Phylogenetic analysis of the genera Bradyrhizobium, Mesorhizobium, Rhizobium and Sinorhizobium on the basis of 16S rRNA gene and internally transcribed spacer region sequences

Some genera within the α-Proteobacteria are associated with plants, inducing the formation of nitrogen-fixing nodules on the roots and stems of leguminous plants or causing gall disease and root hair disease. According to Bergey's Manual of Systematic Bacteriology, the rhizobia and agrobacteria were classified into four genera (Rhizobium, Bradyrhizobium, Agrobacterium and Phyllobacterium) within the family Rhizobiaceae (Jordan, 1984). Since then, rhizobial and agrobacterial taxa have been added to and revised at the genus and species levels. The genus Sinorhizobium was proposed for a group of fast-growing, nodule-forming, soybean rhizobia, Rhizobium fredii (Chen et al., 1988), and this was later confirmed by de Lajudie et al. (1994). Jarvis et al. (1997) created a new genus, Mesorhizobium, for Rhizobium loti, Rhizobium huakuii, Rhizobium ciceri, Rhizobium mediterraneum and Rhizobium tianshanense. Allorhizobium undicola was the name given to nitrogen-fixing strains that nodulated Neptunia natans in Senegal (de Lajudie et al., 1998). Also, agrobacteria were classified as Agrobacterium tumefaciens, Agrobacterium radiobacter, Agrobacterium rhizogenes, Agrobacterium rubi and Agrobacterium vitis on the basis of plasmid-borne phytopathological traits. Most recently, species belonging to the genera Allorhizobium and Agrobacterium have been reclassified in the genus Rhizobium (Young et al., 2001). An overview of all members of these genera is available ().

Rhizobium, Bradyrhizobium, Mesorhizobium, Sinorhizobium and the former Agrobacterium were traditionally classified on the basis of phenotypic characteristics such as nodulation, and pathogenic and physiological properties. However, nodulation and pathogenic properties have become less important in the taxonomic evaluation of these genera. Recently, high sequence variation in the internally transcribed spacer (ITS) region has been shown to be more informative for taxonomic evaluation of Bradyrhizobium strains (van Berkum & Fuhrmann, 2000; Willems et al., 2001b).

In this study, rhizobial strains were isolated from root nodules of several legume hosts in Korea. Their ITS regions and 16S rRNA genes were sequenced and compared with those of reference strains. Phylogenetic analysis of these sequences was performed to compare the dendrograms generated from these two types of sequences and to evaluate the use of the ITS region as a taxonomic marker.

Bacterial strains and media.
The bacterial strains used in this study are listed in Table 1. All strains were grown on yeast extract-mannitol (YM) agar (Vincent, 1970) or nutrient broth (0·5 % peptone, 0·5 % meat extract, pH 7·0) at 30 °C. Strains were preserved using two methods: deep-freezing with 15 % glycerol and freeze-drying with 15 % skimmed milk.

Table 1. Bacterial strains isolated from legume hosts in Korea For each strain, the generic position was determined on the basis of 16S rRNA gene and ITS sequence analysis.

PCR amplification of the 16S rRNA gene and ITS region.
The 16S rRNA gene was amplified using universal primers fD1 and rP2 (Weisburg et al., 1991). The ITS region was amplified using primers R16-1 (5'-CTTGTACACACCGCCCGTCA-3') and R23-3R (5'-GGTACTTAGATGTTTCAGTTC-3'), which were redesigned from the primers of Nakagawa et al. (1994). Primers R16-1 and R23-3R were located at positions 13891408 of the 16S rRNA gene and at 189209 of the 23S rRNA gene sequence of Escherichia coli, respectively (GenBank accession no. J01695) (Brosius et al., 1978). Each PCR mixture contained the following: 20 pmol each primer, 200 µM dNTPs (Promega), 2·5 U Taq DNA polymerase (Promega), approximately 50 ng genomic DNA and Taq polymerase buffer in a final reaction volume of 50 µl. The DNA thermal cycler (Perkin-Elmer) used for PCR amplification was programmed as follows: an initial extensive denaturation step at 94 °C for 5 min; 35 cycles of 94 °C for 1 min, 58 °C for 1 min and 72 °C for 2 min; and a final extension step at 72 °C for 10 min.

