosystems model 310, PerkinElmer Applied Biosystems).

The ITS sequences of P. officinalis Bradyrhizobium spp.strains were corrected using the sequence viewer 4Peaksprogram from Mekentosj B.V. Multiple alignment and phylo-genetic tree construction were performed using the SeaViewprogram version 4 [18]. This interface drives the ClustalOmega program and includes the BioNJ distance-based treereconstruction method [15]. The 16S-23S rRNA ITS se-quences of Bradyrhizobium reference strains were retrievedfrom the DDBJ/EMBL/GenBank databases and included inthe phylogenetic tree, along with the 66 sequences of theBradyrhizobium spp. strains isolated from P. officinalis andother related strains. A bootstrap analysis using 100 replica-tions was performed. The ITS sequences of Bradyrhizobiumspp. strains isolated from P. officinalis were deposited in theGenBank database under accession numbers GQ433640 toGQ433705 (see Table 1). Mothur software [35] was used todifferentiate the operational taxonomic units (OTUs) at acutoff of 0.03 from all the sequences obtained.

Cross-Inoculation Tests and Host Plant Response

Pods were collected from three native populations ofP. officinalis: Crique Alexandre Jacques in French Guiana,Le Moule in Guadeloupe, and the nearby island of Marie-Galante. The choice of these provenances was limited by thelack of viable seeds among the other provenances from Gua-deloupe and French Guiana. They were shelled, and seedswere surface sterilized with a 3 % sodium hypochlorite (w/w)solution for 10 min [13]. Then, they were rinsed several timesin sterile water, transferred aseptically onto 0.08 % water agarin Petri dishes, and incubated for 4–8 days at 30 °C in the dark.Sterilized seeds with approximately 3-cm long tap roots weretransferred into test tubes using the Gibson’s device [16]. Thetubes containing a modified Jensen’s N-free liquid plant nu-trient medium [37] were incubated at 30 °C in a growthcabinet (light intensity 20 W m−2, days length 12 h, 30 °Cday/25 °C night). After 10 days of growth, the seedlings wereinoculated with 1 ml (approximately 109 CFU per ml of YMmedium) of some representatives from each group ofbradyrhizobial strains isolated from nodules of ninePterocarpus provenances: D 6a1 (Dominica), PR 1_3 (PuertoRico), MG 17_1 (Marie Galante, Guadeloupe), G 1857(Rivière Bananier, Guadeloupe), GlmO 3_2 (Le Moule, Gua-deloupe), Dde 17_1 (Deshaies, Guadeloupe), Gbp 9_2 (Belle-Plaine, Guadeloupe), M 8_1 (Martinique), and FG 5_1(Crique Alexandre Jacques, French Guiana). Six replicateswere tested for each bradyrhizobial strain. Uninoculatedplants were added as controls. The experiment was set up asa two-factor analysis consisting of 3 provenances ofP. officinalis and 10 bradyrhizobial strains (including the un-inoculated control treatment).

The relative effects of inoculation on plant growth werecompared by measuring the height (cm) and dry weight (g dryweight) of leaves, shoots, and roots (7 days at 60 °C) of plantsafter 4 months of growth. Plants were also scored for nodula-tion. After drying, ground leaf samples were mineralizedthrough heating at 500 °C and digested in hydrochloric acidfor determination of N. The total N contents of leaves wereassessed using Technicon AutoAnalyzer and expressed as thepercentage of the leaf dry weight.

All data were subjected to two-way analyses of variance,and mean values were compared by the Newman and Keulsmultiple range test using the XLSTAT TM package (version2012, Addinsoft, Paris, France). The mean number of noduleswas calculated from arcsine (square root) transformed data.

Results

Nearly 100 bacterial strains were isolated from 400 rootnodules sampled from the 16 swamp forest sites of the naturaldistribution area of P. officinalis in the Caribbean islands andFrench Guiana. About one third of these isolates werediscarded since they did not renodulate P. officinalis. Ampli-fication and sequencing of the 16S-23S rRNA intergenicspacer region (ITS) were successfully performed on 66 ofthese isolated strains. Of these 66 sequenced isolates, 2 orig-inated from one site inMarie-Galante island, 3 from one site inDominica, 4 from one site in Martinique, 7 from two sites inFrench Guiana, 10 from one site in Puerto Rico, and 40 fromnine sites in Guadeloupe, i.e., a mean of 4.4 isolates per siteoverall (Table 1). All P. officinalis associated bacteriabelonged to the Bradyrhizobium genus.

