ABSTRACT
We applied a multilocus phylogenetic approach to elucidate the origin of serradella and lupin Bradyrhizobium strains that persist in soils of Western Australia and South Africa. The selected strains belonged to different randomly amplified polymorphic DNA (RAPD)-PCR clusters that were distinct from RAPD clusters of applied inoculant strains. Phylogenetic analyses were performed with nodulation genes (nodA, nodZ, nolL, noeI), housekeeping genes (dnaK, recA, glnII, atpD), and 16S-23S rRNA intergenic transcribed spacer sequences. Housekeeping gene phylogenies revealed that all serradella and Lupinus cosentinii isolates from Western Australia and three of five South African narrow-leaf lupin strains were intermingled with the strains of Bradyrhizobium canariense, forming a well supported branch on each of the trees. All nodA gene sequences of the lupin and serradella bradyrhizobia formed a single branch, referred to as clade II, together with the sequences of other lupin and serradella strains. Similar patterns were detected in nodZ and nolL trees. In contrast, nodA sequences of the strains isolated from native Australian legumes formed either a new branch called clade IV or belonged to clade I or III, whereas their nonsymbiotic genes grouped outside the B. canariense branch. These data suggest that the lupin and serradella strains, including the strains from uncultivated L. cosentinii plants, are descendants of strains that most likely were brought from Europe accidentally with lupin and serradella seeds. The observed dominance of B. canariense strains may be related to this species' adaptation to acid soils common in Western Australia and South Africa and, presumably, to their intrinsic ability to compete for nodulation of lupins and serradella.
Lupins (Lupinus spp.) were introduced as grain and forage crops into Australia and South Africa at the end of the 19th and beginning of the 20th centuries (10). The introduced lupins were species native to the Mediterranean region, including white lupin (Lupinus albus L.), narrow-leaf lupin (Lupinus angustifolius L.), blue lupin (Lupinus cosentinii L.), and yellow lupin (Lupinus luteus L.). Major centers of lupin cultivation became established in the winter rainfall regions of Western Australia and the Western Cape province of South Africa. Today, Australia is the leading world producer of lupin seeds used mainly in stock feed. In addition to the lupins, serradella (Ornithopus spp.) was introduced as a pasture and forage crop in Western Australia. As soils of Western Australia and South Africa did not contain indigenous rhizobia able to form effective nodules on lupins and serradella, inoculation was initially required for successful cultivation of these legumes (40). Following the introduction of inoculum strains in Western Australia in the early 1930s, inoculation of lupins and serradella was regarded as essential for the next 30 to 40 years. The widespread cultivation of lupins subsequently led to the establishment of lupin bradyrhizobia in soils of Western Australia. Effective nodulation of uninoculated narrow-leaf lupins was detected in the 1950s (12), and that of Lupinus cosentinii plants was detected even earlier (6). Effective populations became similarly established in soils of the Western Cape where inoculation of lupin is no longer considered necessary by many farmers. Interestingly, the naturally occurring populations found in both Western Australia and the Western Cape were dominated by strains differing from those used in commercial inoculants (4, 24, 40).
Lupins are nodulated mainly by slow-growing strains classified in the genus Bradyrhizobium (16). Although considerable diversity exists among lupin bradyrhizobia, all lupin strains cluster broadly with Bradyrhizobium japonicum but not with Bradyrhizobium elkanii (1, 19). Phylogenetic analyses of 16S rRNA (rrs) and dnaK gene sequences grouped the lupin Bradyrhizobium strain WM9 within the B. japonicum lineage, which corroborated the earlier studies. Large differences in gene composition were, however, detected in the symbiotic gene region of WM9 with respect to B. japonicum strain USDA 110 (38). The nodA gene sequence position of the lupin strain was also distant with respect to nodA sequences of other bradyrhizobia. Similar incongruence between the nod and dnaK and other housekeeping gene phylogenies was subsequently reported in a study of over 30 Bradyrhizobium strains, originating from a broad spectrum of legumes from different geographical regions (26). This work has differentiated three major clades among the Bradyrhizobium nodA sequences, one of which, called clade II, was composed exclusively of the strains isolated from lupins, broom (Cytisus scoparius), and serradella. Similar phylogenetic patterns detected for nodC and nifH gene sequences corroborated the unique position of the symbiotic genes of lupin and serradella bradyrhizobia (15, 49, 50). These findings led to the delineation of bradyrhizobia infecting legumes of the Genisteae and Loteae tribes in a new biovariety, genistearum. This single ecotype comprises bradyrhizobia that group with the species B. japonicum (B. japonicum bv. genistearum) together with the strains that form a related sister lineage, for which a new name, Bradyrhizobium canariense bv. genistearum, was recently proposed, and at least two other additional lineages called the alpha and beta genotypes (49, 50).
