INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Due to their nitrogen-fixing symbiosis with soil bacteria collectively named rhizobia, legumes play an important role in the nitrogen cycle, especially in the tropics. They are used to restore or increase soil fertility of degraded soils in intercropping systems, as green manure, and to produce medicinal and commercial by-products. In Senegal (West Africa), many native legumes are important plant resources for many purposes, and they are the only alternatives to the costly and pollutive mineral nitrogen fertilizers used in agriculture. Many of these legumes are well adapted to the local arid climatic conditions. As drought and soil erosion progress, they are still present as colonizers. The family Leguminosae is considered to be of tropical or subtropical origin, and many of the recently proposed new rhizobial species originate from leguminous plants from these zones that had not been previously investigated.

After several prospections around the country, we focused on 27 native nodulated leguminous plants species belonging to the genera Abrus, Alysicarpus, Bryaspis, Chamaecrista, Cassia, Crotalaria, Desmodium, Eriosema, Indigofera, Moghania, Rhynchosia, Sesbania, Tephrosia, Zornia, which play an important ecological role and have agronomic potential in arid regions. As very little or no information concerning their associated rhizobia were available so far, we obtained 71 nodule isolates from these legumes in different regions in Senegal and first characterized them by sodium dodecyl sulfate-polyacrylamide gel electrophoresis whole-cell protein profiles and by 16S amplified ribosomal DNA restriction analysis (ARDRA) fingerprint analysis (9). Our main conclusion was that the strains were diverse and belonged to five phylogenetic subgroups inside Bradyrhizobium, but further genotypic characterization of these strains was needed to precisely place them in the general classification.

Taxonomically, rhizobia comprise six genera, Rhizobium, Bradyrhizobium, Mesorhizobium, Sinorhizobium, Azorhizobium (for a review, see reference 44), and Allorhizobium (8). They constitute a phylogenetically heterogeneous group, and their taxonomy is being reexamined. Some rhizobia are more closely related to clinical bacteria, like Afipia, Blastobacter, and other nonsymbiotic bacteria, like Mycoplana and Bartonella, than to other rhizobia (39, 43, 44). In addition, several new rhizobial groups have been described, but have not yet been classified or named (for reviews, see references 24 and 44). In particular, the precise taxonomical status of many Bradyrhizobium sp. strains isolated from various legumes is not clarified (9, 10, 25, 26, 27, 34). For this genus (Bradyrhizobium), several authors have reported the lack of consistency between results obtained by different taxonomic techniques (10, 20, 32, 45).

A need for a confident strategy to investigate diversity among Bradyrhizobium populations was claimed, specifically including molecular methods (9, 34). 16S-23S rRNA intergenic gene spacer (IGS; corresponding to the spacer between 16S and 23S rRNA genes) sequences exhibit a large variability and are useful to identify genomic groups at the intraspecific level (4, 16, 22). Moreover, PCR-restriction fragment length polymorphism (PCR-RFLP) of IGS has been reported to be a useful fingerprinting method to characterize bacterial strains, with a higher discriminating power than the 16S ARDRA method. It has been applied to known rhizobial species, such as Rhizobium leguminosarum, Rhizobium "hedysari," Sinorhizobium meliloti, and Rhizobium galegae (6, 22, 31, 33), and also tropical rhizobia strains (18) and Bradyrhizobium strains from the Canary Islands (37).

The amplification fragment length polymorphism (AFLP) technique (38, 46) is a highly discriminating fingerprinting method, based on the selective PCR amplification of certain restriction fragments from a digest of total genomic DNA. The technique involves three steps: (i) restriction of the DNA with two enzymes and ligation of oligonucleotide adapters, (ii) selective amplification of sets of restriction fragments, and (iii) polyacrylamide gel electrophoresis of the amplified fragments. Originally developed for plant genome studies, this technique has also been used to characterize various bacterial species (for a review, see reference 5). As this technique investigates the whole genome, and as the results were reported to be in good agreement with those obtained by DNA:DNA hybridization, it could be a good alternative to the latter, which is especially difficult to perform with Bradyrhizobium.