The mean length of the 16S-23S rRNA ITS region se-quenced was 443 bp overall. A phylogenetic tree based onthis ITS region was generated by BioNJ method (Fig. 1). Thebacterial strains isolated in this study and the reference strainswere divided into two super-clades. The first super-clade,containing both Bradyrhizobium japonicum genospecies typeIV and type I, included Bradyrhizobium strains originatingfrom all P. officinalis provenances except French Guiana andwas composed of 51 isolates from the Lesser Antilles (Gua-deloupe, Martinique, Marie-Galante, and Dominica) and 7isolates from Puerto Rico. Sequences from this latter groupwere closely related (about 100 % of identity for most of thesequences) to Bradyrhizobium sp. genospecies V LMG11955strain [40] and at a lesser extent (about 97 % of identity formost of the sequences) to Bradyrhizobium yuanmingense typestrain CCBAU 10071T. The second super-clade, containingthe Bradyrhizobium elkanii type strain, included sevenBradyrhizobium strains isolated from nodules of P. officinalissampled in two different sites of French Guiana, i.e., FG 5_1,FRG 7_1, FG 9_1, and FG 15_2 from Crique AlexandreJacques site and FG 7_10_10, FG 7_10_13, and FG

Diversity of Pterocarpus officinalis bradyrhizobia

11_10_1 from Paracou site, and the two Bradyrhizobiumstrains PR 1_3 and PR 1_2 originating from the Puerto Ricosite. All these strains, except the two latter and FG 11_10_1,were closely related to the reference Bradyrhizobium sp.Tv2a-2 strain isolated from Tachigali versicolor, a tree speciesfrom the Papilionoideae subfamily native to Panama. TheB. japonicum and B. elkanii super-clades were subdivided intodistinct clusters, i.e., cluster 1 and clusters 2 to 4, respectively,as defined by the mothur software [35] at a cutoff of 0.03.Thus, all strains originating from the Caribbean islands, ex-cept strains PR 1_3 and PR 1_2, were grouped into cluster 1affiliated to Bradyrhizobium genosp. V. On the other hand, allstrains originating from French Guiana were grouped intocluster 2, except strain FG 11_10_1 belonging to cluster 3 asunique strain. Lastly, strains PR 1_3 and PR 1_2 originatingfrom Porto Rico were grouped into cluster 4.

Among the three P. officinalis provenances tested, seed-lings originating from Marie-Galante produced a signifi-cantly (P<0.001) higher biomass than those from CriqueAlexandre Jacques (French Guiana) and Le Moule

(Guadeloupe) after 4 months of growth, all Bradyrhizobiumstrains combined, as attested by two-way analysis of vari-ance (Table 2). Conversely, seedlings from the Marie-Galante provenance had a lower shoot height than thosefrom the two other provenances (Table 2). By contrast, thethree host plant provenances tested exhibited significantlydifferent abilities to nodulate seedlings from French Guianaproducing 54 and 148 % more nodules than those from LeMoule and Marie-Galante, respectively, all inoculationtreatments combined (Table 2). In the same way, the FrenchGuiana provenance had a significantly (P<0.001) higherleaf N content than the two other ones. On the other hand,the nine Bradyrhizobium strains tested exhibited signifi-cantly (P<0.05) different efficiencies, ranging from strainG 1857 (Rivière Bananier, Guadeloupe) that produced thehighest shoot biomass and strain D 6_1 (Dominica) thelowest one, all provenances combined. As for P. officinalisprovenances, no correlation was found between shoot dryweight and shoot height, and only seedlings inoculated withthe Bradyrhizobium strains Gbp 29_1 and M 8_1 had

Table 1 Origin, name, and GenBank accession number of the 66 bacterial strains isolated from nodules collected from 16 sites of the natural distributionarea of Pterocarpus officinalis in the Caribbean islands and French Guiana

Island/country Site Bradyrhizobium strain (GenBank accession no.)

Dominica Indian River D 6_1 (GQ433679), D 6a1 (GQ433651), D 12_1 (GQ433678)

French Guiana Crique Alexandre Jacques FG 5_1 (GQ433642), FG 7_1 (GQ433641), FG 9_1 (GQ433640),FG 15_2 (GQ433643)

Paracou FG 7_10_10 (GQ433645), FG 7_10_13 (GQ433646), FG 11_10_1(GQ433644)

Guadeloupe Belle-Plaine (Grande terre) Gbp 1_1 (GQ433698), Gbp 2_1 (GQ433671), Gbp 3_1 (GQ433674), Gbp 3_2(GQ433675), Gbp 6_2 (GQ433677), Gbp 8_1 (GQ433699), Gbp 9_2(GQ433654), Gbp 12_2 (GQ433652), Gbp 13_1 (GQ433670), Gbp 20_1(GQ433672), Gbp 21_1 (GQ433673), Gbp 34_2 (GQ433676), Gbpom 1(GQ433697), Gbpom 2 (GQ433686), Gbpom 3 (GQ433687), Gbpom 5(GQ433688), Gbpom 6 (GQ433696)

Blain (Basse terre) G 1856 (GQ433702), G 1864 (GQ433669)

Deshaies (Basse terre) G 1866 (GQ433705), Gde 17_1 (GQ433700)

Grande Ravine (Grande terre) G 1844 (GQ433657), G 1859 (GQ433666)

Jarry (Basse terre) G 1847 (GQ433701)

Le Moule (Grande terre) G 1847 (GQ433659), G 1849 (GQ433660), G 1854 (GQ433663),Glmo 3_2 (GQ433682), Glmo 9_1 (GQ433680), Glmo 12_2(GQ433681)

Pointe à Retz (Grande terre) G 1850 (GQ433661), G 1858 (GQ433665)

Port Louis (Grande terre) G 1851 (GQ433662), G 1860 (GQ433704)