In the present study, we took advantage of the unique phylogenetic position characterizing lupin and serradella bradyrhizobia symbiotic genes to elucidate the origin of the diverse naturalized populations persisting in soils of Western Australia and South Africa. To our knowledge, this is the first attempt to infer the origin of naturalized rhizobia using multilocus phylogenetic analyses based on both symbiotic and housekeeping genes.
MATERIALS AND METHODS
Bacterial strains. Bradyrhizobium strains included in the study and their characteristics are described in Table 1. A table describing reference strains used in the phylogenetic trees is presented in the supplementary material. Yeast extract mannitol agar medium (48) was used for growth and maintenance of strains. All Bradyrhizobium strains were grown at 28°C.
Bradyrhizobium strains used in this study
The Bradyrhizobium strains were collected in the winter rainfall region of Western Australia and the South African Western Cape region. The selected Western Australian strains represented Bradyrhizobium populations isolated from three sites (Albany, Carrabin, and Yelbeni). Isolates were recovered either from the nodules of serradella plants growing at the sites (Ornithopus compressus and Ornithopus sativus) or from the nodules of bait plants grown in soil collected from the sites (O. compressus, O. sativus, and L. angustifolius) (24). The Albany, Carrabin, and Yelbeni sites were characterized by acid (pH 4.6 to 6.4) soils. At all sites, either narrow-leaf lupin or serradella had been grown prior to the isolation of strains. At each site, the inoculant strain WU425 had been extensively applied by farmers for the 20 years prior to sampling, and additionally, at the Yelbeni site, the inoculant strain WSM471 had been introduced (13). The selected strains belonged to distinct randomly amplified polymorphic DNA (RAPD)-PCR clusters, as assessed using the combined datasets (24) for the primers RP04, RP05, and RP01 (35). The analysis of RAPD-PCR profiles revealed large diversity (>50% dissimilarity between most strain types), showing that all populations were composed of multiple strain types (24). The selected strains represent both dominant and unique strain types for the given site; however, all had RAPD-PCR profiles that were distinct from the WU425 and WSM471 inoculant strains isolated in Western Australia from serradella nodules (13, 28). Five strains, including WSM468, WSM469, WSM1420, WU8, and WU140, originated from uncultivated Lupinus cosentinii plants growing at Albany, Western Australia. Two strains, WSM2155 and WSM2156, were isolated from nodulated plants of Daviesia nudiflora Meissn. and Mirbelia floribunda Benth., respectively, growing at Carrabin within 20 m of serradella trials and a narrow-leaf lupin crop. We also included the A13 and W72 strains that had been used as inocula in Western Australia before strains WU425 and WSM471 were developed (5). Four strains originated from native Western Australian legumes, whereas six other strains were isolated by Lafay and Burdon (19) from native legumes growing in New South Wales (Table 1).
Unlike Western Australian strains, all five South African strains were isolated from Lupinus angustifolius nodules. The lupin plants were grown in four moderately acid (pH 5.5 to 5.8) soil samples collected at Hopefield, Langewens, Eendekuil, and Tygerhoek, and one moderately alkaline (pH 7.6) soil from Roodebloem. Strains 1S20, 2S9, 3S16, and 5S10 originated from the acid soils, whereas strain 4S16 came from the alkaline site. The acid soils hosted large, >5,000 rhizobia g of soil−1, and diverse Bradyrhizobium populations, as RAPD-PCR analyses with BOXA1R, ERIC, M13 phage, and REP primers showed. On the other hand, the alkaline soil population was small, <400 rhizobium cells g of soil−1, and homogenous. These strains were serologically distinct from the current inoculant strain WU425, imported from Australia, and the preceding strain VK7 isolated in the Western Cape from lupin (4).
All strains were tested for their ability to nodulate yellow serradella (O. compressus cv. Avila) plants as described elsewhere (38).