Here, 64 Bradyrhizobium sp. strains from small Senegalese legumes were further characterized genotypically by PCR-RFLP analysis of the IGS region between 16S and 23S rRNA genes and by AFLP analysis, including representative strains for Bradyrhizobium species and groups reported in the past (9, 10, 25, 26, 40). The results are compared with previous 16S ARDRA grouping, and taxonomic resolution levels of the three techniques are discussed.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and isolation procedures. The strains used in this study are listed in Table 1, and their places of isolation are indicated on Fig. 1. Bacterial strains were grown as previously described (9, 40).

View larger version (30K): [in this window] [in a new window]   FIG. 1.   Sites of sampling in Senegal.

PCR amplification of IGS (16S-23S rDNA) region. Total DNA was prepared as previously described (9). Alternatively, for some strains, the PCR amplification was done from bacterial cell preparation, prepared as follows: cells were grown on yeast extract-mannitol agar slants (36) for 72 h at 28°C and then washed with sterile distilled water. The cell suspension was adjusted to an optical density (620 nm) of 0.5 by dilution in water. An aliquot of 100 µl of this cell suspension was pelleted, and cells were resuspended in 100 µl of sterile Milli-Q water (Milli-Q plus; Millipore, Saint-Quentin-Yvelines, France) and lysed with 100 µl of 10 mM Tris-HCl (pH 8.2) and 13 µl of proteinase K (1 mg/ml) (Merck-Belgolabo, Overijse, Belgium) during 2 h at 55°C. Then, the cells were boiled for 10 min to denature the enzyme. The PCR was carried out with 250 ng of DNA or with 10 µl of bacterial cell suspension as template DNA.

RFLP analysis. Aliquots (10 µl) of PCR products were digested with 5 U of restriction endonuclease in 20-µl reaction volumes by using the manufacturer's recommended buffer and incubation conditions. The following restriction enzymes were used: AluI and NdeII (Boehringer Mannheim Biochemica, Brussels, Belgium), DdeI (Amersham Pharmacia Biotech Benelux, Roosendaal, The Netherlands), and HaeIII, HhaI, HinfI, MspI, and RsaI (New England Biolabs Inc., Leusden, The Netherlands). Restricted DNA was analyzed by horizontal electrophoresis in 4% Nusieve 3:1 or Metaphor agarose (FMC, Rockland, Maine) gels. DNA molecular-weight-marker VIII (Boehringer Mannheim) was used as standard for gel calibration. Electrophoreses were carried out at 80 V for 4 h with standard gels (11 by 14 cm or 10 by 15 cm) on a Bethesda Research Laboratories Horizon 11-14 apparatus or with a DNA Sub Cell unit (15-by-10-cm tray) (20 wells; Bio-Rad Laboratories NV, Nazareth-Eke, Belgium).

AFLP analysis. The experimental protocol used was modified from that of Vos et al. (38) as described by Huys et al. (14) and was essentially similar to that used in the study of Willems et al. (40). In short, 1 µg of total genomic DNA was digested by two restriction enzymes. Double-stranded oligonucleotide adapters with a single-stranded overhang homologous to the 5' and 3' ends generated during restriction were ligated to the DNA fragments. The ligated DNA fragments were amplified by PCR using primers complementary to the adapter and restriction site sequence with additional selective nucleotides at their 3' end. The amplified products were separated by polyacrylamide gel electrophoresis, and the resulting banding pattern was revealed through autoradiography. Bordetella holmesii strain LMG 15945 was used as the reference strain because AFLP analysis of this strain generated a banding pattern displaying evenly distributed and well-separated bands over the entire length of the gel.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

IGS PCR-RFLP analysis. (i) Cluster analysis of IGS RFLP patterns. The 16S-23S rDNA IGS region of 104 Bradyrhizobium strains was amplified by PCR and restricted with eight endonucleases. One to six DNA fragments were generated by each restriction enzyme. For each strain, the patterns obtained with the eight enzymes were combined, resulting in 70 different combinations referred to as IGS rDNA types (Table 2).

View larger version (40K): [in this window] [in a new window]   FIG. 2.   Dendrogram (UPGMA) obtained by numerical analysis comparison of the normalized and combined IGS patterns by PCR-RFLP analysis. Clusters of strains were delineated above 70% similarity.