Rivière Bananier (Basse terre) G 1852 (GQ433656), G 1857 (GQ433703), G 1863 (GQ433668)

Viard (Basse terre) G 1846 (GQ433658), G 1855 (GQ433664), G 1862 (GQ433667)

Marie-Galante Marie Galante MG 8_1 (GQ433685), MG 17_1 (GQ433684)

Martinique Galion M 1a2 (GQ433650), M 1a3 (GQ433653), M 1b1 (GQ433649),M 8_1 (GQ433683)

Puerto Rico Sabana Seca PR 1_2 (GQ433648), PR 1_3 (GQ433647), PR 2_3 (GQ433692),PR 5_1 (GQ433695), PR 13_2b (GQ433689), PR 13_3 (GQ433690),PR 16_2 (GQ433691), PR 16_3 (GQ433655), PR 20_2 (GQ433694),PR 20_3 (GQ433693)

C. Le Roux et al.

significantly (P<0.001) lower shoot heights than thosefrom other treatments, except the uninoculated controlplants (Table 2). We did not observe any host specificityof the Bradyrhizobium strains tested in terms of infectivity,since they all indifferently formed nodules among the threeP. officinalis provenances tested, regardless of the geo-graphical or host origin of the strains (Table 2). However,the number of nodules differed significantly (P<0.001)according to the strain tested. This latter parameter wascorrelated with leaf N content (r=+0.79) but not with both

plant dry weight and shoot height. There was also a signif-icant interaction (P<0.001) between the three host plantprovenances and the nine Bradyrhizobium strains tested forall the parameters analyzed.

When homologous and heterologous combinations be-tween Bradyrhizobium strains FG 5_1, MG 17_1, and Glmo3_2 and P. officinalis provenances from Crique AlexandreJacques (FG), Marie-Galante (MG), and Le Moule (Glmo)were only considered through a separate two-way analysis ofvariance, we observed significant effects of both

LMG8443T (AF338176)

BradyrhizobiumB. denitrificans

genosp. IX LMG10677 (AJ534597)

Bradyrhizobium genosp. VII LMG10671 (AJ279272)

B. elkanii genosp. II LMG6134T (AJ279308)

Bradyrhizobium genosp. X LMG10705 (AJ534592)

Bradyrhizobium genosp. XI LMG11951 (AJ534594)

Bradyrhizobium genosp. VI LMG8293 (AJ279311)

B. pachyrhizi PAC48T (AY628092)89

94

80

B. lablabi CCBAU 23086T (GU433583)

B. jicamae PAC68T (AY628094)

98

100

Bradyrhizobium sp. STM3567 (FJ002413)

PR 1_3 (2)FG 11_10_1 (1)

Bradyrhizobium sp. Tv2a-2 (AY187818)

FG 7_10_13 (1) FG 5_1 (1)

FG 7_10_10 (1) FG 15_2 (1) FG 9_1 (1)FG 7_1 (1)

93

100

86

99

87

82

89

95

Gbp 9_2 (1)

G 1861 (4) M 1a3 (1)M 1b1 (4)

G 1866 (1)

Gbpom 6 (5) G 1852 (1)G 1844 (39)

PR 16_3 (1)Bradyrhizobium genosp. V LMG 11955 (AJ534590)

96

Bradyrhizobium sp. CIRADAc12 (DQ311111)

B. yuanmingense CCBAU 10071T (AY386734)

88

B. daqigense CCBAU15774T (HQ231312)

B. japonicum genosp. IV LMG 8321 (AJ279317)

B. iriomotense LMG 24129T (AB300993)

B. arachidis CCBAU 051107T (HM107198)

B. huanghuaihaiense CCBAU 23303T (HQ428043)

B. japonicum genosp. I LMG 6138T (AJ279264)

B. canariense BTA-1T (AY386708)

B. betae LMG 21987T (AJ631967)

B. liaoningense genosp. III LMG 18230T (AJ279301)97

85

0. 02

Cluster 1 (insular strains)

Cluster 2 (continental strains)

Cluster 4 (insular strains)Cluster 3 (continental strain)

Fig 1 BioNJ phylogenetic tree based on 16S-23S rRNA intergenicsequences of 66 Bradyrhizobium spp. strains isolated from Pterocarpusofficinalis, reference and related strains. Number of strains with similarsequences are indicated in brackets. The clusters were defined by mothursoftware [35] and corresponded to OTUs differentiated at a cutoff of 0.03.Only bootstrap probability values higher than 80 % (100 replications) aregiven at the branching points. Gaps were not considered. Scale indicated2% sequence divergence.Bradyrhizobium denitrificanswas chosen as anoutgroup. Strains with similar sequences not represented in tree: G 1844=

G 1846, G 1847, G 1849, G 1850, G 1851, G 1854, G 1855, G 1858, G1859, G 1862, G 1863, G 1864, Gbp 13_1, Gbp 2_1, Gbp 20_1, Gbp21_1, Gbp 3_1, Gbp 3_2, Gbp 34_2, Gbp 6_2, D 12_1, D 6_1, Glmo9_1, Glmo 12_2, Glmo 3_2, M 8_1, MG 17_1, MG 8_1, Gbpom 2,Gbpom 3, Gbpom 5, PR 13_2b, PR 13_3, PR 16_2, PR 2_3, PR 20_2,PR 20_3, PR 5_1; PR 1_3=PR 1_2; M1b1=M 1a2, D 6a1, Gbp 12_2;G1861 = G1856, G1857, G1860; Gbpom 6=Gbpom 1, Gbp 1_1, Gbp8_1, Gde 17_1