Molecular techniques.For PCR amplification experiments, total genomic DNA was isolated using the sodium dodecyl sulfate-proteinase K lysis procedure described by Moulin et al. (26) or using the QIAGEN QIAamp DNA mini kit following the recommendations of the producer. All primers used for PCR amplification are listed in Table 2. The dnaK sequences were amplified and sequenced by following the procedure described elsewhere (39). Other markers—nodA, nodZ, noeI, nolL, intergenic transcribed spacer (ITS), atpD, glnII, and recA gene fragments—were amplified using the following procedure: 95°C for 2 min followed by 35 cycles of 95°C for 45 s, 58°C for 30 s, and 72°C for 1.5 min (2.5 min for nodA) and a final elongation step of 7 min at 72°C, as recommended for the FastStart High-Fidelity PCR system by the producer (Roche Diagnostics GmbH, Germany). The annealing temperature for amplification of nodA, nodZ, and noeI was 53°C, whereas the annealing temperature for nolL was 50°C. The PCR products were purified with a QIAquick gel extraction kit (QIAGEN, Germany). The sequencing reactions were performed as described elsewhere (26). The sequences were deposited in the EMBL database under the accession numbers given in Table 3.
Oligonucleotides used for PCR amplificationa
PCR amplification and accession numbers of sequenced PCR productsa
Sequence analyses.Multiple nucleotide sequence alignments were generated using ClustalX (45) and optimized manually using Genedoc software (27). All phylogenetic analyses were performed using PAUP, version 4.0b10 (44). Due to the large number of sequences, parsimony and maximum-likelihood (ML) analyses were performed using the heuristic search option of PAUP.
An intragenic recombination test was assessed on each alignment using recombinational detection prediction (23), which combines the RDP, GeneConv, Bootscan, MaxChi, Chimaera, and SiScan programs. Intragenic recombination (by at least two programs) was detected only for strain 4S16 in the glnII phylogeny. This strain was retained in the phylogenetic trees, though its placement should be considered as unresolved in the glnII phylogeny.
For the nodA, nodZ, noeI, nolL, atpD, dnaK, glnII, and recA gene phylogenies, ML and neighbor-joining (NJ) analyses were performed using PAUP (44). The best-fit model for each gene was assessed using MODELTEST3.6 (33). ML phylogenies using a codon partition were also performed, i.e., specific substitution rates were evaluated following the codon structure of the DNA sequence (this model allows one to estimate the site-specific substitution rates and thus evaluate data saturation, which may disturb the phylogenetic signal). The best substitution models found for each marker were as follows: dnaK, TrN+G (Tamura-Nei plus Gamma); glnII and recA, GTR+I+G (General Time Reversible with invariant sites and a gamma rate distribution); atpD, codon site-specific rates.
The nodA phylogeny was constructed using the same ML model as described by Moulin et al. (26). The base frequencies, Ti/Tv ratios, and substitution rates were estimated at each position from the data using ML analyses. The third position in the codon was excluded from the analyses, as it accumulated too much saturation (26). The ITS phylogeny was assessed using parsimony analyses as described by Vinuesa et al. (49). Parsimony analyses allow gaps to be treated as informative (as a fifth base) and were preferred due to the high number of insertions/deletions found in the 16S-23S intergenic sequence. The ITS alignment consisted of a 1,056-bp fragment, of which 374 bp showed insertion/deletion events. This alignment can be obtained upon request. The ITS phylogeny was assessed using only Bradyrhizobium strains, as ITS from outer-group bacteria would be impossible to align correctly due to their large sequence divergence. For the atpD, recA, dnaK, and glnII phylogenies, the bootstrapping analyses were performed using the heuristic searches under distance and ML models (using PAUP) with 1,000 and 100 replicates, respectively. Due to the large datasets, bootstrap values on nodA and ITS phylogenies were assessed using only parsimony and NJ analyses.
The partition homogeneity tests (100 random trees; 1,000 replicates) and Shimodaira-Hasegawa tests of congruence of tree topologies were performed using PAUP on a restricted data set of 13 Bradyrhizobium strains (USDA94, USDA76, 1S20, 4S16, USDA110, WSM2150, WSM2154, WSM2157, WSM471, WSM2150, 3S16, BC-C2, and WU425) for which sequences were available for all markers. The result of this analysis is presented in the supplemental material.
Nucleotide sequence accession number.Sequences were deposited in the EMBL database under the accession numbers given in Table 3.
RESULTS
Bradyrhizobium canariense dominates serradella and lupin isolates from Western Australia and South Africa.The nodulation and nitrogen fixation abilities of the strains were evaluated on yellow serradella (O. compressus cv. Avila). Pink nodules were formed on the plants inoculated with Western Australian serradella strains and with four of the five South African narrow-leaf lupin strains. The exception was strain 4S16 (the alkaline soil isolate), which formed only small, white nodules. Effective nodules were also found on serradella plants inoculated with the blue lupin (L. cosentinii) isolates and with strains WSM2155 and WSM2156 (isolated from two Western Australian native species). No nodules were found on plants inoculated with other Western Australian and New South Wales native legume isolates (Table 1).