(ii) Size of IGS PCR products. We estimated the sizes of IGS PCR products by summing the sizes of the restriction fragments. All strains produced a single PCR product ranging from 780 to 1,038 bp, depending on the strain (Table 2). The differences observed in the size of the PCR products could be in part explained by the presence of several tRNA genes varying in number and type in the IGS regions (16, 22, 23).

AFLP analysis. (i) Optimization of the AFLP technique for Bradyrhizobium strains. Several enzyme combinations were tested to determine the most suitable one, i.e., one producing a large number of fragments (30 to 50) of many different lengths resulting in an evenly distributed banding pattern. For Bradyrhizobium organisms which have a high GC percentage in their genomes, the combination of TaqI (T/CGA) and ApaI (GGGCC/C) proved the most useful.

(ii) Numerical analysis of Bradyrhizobium AFLP patterns. In our AFLP study, we included 63 strains from small legumes, 14 reference strains representative for B. japonicum, B. elkanii, and Bradyrhizobium liaoningense, 18 Bradyrhizobium sp. (F. albida) and 10 Bradyrhizobium sp. (Aeschynomene) strains representative for AFLP groups 6, 16, 17, 19, 20, 28, 29, 31, 33, and 34 described elsewhere (40), and 13 other Bradyrhizobium sp. strains from various host plants (10, 21, 26, 27, 34).

View larger version (25K): [in this window] [in a new window]   FIG. 3.   Dendrogram obtained by UPGMA analysis of AFLP patterns of Senegalese isolates, some representative strains from Brazil, from F. albida, and from Aeschynomene, and the reference strains of Bradyrhizobium. The clusters were delineated at above 50% similarity. The clusters 6, 12, 15, 16, 17, 19, 20, 28, 29, 31, 32, 33, and 34 correspond to previously described AFLP clusters (42). The new groups identified in this study (35 to 48) are in bold.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

For Bradyrhizobium, many groups have been identified during recent years, but their taxonomic status has remained unclear due to the inconsistency of the results obtained by different taxonomic methods. There is a need to use several methods, preferably genotypic ones, to draw more reliable taxonomic conclusions.

In a previous report, we isolated and performed initial characterization of 71 nodule isolates from small legumes in Senegal by sodium dodecyl sulfate-polyacrylamide gel electrophoresis of total proteins and by 16S ARDRA and compared them to a number of Bradyrhizobium reference strains. We concluded that these Bradyrhizobium strains were diverse and formed seven phylogenetic groups. Here, we continued the study of these strains by two genomic techniques, PCR-RFLP analysis of the IGS region between 16S and 23S rRNA genes and AFLP analysis. We focused on the applicability and comparison of the taxonomic resolution level of these two techniques to characterize Bradyrhizobium strains and help elucidate taxonomic problems encountered in this genus. Both IGS PCR-RFLP and AFLP analyses confirmed that Bradyrhizobium strains from small legumes are diverse.

By IGS PCR-RFLP using eight restriction enzymes, the 104 strains studied produced 70 types of combined restriction profiles, forming 16 groups. By AFLP analysis, the strains formed 27 groups. Table 3 shows comparative results obtained with essentially the same strains by the two techniques IGS PCR-RFLP and AFLP analyses and also by previous 16S ARDRA (9). There is very good agreement between results from the three techniques, except for the two Bradyrhizobium sp. (F. albida) strains ORS 187 and ORS 110. The 16S ARDRA and AFLP results for ORS 187 are consistent, but not with the IGS results; the 16S ARDRA and IGS results for ORS 110 do not match. In most cases, each AFLP cluster contained strains showing the same or very similar rDNA IGS types and belonging to one IGS PCR-RFLP cluster.

The strains belonging to the 16S ARDRA groups A, B, E, and F (B. japonicum lineage) belonged to 43 IGS PCR-RFLP types (corresponding to 12 main IGS PCR-RFLP clusters) and to 17 AFLP clusters. Interestingly, strains of 16S ARDRA group A, exclusively consisting of photosynthetic strains, were also separate from nonphotosynthetic strains in AFLP and IGS PCR-RFLP groups. The strains belonging to 16S ARDRA groups C and G (B. elkanii lineage) belonged to 20 IGS PCR-RFLP types (corresponding to six IGS PCR-RFLP main clusters) and 10 AFLP clusters.