Diversity of Pterocarpus officinalis bradyrhizobia

Bradyrhizobium strain and P. officinalis provenance factorsand their interaction for almost all parameters analyzed(Table 3). Thus, significantly higher nodule numbers and leafN contents were found in the three homologous associationstested compared to the heterologous ones (Table 3). The shootheight and plant dry weight were higher in two and onehomologous combinations, respectively.

Discussion

Based on the sequencing of the 16S-23S rRNA ITS region,the 66 bacterial nodulating strains isolated from the nineP. officinalis provenances were shown to belong exclusivelyto Bradyrhizobium, whatever their geographical origin. Eventhough Bradyrhizobium spp. have been reported as beingpreferentially associated with a majority of legume tree spe-cies from tropical rainforests, regardless of their subfamilyaffiliation [9, 10, 24, 31], such exclusivity of host species likeP. officinalis towards Bradyrhizobium has rarely been reportedin the literature since rainforest legume tree species are oftenfound associated with low proportions of non-bradyrhizobial

s t r a ins a l though predominan t ly nodu la ted byBradyrhizobium.

The Bradyrhizobium strains isolated in this study exhibitedintrageneric diversity and ranged into either of the two widelyrecognized B. japonicum and B. elkanii super-clades. Thisphylogenetic differentiation depends here on the geographicalorigin of the bacterial strains since all P. officinalis strainsoriginating from the Caribbean islands, except PR 1_3 andPR 1_2, clustered with B. japonicum branch, whereas allrhizobial isolates originating from the Amazonian areabelonged to the B. elkanii branch. Hence, the P. officinalis-Bradyrhizobium symbiosis remains quite original since, bycontrast, there is no correlation between the phylogeny ofbacterial isolates and their geographical origin in most tropicaltree legume-rhizobium associations as observed in Austral-asian acacias [4, 21, 22].

The B. japonicum and B. elkanii super-clades weresubdivided into one and three distinct clusters, respectively,as defined by the mothur software at a cutoff of 0.03. Withinthe super-clade that includes B. elkanii, strain FG 11_10_1grouped in cluster 3 and strains PR 1_3 and PR 1_2 groupedin cluster 4 were not related with any known referenceBradyrhizobium sp. strain, the closest reference strain being

Table 2 Effects of inoculationwith nine bradyrhizobial strains on plant dry weight, shoot height, nodule number, and leaf N content of threePterocarpusofficinalis provenances after 4 months of plant growth

Factor tested1 Treatment Plant dry weight(g plant−1)

Shoot height(cm)

Nodule number(per plant)

Leaf N content(g kg−1)

Bradyrhizobium strain G 1852 1.72 a 38.4 a 19.2 abc 12.8 cd

M 8_1 1.59 b 35.6 ab 15.9 bc 11.4 de

FG 5_1 1.58 b 37.9 a 25.0 ab 15.5 b

Gde 17_1 1.52 bc 37.4 a 19.2 abc 11.6 de

Gbp 29_1 1.52 bc 34.2 b 12.3 c 10.4 e

PR 13_3 1.49 bcd 39.2 a 11.3 c 14.2 bc

Glmo 3_2 1.49 bcd 38.6 a 26.6 a 15.1 b

MG 17_1 1.43 cd 37.0 a 21.1 abc 17.3 a

D 6_1 1.40 d 38.2 a 17.5 abc 14.5 bc

Uninoculatedcontrol2

1.19 29.5 0 7.4

Pterocarpus provenance Marie-Galante (MG) 1.83 a 35.3 b 11.0 c 13.0 b

French Guiana, Crique A.Jacques (FG)

1.39 b 39.1 a 27.3 a 16.7 a

Guadeloupe, Le Moule (Glmo) 1.36 b 37.8 a 17.7 b 11.3 c

Origins of Bradyrhizobium strains are indicated in Table 11 The effects of Bradyrhizobium strain and P. officinalis provenance factors were tested through two-way analyses of variance showing significant effectsof both factors and their interaction on the four parameters analyzed at P<0.001 (6 replicates were used in the 30 strain×provenance combinationstested). For each parameter, the means per Bradyrhizobium strain or per P. officinalis provenance followed by different letters are significantly differentaccording to the Newman and Keuls multiple range test at P=0.052 The uninoculated control treatments were not included in both the two-way Anova and the Newman and Keuls tests presented here, (i) considering thefar lower values obtained for the four parameters measured compared to those obtained with the inoculated treatments and (ii) in order to not interfere inthe ranking of the different Bradyrhizobium strains tested. The two-way ANOVA including the uninoculated control treatments showed similar highlysignificant effects at P<0.001 of both factors and their interaction on all parameters (results not reported here)

C. Le Roux et al.

Bradyrhizobium genospecies IX LMG10677 strain with asequence identity of 81 % according to Blast results. Moreprecise taxonomic position should be further clarified forthese three strains by DNA-DNA hybridization analysis orMulti Locus Sequence Analysis [38] to know whether theycould constitute a new species or genospecies.