The taxonomic position of each Bradyrhizobium strain used in this study was initially verified by phylogenetic analysis of dnaK gene sequences. The dnaK tree was constructed with partial (278 bp) sequences, encoding the variable 3′ end gene fragment (Fig. 1A). This part of the 70-kDa chaperone gene accumulates a large number of substitutions, giving a tree topology that corroborates trees of other gene phylogenies, including 16S rRNA, recA, and glnII gene markers (39). We decided to use additional markers, taking into account the uncertainty related to phylogenetic inference based on a single gene marker analysis, especially for the genus Bradyrhizobium (29). The selected gene markers included the ITS region between the 16S rRNA (rrs) and 23S rRNA (rrl) genes, ATP synthase beta chain (atpD), glutamine synthetase II (glnII), and recombination protein recA genes. All of these gene markers have been used by other workers for phylogenetic reconstructions of Rhizobiaceae (9, 46, 49, 50, 51). It is noteworthy that the selected genes occupy distinct positions on the chromosome of Bradyrhizobium japonicum strain USDA 110 (17). Thus, assuming that synteny occurs between Bradyrhizobium chromosomes (31), we expected that these gene markers would yield phylogenetic reconstructions indicative of the whole genome (not just single genes).
(A) ML tree based on partial dnaK sequences; (B) maximum-parsimony tree based on complete 16S-23S rRNA intergenic transcribed sequences. New strains from this study are indicated in boldface type. Host plant and geographical origin are indicated between brackets. The scale bar (only for the ML dnaK tree) indicates the number of substitutions per site. The bootstrap values indicated were calculated under distance criteria (1,000 replicates). Abbreviations: B., Bradyrhizobium; BC, B. canariense; B. jap., B. japonicum; B. liaon., B. liaoningense; B. yuan., B. yuanmingense; B. IX, Bradyrhizobium genospecies IX (defined by Willems et al.) (52); Rho., Rhodopseudomonas; NSW, New South Wales Australia; WA, Western Australia; CI, Canary Islands; SA, South Africa; USA, United States; La, Lupinus angustifolius; Lc, Lupinus cosentinii; Oc, Ornithopus compressus; Op, Ornithopus pinnatus; Os, Ornithopus sativa; Cham., Chamaecytisus proliferus; Ad, Adenocarpus; Te. and Tel., Teline; Lesp., Lespedeza; L., Lupinus; G., Glycine; F, Faidherbia; inoc., inoculum.
ML and NJ phylogenies of dnaK revealed that, apart from two lupin isolates, 1S20 and 4S16 from South Africa, the remaining lupin and serradella strains formed a well-defined branch (78% bootstrap support), that was related to (but distinct from) the Bradyrhizobium japonicum sequences (Fig. 1A). The only sequence that was determined earlier (39), and which clustered in this new branch, was the sequence of B. canariense strain BC-C2. Two isolates from Australian native legumes (WSM2155 and WSM2156) were also located in this cluster. Pairwise sequence similarity within this cluster ranged from 94% to 100%, although the majority of sequences were 98 to 99% related. The sequence of the inoculant strain W72 was identical to the sequences of strains WSM866, WSM2153, WSM2154, and 5S10. Similarly, two widely used inoculant strains, WU425 and VK7, showed identical dnaK sequences, as well as strains WSM2150, WSM2156, and WU140; WSM468, WSM2149, and WU8; and WSM2151, WSM2155, WSM2158, WSM2159, and WSM2458.
ML and NJ phylogenetic trees of the ITS sequences of 25 strains confirmed the relationships found for lupin and serradella strains using the dnaK phylogeny. Each tree contained a well-supported branch which, in addition to the lupin and serradella strains under investigation, was composed exclusively of previously sequenced B. canariense strains (Fig. 1B). The B. canariense branch essentially showed the same grouping as the dnaK tree. Thus, strains A13, W72, WU425, VK7, and WSM2154 had identical ITS sequences, as did WSM2150 and WU140, WSM468 and WU8, and WSM2151 and WSM2155. B. canariense strain ISLU16 was identical to strains A13, W72, WU425, VK7, and WSM2154. The remaining strains from Western Australia shared ≥98% similarity with the other B. canariense strains. Similar topology was revealed in the recA phylogenetic tree obtained with 23 recA strain sequences (Fig. 2A), confirming the distinct position of lupin and serradella isolates and their grouping with B. canariense. The grouping with the B. canariense branch was further supported by ML and NJ trees of glnII (Fig. 2B) and atpD (Fig. 2C), although the latter two genes were sequenced in only eight strains.