This confirms, as expected from literature dealing with other bacterial groups, that IGS PCR-RFLP (4, 22) and AFLP (15) analyses are more discriminative than 16S ARDRA for Bradyrhizobium and do almost differentiate at the strain level. Moreover, it has been reported that the AFLP method was more efficient for assessing intrapathovar diversity of the genus Pseudomonas than the rapid amplified polymorphic DNA method (7).

For screening purposes, fine discriminating genotypic techniques such as IGS PCR-RFLP and AFLP are recommended; IGS PCR-RFLP has some advantages because the experimental protocol is sure, simple, and similar to that used for other studies on the 16S rDNA gene and is less laborious than the AFLP technique. Our results agreed with the study of Leblond-Bourget et al. (23) on Bifidobacterium species, showing that IGS rDNA regions are useful for rapid identification or intraspecific phylogenetic studies, while 16S rRNA sequences are a good tool to infer interspecific links (11).

As expected from the literature (3, 9, 13, 19, 30, 40, 41, 45), our results show some heterogeneity among B. japonicum reference strains. Four of the B. japonicum reference strains included, LMG 4262, LMG 4265, LMG 4272, and LMG 4274, were of different plant origins, Albizzia, Ulex, Pueroria, and Vigna, respectively. They all occupied separate positions by both IGS PCR-RFLP and AFLP analyses. On the contrary, a majority of the B. japonicum reference strains originating from Glycine max clustered in two main groups, IGS II/AFLP 15 and IGS VII/AFLP 12. One exception is B. japonicum strain LMG 8321 grouping together with isolates from small legumes and from F. albida in IGS V/AFLP 20. By the two techniques, and also by 16S ARDRA (9), B. japonicum strain USDA 110 grouped separately from the B. japonicum type strain LMG 6138. This result corroborated other studies based on DNA:DNA hybridizations (13) and 16S rDNA analysis (3, 37).

By the three techniques, the B. elkanii strains grouped together; moreover, by AFLP analysis, they formed a separate group.

By ARDRA, both B. liaoningense strains were found in the same group as B. japonicum type strain LMG 6138; here, they were distinct from B. japonicum and formed a separate group by both IGS PCR-RFLP and AFLP analyses.

Of the eight restriction enzymes used for IGS PCR-RFLP in our study, HinfI and RsaI were the least discriminative. Except for three strains, LMG 15268, LMG 8888, and LMG 4274, all the strains could be differentiated by using the combination of the IGS PCR-RFLP patterns obtained with two enzymes, NdeII and MspI.

No clear relationship could be evidenced between plant origin, host range (9), and the grouping of strains by AFLP and IGS PCR-RFLP analyses. On the contrary, we found a certain degree of relationship between AFLP and IGS PCR-RFLP groups and the geographical origins of the strains, especially between north and south Senegal, corresponding to different ecological regions. Senegal is a constrasting country, from its northern part being in the sahelian zone (dry steppe receiving 100 to 300 mm rainfall per year) to the southern part being in the guineo-soudanian zone (deciduous forest receiving 800 to 1,200 mm of rainfall per year). In general, groups mainly consisted of isolates either from north Senegal (IGS groups V, VI, XIII, and XIV corresponding to AFLP groups 20, 29, 33, 34, 40, and 46) or south or central Senegal (IGS groups I, III, IV, X, XI, and XII corresponding to AFLP groups 19, 31, 35, 36, 37, 38, 42, 43, 44, 45, and 47). However, this is not a general rule since some exceptions exist, like IGS group VIII and AFLP group 33 consisting of strains originating from everywhere in Senegal.

Here, we extended the screening of our Bradyrhizobium strain collection by AFLP analysis. The AFLP technique has been described as having a similar discriminating power as DNA:DNA hybridizations, making it a potentially useful taxonomic tool for species delineation for a number of different bacterial groups (7, 14, 15). This result is not unexpected since this method studies the whole bacterial genome, resulting in a profile with many bands (15). DNA:DNA hybridizations remain required to fully evaluate the discriminative power of the technique in the case of Bradyrhizobium.