Although two strains isolated from Porto Rico, i.e., PR 1_3and PR 1_2 were not related to cluster 1 as all the 57 otherstrains originating from the Caribbean islands, they representan exception among this population. Since strains from FrenchGuiana were underrepresented with an unbalanced proportioncompared to that of islands (2 sampling sites in French Guianavs 14 sites distributed in five islands), isolation of a highernumber of strains from French Guiana equivalent to that ofCaribbean strains would have been necessary to better com-pare the levels of diversity and heterogeneity between bothprovenances.

The high degree of similarity between Bradyrhizobiumstrains isolated from the Caribbean islands could be explainedby the monospecific status of the P. officinalis populations inthis region where they can represent up to 90 % of the tree

stratum [26, 32]. By contrast, in French Guiana, P. officinalis,which represents about 25 % of the tree canopy and 37.5 % ofnodulated leguminous plants [20], coexists with other treespecies in highly diverse forests that contain a majority oflegume tree species mostly nodulated by bradyrhizobia [9,25]. This higher diversity of plant species did not seem toinduce a higher diversity and heterogeneity amongBradyrhizobium strains isolated from P. officinalis in FrenchGuiana although they were not represented enough to con-clude. Generally, plant species richness is known to impactecosystem processes through the diversification in composi-tion and functions of soil microbial communities [41]. Inaddition to plant species diversity as a factor of microbialdiversification, the degree of heterozygosity of P. officinalisat intraspecific level could enhance the diversity observedamong bradyrhizobia as stated by a previous study performedin the same locations showing that the continental populationsof P. officinalis from French Guiana showed higher values ofheterozygosity than the island populations [27]. However,while many studies have shown that rhizobial diversity variedaccording to plant line, variety, or cultivar in annual crop

Table 3 Effects of cross inoculation of three Pterocarpus officinalis provenances with three bradyrhizobial strains on plant dry weight, shoot height,nodule number, and leaf N content after 4 months of plant growth

Factor tested1 Treatment2 Plant dry weight(g plant−1)

Shoot height(cm)

Nodule number(per plant)

Leaf N content(g kg−1)

P. officinalis provenance Marie-Galante (MG) 1.85 a 34.8 b 30.9 a 13.4 b

French Guiana,Crique A. Jacques (FG)

1.36 b 40.0 a 20.8 b 21.5 a

Guadeloupe, Le Moule (Glmo) 1.29 c 38.6 a 20.9 b 13.0 b

Significance level P<0.001 P<0.001 P<0.05 P<0.001

Bradyrhizobium strain FG 5_1 1.58 a 37.9 a 25.0 a 15.5 b

Glmo 3_2 1.49 b 38.6 a 26.6 a 15.1 b

MG 17_1 1.42 c 37.0 a 21.1 a 17.3 a

Significance level P<0.001 NS NS P<0.05

Interaction provenance×strain MG×FG 5_1 1.81 34.8 14.8 9.3

MG×Glmo 3_2 1.87 36.2 35.0 9.0

MG×MG 17_1 1.86 33.5 43.0 22.0

FG×FG 5_1 1.48 40.4 40.3 26.7

FG×Glmo 3_2 1.40 40.0 11.2 18.7

FG×MG 17_1 1.19 39.5 10.8 19.2

Glmo×FG 5_1 1.46 38.5 19.8 10.5

Glmo×Glmo 3_2 1.19 39.6 33.5 17.7

Glmo×MG 17_1 1.21 37.9 9.5 10.8

Significance level P<0.001 NS P<0.001 P<0.001

Origins of Bradyrhizobium strains are indicated in Table 1

NS not significant1 The effects of Bradyrhizobium strain and P. officinalis provenance factors were tested through two-way analyses of variance (six replicates were used inthe nine strain×provenance combinations tested). For each parameter, the means per Bradyrhizobium strain or per P. officinalis provenance followed bydifferent letters are significantly different according to the Newman and Keuls multiple range test at P=0.052 The homologous Bradyrhizobium strain×P. officinalis provenance associations are indicated in bold characters. When homologous associations gavehigher values than heterologous ones for each parameter analyzed, these values were indicated in bold characters

Diversity of Pterocarpus officinalis bradyrhizobia

legume species [30], little is known about the relationshipbetween genetic variation among natural populations andprovenances of legume tree species at intraspecific level anddiversity of their associated rhizobia.

It is well-known that soil composition influences rhizobialdiversity [39, 42]. However, the separation ofBradyrhizobiumstrains into two major clades according to their geographicalorigin is probably not due to different soil properties betweenthe island and mainland areas. Even though the Guianeseforest soils are very different from those of the Caribbeanislands, P. officinalis is encountered on a large variety of soilsin the insular areas and even within the same island, as inGuadeloupe where different P. officinalis forest types can bedistinguished according to soil substrate like volcanic acidicsoil, carbonate-rich sedimentary soil, peat-swamp soil, or soilcontaining high clay content [6].