ML phylogenetic trees based on partial recA (A), glnII (B), and atpD (C) sequences. New strains from this study are indicated in boldface type. Scale bars indicate the number of substitutions per site. The bootstrap values indicated were calculated under distance criteria (1,000 replicates); they are in boldface type when a bootstrap value of ≥80% was obtained also using ML models (100 replicates). Abbreviations are defined in the legend to Fig. 1.
Strains 1S20 and 4S16 were the only exceptions among the lupin and serradella isolates that consistently grouped outside the B. canariense branch. However, their placement differed between the phylogenetic trees, according to the marker gene used. On the dnaK tree (Fig. 1A), 1S20 formed a distinct clade with BDV5029, BDV5493, and USDA3001, whereas on the recA tree (Fig. 2A), it grouped with BDV5029. Strain 1S20 also occupied distinct positions on the atpD and glnII trees, although the placement was weakly supported (Fig. 2B and C). On the ITS tree, strain 1S20 grouped with 4S16, close to B. japonicum type strain USDA 6. On the recA and atpD trees, strain 4S16 grouped close to Bradyrhizobium liaoningense strain LMG18230, while on the dnaK tree strain 4S16 occupied a distinct position. These phylogenies suggested that strain 1S20 may represent a new lineage, whereas 4S16 may belong to B. liaoningense or be related to this species.
The dnaK and recA trees showed that native Australian legume bradyrhizobia (other than strains WSM2155 and WSM2156) were either situated within the B. japonicum clade (BDV5325), occupied a position close to the Bradyrhizobium yuanmingense cluster (WSM1743), or formed a cluster with B. elkanii strains (WSM1704). Strain BDV5057 occupied a distinct position on the dnaK tree, whereas it was placed in the B. yuanmingense cluster on the ITS tree and closer to B. japonicum on the recA tree (Fig. 1 and 2). Two other strains, WSM1735 and WSM1790, formed a separate clade on the ITS and recA trees, whereas on the dnaK tree, they grouped with B. japonicum.
Partition homogeneity (incongruence length difference) tests and Shimodaira-Hasegawa tests of congruence of tree topologies (see the supplemental material) confirmed in part the analogy of topologies obtained in the ITS, dnaK, recA, atpD, and glnII phylogenies but only on restricted datasets. Topology congruence was found between the ITS, atpD, and dnaK trees but not between the glnII and recA trees. Partition homogeneity tests also reflected homogeneity between markers on small partitions (2 to 3 markers) but not on partitions implementing all markers (see the supplemental material). Such statistical incongruence between housekeeping genes has already been reported, even if trees show identical topologies (9, 46). We thus relied on bootstrap values to distinguish clades in the phylogenetic trees. Our data indicated that the Western Australia serradella isolates and three South African narrow-leaf lupin isolates grouped with B. canariense strains with high bootstraps and thus represent authentic B. canariense isolates, although DNA-DNA hybridizations would be needed to confirm our results.
Similarity of lupin and serradella nodA gene sequences to clade II bradyrhizobia.Following the strategy described earlier (26), we amplified in all strains from lupin and serradella nodules a DNA fragment that comprised the intergenic region between nodD and nodA, the complete nodA sequence, and part of the nodB gene. An exception was WSM2158, for which only a partial sequence covering the 5′ half of the gene was obtained. Moreover, we amplified the nodA gene in three strains from legumes native to Western Australia, except strain WSM1704, for which we could not obtain a nodA, nor any other nod gene (nodZ, nolL, and noeI) product. Complete nodA gene sequences were, however, amplified in five (of the six) Bradyrhizobium strains that originated from New South Wales native legume species.
All nodA sequences of the strains isolated from lupins and serradella were closely related, with identities ranging from 90 to 100%. In the ML tree, all sequences formed a common branch (100% bootstrap support) with other lupin and serradella strains of European or North American origin as well as with strains from other species of the tribe Genisteae isolated in Europe and Japan (Fig. 3) and strains WSM2155 and WSM2156, isolated from Western Australian native legumes. This branch was earlier referred to as clade II (26). However, only nodA sequences of L. cosentinii isolate WU140 and the narrow-leaf lupin 4S16 were intermingled with formerly sequenced strains. The remaining lupin and serradella strains from Western Australia and South Africa formed a distinct subgroup within clade II, with 95% bootstrap support. This subgroup is composed of very similar sequences (identity ranging from 97% to 100%). The strains WSM2151, WSM2152, WSM2155, WSM2157, and WSM2159 had identical sequences. Similarly, strains WU425, WSM2153, and 3S16; WSM2148 and WSM2156; and WSM2154 and WSM2458 had 100% similarity. This branch included nodA sequences of strains WSM866 and WSM1253 originating from the Mediterranean (Greece) which have never been applied as inocula in Western Australia.
ML phylogenetic tree of nodA sequences. The tree was constructed by the ML approach with a model described in Materials and Methods. Details of the model are summarized in the inset. Only the Bradyrhizobium-Burkholderia-Methylobacterium branch is shown. The scale bar indicates the number of substitutions per site. Bootstrap values (percentage of 1,000 replicates under distance criteria, only values >75%) are given at the branching nodes. Bradyrhizobium country of origin and host plant are in brackets. New strains from this study are indicated in boldface type. Clade numbers correspond to the nodA phylogeny defined by Moulin et al. (27). Abbreviations are defined in the legend to Fig. 1.
The nodA sequence of WSM1743 grouped within clade III (26), which includes sequences of strains mainly of (sub)tropical origin. Two other native legume isolates from Western Australia, WSM1735 and WSM1790, formed a separate clade that occupied a sister position with respect to clade II. This new clade is termed clade IV. The nodA sequences of the two strains were 98% similar. All five New South Wales native legume isolates grouped in a branch called clade I together with nodA sequences of two strains isolated in South Africa and Brazil, respectively, from Acacia spp. native to Australia (26). All nodA sequences within clade I showed 98 to 99% identity. All deduced NodA proteins were composed of 211 amino acid residues (unlike the majority of Bradyrhizobium NodA proteins built of 210 residues), with a single codon insertion (aspartate 183) in the C-terminal part, which seems a distinctive characteristic of this clade (data not shown).
PCR amplification revealed the presence of the nodZ gene in all lupin and serradella isolates. Additionally, this gene was amplified in strain WSM1743 of clade III, and in strains WSM1735 and WSM1790 of clade IV. The nodZ gene was not amplified from the template DNAs of the New South Wales isolates which represented clade I (Table 3). This gene is responsible for fucosylation of Nod factor reducing end glucosamine, which appears to be a common Nod factor modification in the genus Bradyrhizobium (8, 34, 36). Phylogenetic sequence analysis of the nodZ sequences of strains belonging to clades II, III, and IV yielded essentially the same branching pattern as that of the nodA tree. The above data were consistent with the previous study (26), which showed that strains belonging to clades II and III carry the nodZ gene, whereas the strains of clade I do not (Fig. 4A).
ML trees of nodZ (A), nolL (B), and noeI (C) sequences. New strains from this study are indicated in boldface type. Scale bars indicate the number of substitutions per site. Bootstrap values (percentage from 100 replications) are indicated at nodes when higher than 50% (using ML). Strains or groups of strains belonging to clades defined in the nodA phylogeny (Fig. 3) are indicated on nodes. Abbreviations: L., Lupinus; O., Ornithopus; A., Aeschynomene; Ast., Astragalus; R., Rhizobium; M., Mesosrhizobium; C., Crotalaria.
Further analysis showed that all of the lupin and serradella bradyrhizobia except WSM1420 (for which no PCR product was obtained) carried the nolL gene, which is involved in acetylation of Nod factor fucose residues (2). In lupin strain WM9, a nolL homologue is located 5 kb downstream of the nodW gene (38). Considering that in soybean strain USDA 110, the common nodABC genes are separated from the nodVW operon by 300 kb (17), nolL may be similarly distant in lupin strains and may represent a distinct part of the symbiosis island. Previously, nolL was detected in all strains belonging to nodA clade II and in two strains of clade III but not in clade I (26). In the present work, we found it difficult to amplify this gene even in lupin and serradella bradyrhizobia, although we used more specific primers. Thus, we decided to exclude the strains of the remaining branches (I, III, and IV) from PCR amplification assays. The nolL sequences obtained were more divergent than those of the nodA and nodZ genes, with pairwise comparisons showing sequence similarities ranging from 87% to 100%. Nevertheless, the nolL phylogenetic tree displayed similar topology to that of trees derived from the nodA and nodZ genes (Fig. 4B).
The noeI gene is involved in the methylation of the Nod factor fucose residue, a widespread modification in broad-host-range rhizobia (14). The noeI in the genome of B. japonicum strain USDA 110 is located nearly 40 kb downstream of the common nod gene operon (17). Although this gene is widespread in clade III bradyrhizobia, it appears to be missing in clades I and II (26). Our present study confirmed the earlier data. The noeI gene was found in a strain belonging to clade III (WSM1743) and in strains WSM1735 and WSM1790 of the new clade IV. It was not, however, detected in the lupin and serradella strains nor in the clade I strains. Phylogenetic analysis of the noeI sequences confirmed the distant position of clade IV strain WSM1735 with respect to the strains of clade III. Although this corroborated the nodA-derived phylogeny, the position of WSM1735 on the noeI tree remained unresolved (Fig. 4C).