It is also noticeable that the water salinity level in floodedareas has no effect on taxonomic affiliation of theBradyrhizobium strains isolated. This was more particularlyillustrated by the very high homology (99.8 % identity in 16S-23S rRNA sequence affiliated to Bradyrhizobium sp.genospecies V) between the strains originating from Belle-Plaine site (Guadeloupe), where the salinity gradient rangedfrom 26 to 2‰ at spatial level and from 22 to 5‰ at the end ofthe dry and wet season, respectively [12]. Furthermore, it waspreviously shown that both strains Gbpom 2 and Gbpom 6from Belle-Plaine site, originating from the areas with thelowest and highest salt concentrations, respectively, had asimilar response to salt tolerance in liquid culture medium aswell as in symbiosis with P. officinalis [33].

Under similar swampy conditions, Bradyrhizobium wasalso shown to be the dominant group of symbiotic N2-fixingbacteria associated with legume trees in French Guiana [20]and Brazilian Amazonia [9, 24, 25], although the precisetaxonomic affiliation of the few isolated strains was not de-termined in these studies. By contrast with salinity, wateracidity of swamp forests could have a selective effect onrhizobial populations that would result in the predominanceof Bradyrhizobium since these slow-growing bacteria areknown to have a better adaptation and tolerance to acidicconditions compared to the other rhizobia genera [19, 20].Besides, these Bradyrhizobium strains are able to toleratealternating cycles of drought and moisture as well as anoxicconditions during the different seasons affecting the soils oftropical swamp forests [20].

In this study, the representative strains of Bradyrhizobiumstrains isolated from the six different islands or regions inves-tigated, i.e., Guadeloupe, Marie-Galante, Martinique, Domi-nica, Puerto Rico, and French Guiana, were able to crossnodulate seedlings of three P. officinalis provenances fromGuadeloupe, French Guiana, and Marie-Galante. Our resultsclearly showed that the Bradyrhizobium strains isolated fromP. officinalis were not specific in terms of infectivity at host

plant provenance level. However, we found thatBradyrhizobium spp. can be highly specific in terms of strainefficiency despite their broad nodulation host range, as foundin other studies [7, 14, 22, 28]. The variation of efficiencyobserved between strains, based on shoot dry weight of thehost plant, was not related to their phylogenetic position, sincestrains with similar sequences might have different efficien-cies and, conversely, strains from 2 different clusters couldhave similar efficiencies (i.e., FG 5_1 compared to M 8_1,Gde 17_1, Gbp 29_1, PR 13_3, or Glmo 3_2). In addition, thehomologous associations tested between three bradyrhizobialstrains and three P. officinalis provenances were significantlymore infective and effective than the heterologous ones. Moreprecisely, the number of nodules and leaf N content signifi-cantly increased in each of the three P. officinalis provenancesinoculated with Bradyrhizobium strains collected from thesame site. By contrast, shoot height and shoot biomass ofthe strain-host homologous associations were not significantlydifferent compared to heterologous ones. However, leaf Ncontent is a more appropriate and accurate criterion forassessing rhizobium efficiency as it directly reflects theamount of N fixed in this experiment considering thatP. officinalis seedlings were grown onN-free nutrient medium.The present work also demonstrated that, within the cluster ofinsular bacterial strains, the range of effectiveness ofbradyrhizobial strains towards the P. officinalis provenancesdiffered. This could be due to the absence of correlationbetween the phylogeny of 16S-23S rRNA gene ITS and thatof symbiotic genes [11, 22]. Further works should be done toexplore the diversity of nodulation and nitrogen fixationgenes, especially nodA, nodC, and nifH, among theBradyrhizobium populations isolated in this study.

Our study shows the importance of constituting a collectionof rhizobial strains as broad and representative as possiblecovering the distribution area of the host plant in view ofinoculating selected strains in the context of ecosystem resto-ration programs. Moreover, the higher nitrogen-fixing effi-ciency found betweenBradyrhizobium strains andP. officinalisprovenances of the same origins, which must be confirmed ona larger number of homologous combinations, allows theimplementation of environmentally friendly inoculation tech-niques by preventing the introduction of exotic strains intonatural ecosystems while introducing strains adapted to localedaphoclimatic conditions.

Acknowledgments We would like to thank the French MinistryMEDD (Ministère de L’Ecologie et du Développement Durable) for thefinancial support to implement the Pterocarpus officinalis project. Thiswork was financially supported by the Guadeloupe Archipelago Regionand the Social European Fund. We are very grateful to colleagues fromthe University of Antilles-Guyane, especially A. Rousteau and A. Mitch-ell, as well as E. Rivera-Ocasio from the University of Puerto Rico, RíoPiedras for his assistance in the collection of biological samples, and to A.Vaillant and M. Poitel from CIRAD for their technical assistance in the

C. Le Roux et al.

laboratory. We thank H. Sanguin from CIRAD for his assistance in theuse of mothur software and L. Dedieu from Cirad for her careful andcritical reading of our manuscript.