DISCUSSION
Application of legume inoculants may lead to the establishment of inoculated strains of rhizobia in soils, especially when a new crop is introduced. Well-established soil populations often, however, contain rhizobia belonging to serologic and genomic types different from those of applied inoculum strains. For example, appreciable proportions of Bradyrhizobium populations infecting lupins and serradella in Western Australia and South Africa differ from the inoculum strains WU425 and WSM471 or VK7 used in the respective countries (4, 24). This genetic heterogeneity among naturalized rhizobia could have resulted from three different processes: (i) accidental introduction on soil-contaminated seeds (32) or through inoculation using soil from fields of legumes established elsewhere (40), (ii) acquisition by native strains of symbiosis genes from introduced inoculum strains by lateral gene transfer (41), or (iii) the presence of indigenous rhizobia capable of infecting the new legume species (20).
Examination of the phylogenetic relationships of nodulation and housekeeping genes showed that serradella and lupin bradyrhizobia in Western Australia and South Africa are closely related to strains originating from southern Europe. Together with the related finding that Rhizobium etli strains are carried by bean seeds (32), this evidence favors the first possibility, that they were introduced on imported seeds from Europe. Our data show remarkable consistency in phylogenetic placement of the majority of lupin and serradella strains within the respective gene trees, in all of which they form discrete branches (Fig. 1 and 4). This corresponds well with the present evidence that lupin and serradella bradyrhizobia originating from Europe form separate clades on nodA and nodC trees (15, 26, 38, 49, 50). Their unique position on symbiotic gene trees allows lupin and serradella bradyrhizobia of mainly European origin to be distinguished from bradyrhizobia nodulating legumes other than Genisteae and Loteae. The nodA sequences within clade II show ≥90% identity, which is much higher, for instance, than identity within divergent clade III (Fig. 3 and 4). Excluding strain WU140, which shows 90% sequence identity, the remaining nodA sequences of the lupin and serradella strains from Western Australia and South Africa showed 97 to 100% identity and were practically indistinguishable from the two strains from Greece included in this study. This additionally contradicts the hypothesis of an indigenous origin of nod gene sequences in the lupin and serradella symbionts. In Western Australia, for example, one would expect higher divergence between European strains nodulating lupins and serradella and native Australian strains infecting these legumes. Likewise, the general similarity of nod gene sequences of the South African lupin isolates to the lupin and serradella strains from Western Australia and Greece and their dissimilarity with respect to other sequenced African isolates (26) makes it unlikely that their sequences are of native origin.
There is evidence that lateral gene transfer is responsible for dispersal of symbiosis genes among diverse bacteria of the α- and β-Proteobacteria groups (25, 37, 42, 43). The second possibility of transfer to indigenous strains is extremely unlikely, however, considering the observed relatedness between housekeeping genes of the Australian and South African isolates and the European lupin and serradella strains (discussed in the next paragraph). Lateral gene transfer of the symbiosis loci provides, however, an acceptable explanation for the presence of similar nodA sequences of clade II in different Bradyrhizobium lineages (26). In the present study, for example, we found that the nodA clade II sequences are carried by strains of B. canariense sp. nov. and two Bradyrhizobium lineages (as shown by the housekeeping gene phylogenies), one of which (4S16) may be related to B. liaoningense, while the other strain (1S20) may form a new clade in this genus.
Rhizobia from many native legumes in Western Australia effectively nodulated several lupin species (L. albus, L. cosentinii, and L. pilosus) but were unable to infect L. angustifolius, O. compressus, or O. sativus (11, 20, 21, 22). The present study does not, however, support the third possibility that the bradyrhizobia isolated from L. cosentinii have an Australian origin. First, housekeeping gene phylogenies reveal that the blue lupin (L. cosentinii) isolates belong to the same Bradyrhizobium canariense clade as strains of serradella from Western Australia and three of five South African narrow-leaf lupin strains (Fig. 1 and 2). Second, the nodulation genes of the blue lupin isolates group within the same clade II as the strains isolated from serradella species and narrow-leaf lupin (Fig. 3 and 4). Furthermore, most of the strains isolated from Australian native legumes showed no phylogenetic affinity of their nodulation genes to the lupin and serradella isolates (with the exception of strains WSM2155 and WSM2156). Instead, they group in the distinct nod gene clusters I, III, and IV. It appears possible that the nodA gene clade I may be of Australian origin despite the fact that the first two clade I nodA sequences originated from strains isolated in South Africa and Brazil. This clade comprises 98 to 99% similar nodA sequences, whereas the housekeeping gene phylogenies indicate various genomic types. Moreover, the New South Wales strains were collected in different places from the nodules of native legumes representing various legume tribes, including two, Bossiaeeae and Mirbelieae, that are endemic to Australia (Table 1). Likewise, clade IV may represent another set of strains that coevolved with native Australian legumes. This implies that geographical separation preventing strain migration and recombination of symbiotic loci may be an essential factor responsible for the observed divergence patterns of nodulation genes. This would agree with recent observations that nifD sequences group according to a strain's geographical origin, unlike 16S rDNA sequences, for which such a correlation has not been found (30). It remains to be established whether the phylogenetic pattern shown by the nodA tree is due to a vicariance related to tectonic isolation of Australia some 32 million years ago or climatic changes contributing to rapid radiation of Australian sclerophyllous flora (including legumes) in the mid-Cenozoic period between 25 and 10 million years ago (7).
Although phylogenetic evidence in this study does not indicate that our lupin and serradella strains are of indigenous origin, two isolates from native Western Australian legumes (strains WSM2155 and WSM2156) were capable of nodulating serradella (Table 1). Both shared a high level of dnaK, ITS, recA, nodA, nodZ, and nolL sequence homology with the lupin, serradella, and other B. canariense strains and probably represent lupin and serradella bradyrhizobia. As the respective legumes from which these strains were isolated (Daviesia nudiflora and Mirbelia floribunda) were located within 20 m of serradella trials and a lupin crop, we believe that Daviesia nudiflora and Mirbelia floribunda are promiscuous legumes that can be nodulated by lupin and serradella bradyrhizobia as well as indigenous rhizobia. This would accord with the observation that five native legumes (seed harvested from the Carrabin site) were nodulated by a cocktail of lupin and serradella strains originating from the Carrabin and Albany sites in Western Australia (24). The presence of alternate hosts may have facilitated the spread of lupin and serradella rhizobia throughout Western Australia after introduction from Europe.
Previous RAPD-PCR studies detected considerable genomic diversity among the lupin and serradella bradyrhizobia populations found in Australia and South Africa (4, 24). Although gene sequences, especially nod sequences of the lupin and serradella strains, were shown to be closely related and sometimes identical, well-defined distinct groups were detected. This diversification is unlikely to have occurred in the relatively short period of about a century following their introduction into soils of Western Australia and South Africa, since the divergence rate of conserved housekeeping genes is too low to explain the observed sequence diversity (46). It rather appears that these soils were initially infected and colonized by a variety of related strain types. Interestingly, all of the lupin and serradella strains from Western Australia and three of five South African strains group with Bradyrhizobium canariense sp. nov. B. canariense is a newly described species that originates from the Mediterranean region and is well adapted to acid soils common in Western Australia and South Africa (49, 50). Soil conditions often determine strain dominance. For example, soybean serocluster 135 (strains classified as B. liaoningense) is dominant in soils of high pH in the United States and South Africa (3, 47). It is noteworthy, therefore, that lupin strain 4S16 from the alkaline Roodebloem soil grouped with B. liaoningense in the recA and atpD trees (Fig. 2). Presumably, soil environment influences the survival of certain phylogenetic types. Dominance of the B. canariense strains in soil may also be related to compatibility with lupin and serradella, as reflected by the position of the nod genes of lupin and serradella strains on the phylogenetic trees presumably resulting from an evolutionary process interwoven with that of their natural hosts.
In conclusion, this study provided evidence that accidental cotransfer of effective lupin Bradyrhizobium strains, possibly by means of soil-contaminated seeds, accompanied the introduction of lupin and serradella from Europe to soils of Western Australia and South Africa. As in the transfer of seed-borne diseases, precautions against similar occurrences should be taken when introducing new legume crops to a region. Efforts to optimize nitrogen fixation through the inoculation of select rhizobial strains may be fruitless if less-effective and highly competitive strains are cointroduced.
ACKNOWLEDGMENTS
We thank Andrzej B. Legocki and Wanda Małek for their help in this project. We are grateful to Jeremy Burdon and Noelle Amarger for providing Bradyrhizobium strains. We also thank Dorota Gurda, Anna Rucka, Agnieszka Chlebicka, and Łukasz Markiewicz for their help during the preparation of the manuscript.
This work was financed by grant 2P04C 063 26 (to T.S.) from the Polish State Committee for Scientific Research (KBN).
FOOTNOTES
- Received 21 April 2005.
- Accepted 19 July 2005.
- Copyright © 2005 American Society for Microbiology