References

1. Aguilar OM, Riva O, Peltzer E (2004) Analysis ofRhizobium etli andof its symbiosis with wild Phaseolus vulgaris supports coevolution incenters of host diversification. PNAS 101:13548–13553

2. Bâ AM, Samba R, Sylla S, Le Roux C, NeyraM, Rousteau A, ImbertD, Toribio A (2004) Caractérisation de la diversité desmicroorganismes symbiotiques de Pterocarpus officinalis dans desforêts marécageuses de Guadeloupe et Martinique. Rev Ecol 59:163–170

3. Bacon P (1990) Ecology and management swamp forest in theGuianas and Caribbean region. In: Lugo AE, Brinson M, Brown S(eds) Ecosystems of the World 15. Forested Wetlands, Elsevier,Amsterdam, pp 213–250

4. Bala A, Murphy P, Giller KE (2003) Distribution and diversity ofrhizobia nodulating agroforestry legumes in soils from three conti-nents. Mol Ecol 12:917–930

5. Barrios E, Herrera R (1994) Nitrogen cycling in Venezuelan tropicalseasonally flooded forest: soil nitrogen mineralization and nitrifica-tion. J Trop Ecol 10:399–416

6. Bonhême I, Rousteau A, Imbert D, Saur E (1998) La forêtmarécageuse à Pterocarpus officinalis : sa situation en Guadeloupe.Bois For Trop 258:59–68

7. Boukhatem ZF, Domergue O, Merabet C, Sekour S, Bouazza F, deLajudie P, Duponnois R, Bekki A, Galiana A (2012) Symbioticcharacterization and diversity of rhizobia associated with native andintroduced Acacias in arid and semi-arid regions in Algeria. FEMSMicrobiol Ecol 80:534–547

8. Cintrón BB (1983) Coastal freshwater swamp forests: Puerto Rico'smost endangered ecosystem. In: Lugo AE (ed) Los bosques de PuertoRico, Institute of Tropical Forestry, USDA 31. Forest Service, RioPiedras, Puerto Rico, pp 249–282

9. de Faria S, Diedhiou AG, de Lima HC, Ribeiro RG, Galiana A,Alexandre F, Castilho AF, Henriques JC (2010) Evaluating thenodulation status of leguminous species from the Amazonian forestof Brazil. J Exp Bot 63:4191–4212

10. Diabate M, Munive A, de Faria SM, Bâ AM, Dreyfus B, Galiana A(2005) Occurrence of nodulation in unexplored leguminous treesnative to the West African rain forest—application to rhizobial inoc-ulation of native species. New Phytol 166:231–239

11. Diouf D, Fall D, Chaintreuil C, Bâ AT, Dreyfus B, NeyraM, Ndoye I,Moulin L (2009) Phylogenetic analyses of symbiotic genes andcharacterization of functional traits of Mesorhizobium spp. strainsassociated with the promiscuous species Acacia seyal Del. J ApplMicrobiol 108:818–830

12. Dulormne M, Musseau O, Muller F, Bâ AM (2010) Effects of NaClon growth, water status, N2 fixation, and ion distribution inPterocarpus officinalis seedlings. Plant Soil 327:23–24

13. Fougnies L, Renciot S, Muller F, Plenchette C, Prin Y, de Faria SM,Bouvet JM, Sylla S, Dreyfus B, Bâ AM (2007) Arbuscular mycor-rhizal colonization and nodulation improve flooding tolerance inPterocarpus officinalis Jacq. seedlings. Mycorrhiza 17:159–166

14. Galiana A, Chaumont J, Diem HG, Dommergues YR (1990)Nitrogen-fixing potential of Acacia mangium and Acaciaauriculiformis seedlings inoculated with Bradyrhizobium andRhizobium spp. Biol Fertil Soils 9:261–267

15. Gascuel O (1997) BIONJ: an improved version of the NJ algorithmbased on a simple model of sequence data. Mol Biol Evol 14:685–695

16. GibsonAH (1980)Methods for legumes in glasshouse and controlledenvironment cabinet. In: Bergersen FJ (ed) Methods for EvaluatingBiological Nitrogen Fixation. Wiley-Interscience, Chichester, pp139–184

17. GomezA,Gonzales-Martinez SC, Collada C, Climent J, Gil L (2003)Complex population genetic structure in the endemic Canary Islandpine revealed using chloroplast microsatellite markers. Theor ApplGenet 107:1123–1131

18. Gouy MS, Guindon S, Gascuel O (2010) SeaView version 4: amultiplatform graphical user interface for sequence alignment andphylogenetic tree building. Mol Biol Evol 27:221–224

19. Graham P (1992) Stress tolerance in Rhizobium and Bradyrhizobium,and nodulation under adverse soil conditions. Can J Microbiol 38:475–484

20. Koponen P, Nygren P, Domenach AM, Le Roux C, Saur E,Roggy JC (2003) Nodulation and dinitrogen fixation of le-gume trees in tropical freshwater swamp forest in FrenchGuiana. J Trop Ecol 19:655–666

21. Lafay B, Burdon JJ (2001) Small-subunit rRNA genotyping ofrhizobia nodulating Australian Acacia spp. Appl Environ Microbiol67:396–402

22. Le Roux C, Tentchev D, Prin Y, Goh D, Japarudin Y, Perrineau MM,Duponnois R, Domergue O, de Lajudie P, Galiana A (2009)Bradyrh i zob ium nodu la t i ng the Acac ia mang ium xA. auriculiformis interspecific hybrid are specific and differ fromthose associated with both parental species. Appl EnvironMicrobiol 75:7752–7759

23. McKey D (1994) Legumes and nitrogen: the evolutionary ecology ofnitrogen-demanding lifestyle. In: Sprent JL, McKey D (eds)Advances in Legume Systematics 5, The Nitrogen Factor. RoyalBotanic Gardens, Kew, pp 211–229

24. Moreira FMS, Da Silva MF, de Faria SM (1992) Occurrence ofnodulation in legume species in Amazon region of Brazil. NewPhytol 121:563–570

25. Moreira FMS, Haukka K, Young JPW (1998) Biodiversity ofrhizobia isolated from a wide range of forest legumes in Brazil.Mol Ecol 7:889–895

26. Muller F, Vaillant A, Bâ AM, Bouvet JM (2006) Isolation andcharacterization of microsatellite markers in Pterocarpus officinalisJacq. Mol Ecol Notes 6:462–464

27. Muller F, VocciaM, BâAM, Bouvet JM (2009) Genetic diversity andgene flow in a Caribbean tree Pterocarpus officinalis Jacq.: a studybased on chloroplast and nuclear microsatellites. Genetica 135:185–198

28. Parker MA (2003) Genetic markers for analysing symbiotic relation-ships and lateral gene transfer in Neotropical bradyrhizobia.Mol Ecol12:2447–2455

29. Ponsonnet C, Nesme X (1994) Identification of Agrobacteriumstrains by PCR-RFLP analysis of pTi and chromosomal regions.Arch Microbiol 161:300–309

30. Rangin C, Brunel B, Cleyet-Marel JC, PerrineauMM, Béna G (2008)Effects of Medicago truncatula genetic diversity, rhizobial competi-tion, and strain effectiveness on the diversity of a naturalSinorhizobium species community. Appl Environ Microbiol 74:5653–5661

31. Rasolomampianina R, Bailly X, Fetiarison R, Rabevohitra R, BénaG, Ramaroson L, Raherimandimby M, Moulin L, de Lajudie P,Dreyfus B, Avarre JC (2005) Nitrogen-fixing nodules from rosewood legume trees (Dalbergia spp.) endemic to Madagascar hostseven different genera belonging to α- and β-Proteobacteria. MolEcol 14:4135–4146

32. Rivera-Ocasio E, Aide M, Rios-Lopez N (2007) The effects ofsalinity on the dynamics of a Pterocarpus officinalis forest stand inPuerto Rico. J Trop Ecol 23:559–568

33. Saint-Etienne L, Paul S, Imbert D, Dulormne M, Muller F, TorobioA, Plenchette C, Bâ AM (2006) Arbuscular mycorrhizal soil

Diversity of Pterocarpus officinalis bradyrhizobia

infectivity in a stand of the wetland tree Pterocarpus officinalis alonga salinity gradient. For Ecol Manag 232:86–89

34. Saur E, Bonhême I, Nygren P, Imbert D (1998) Nodulation ofPterocarpus officinalis in the swamp forest of Guadeloupe (lesserAntilles). J Trop Ecol 14:761–770

35. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M et al(2009) Introducing mothur: open-source platform-independentcommunity-supported software for describing and comparing micro-bial communities. Appl Environ Microbiol 75:7537–7541

36. Sprent JI (1999) Nitrogen fixation and growth of non-crop legumespecies in diverse environments. Perspectives in plant ecology. EvolSyst 2:149–162

37. Vincent JM (1970) A manual for the Practical Study of the Root-Nodule Bacteria. Blackwell Scientific, Oxford

38. Vinuesa P, Rojas-Jiménez K, Contreras-Moreira B, Mahna SK,Prasad BN, Moe H, Selvaraju SB, Thierfelder H, Werner D (2008)Multilocus sequence analysis for assessment of the biogeography and

evolutionary genetics of four Bradyrhizobium species that nodulatesoybeans on the Asiatic continent. Appl EnvironMicrobiol 74:6987–6996

39. Wang ET, Martinez-Romero J, Martinez-Romero E (2002) Geneticdiversity of rhizobia from Leucaena leucocephala nodules inMexican soils. Mol Ecol 8:711–724

40. Willems A, Coopman R, Gillis M (2001) Comparison of sequenceanalysis of 16 s-23s rDNA spacer regions, AFLP analysis and DNA-DNA hybridizations in Bradyrhizobium. Int J Syst Evol Microbiol51:623–632

41. Zak D, Holmes W, White D, Peacock AD, Tilman D (2003) Plantdiversity, soil microbial communities, and ecosystem function: arethere any link? Ecology 84:2042–2050

42. Zhang YM, Li Y, Chen WF, Wang ET, Tian CF, Li QQ, Zhang YZ,Sui XH, ChenWX (2011) Biodiversity and biogeography of rhizobiaassociated with soybean plants grown in the North China Plain. ApplEnviron Microbiol 77:6331–6342

C. Le Roux et al.