Cloning and sequencing.
PCR products were electrophoresed on 1·5 % agarose gel. In the ITS region, one or two bands from each of the strains were observed. Each band was eluted and purified with a QIAquick gel extraction kit (Qiagene). Purified DNAs were ligated into pGEM-T easy vector (Promega) and ligated plasmids were then used to transform Escherichia coli DH5αF' cells (Sambrook et al., 1989). Nucleotide sequences were determined with an Applied Biosystems 377 sequencer. Sequencing of the ITS region was conducted with three primers, R16-1, R23-3R and ILEF (5'-GTAGCTCAGTTGGTTAGAGC-3'). ILEF was designed from a conserved region (tRNA^Ile coding region) within the ITS region.

Phylogenetic analysis.
Nucleotide sequences were aligned using the program MEGALIGN (DNASTAR). An evolutionary distance matrix was generated as described by Jukes & Cantor (1969). An evolutionary tree for the datasets was inferred by the neighbour-joining method of Saitou & Nei (1987) using the neighbour-joining program of MEGA2 (). The stability of relationships was assessed by performing bootstrap analysis of the neighbour-joining data based on 1000 resamplings.

16S rRNA gene and ITS region sequence determination
A total of 128 bacterial strains was isolated from more than 23 legume hosts in Korea (Table 1). Phylogenetic analysis was carried out using 16S rRNA gene sequences of about 1450 nt from all strains used in this study, corresponding to nt 501501 of the Escherichia coli 16S rRNA gene sequence (Brosius et al., 1978). Most of the strains produced one band in PCR amplification of the ITS region. However, six strains (Mesorhizobium mediterraneum USDA 3392^T, Rhizobium gallicum R602sp^T, Rhizobium rubi DSM 6772^T, Sinorhizobium kostiense LMG 19227^T, Leb-2 and Glm-15) produced two bands. The sequence similarities between two clones of the same strain ranged from 77 % (two clones of S. kostiense LMG 19227^T) to 98 % (two clones of R. gallicum R602sp^T) and all these sequences were included in the phylogenetic analysis. Phylogenetic analysis of these ITS clones revealed that each pair of ITS clones for these strains, except for Glm-15, whose two ITS clones were positioned in separate clades, were closely related (Figs 1 and 3).

(37K): Fig. 1. Phylogenetic trees based on the 16S rRNA gene (a) and ITS region sequences (b) for the genus Bradyrhizobium and related strains. The branching pattern was produced by the neighbour-joining method. Numbers at nodes indicate levels of bootstrap support based on neighbour-joining analysis of 1000 resampled datasets. Bootstrap values below 40 % are not shown. Genospecies are noted according to Willems et al. (2001b).

(39K): Fig. 3. Phylogenetic trees based on ITS region sequences for the genus Rhizobium and related strains (a) and the genus Mesorhizobium and related strains (b). For further details see the legend to Fig. 2.

Although, in general, the high sequence variation of the ITS region allows discrimination between closely related strains, the presence of multiple types of ITS region makes interpretation of phylogenetic relationships difficult (Chun et al., 1999; Gürtler & Stanisich, 1996). This study showed that most rhizobial strains contained one type of ITS region sequence, suggesting that this region may be useful as a marker for phylogenetic analysis. For phylogenetic analysis of ITS region sequences, the full-length ITS region with about 10 nt of the 3'-terminal region of the 16S rRNA gene and about 40 nt of the 5'-terminal region of the 23S rRNA gene was used. All ITS region sequences analysed in this study varied from 751 to 1236 nt and contained two deduced tRNA genes, tRNA^Ile and tRNA^Ala. The ILEF primer for internal sequencing of the ITS regions was designed from the conserved tRNA^Ile region. Variable sequences within the ITS region were interspersed throughout the entire region, except for the two tRNA sequences. The ITS regions for all strains revealed high sequence variation (24100 %). To simplify the tree topology, strains with high sequence similarities, i.e. >99·8 % (16S rRNA gene) and >99·7 % (ITS region), were grouped into boxes at the same position in our phylogenetic trees (Figs 1, 2 and 3).

(39K): Fig. 2. Phylogenetic tree based on the 16S rRNA gene for the genera Mesorhizobium, Rhizobium and Sinorhizobium and related strains. The branching pattern was produced by the neighbour-joining method. Numbers at nodes indicate levels of bootstrap support based on neighbour-joining analysis of 1000 resampled datasets. Bootstrap values below 40 % are not shown.

Phylogenetic analysis of the Bradyrhizobium group
For 16S rRNA gene sequence analysis of the Bradyrhizobium group, a total of 73 strains, including four type strains, 60 Korean isolates and nine strains of van Berkum & Fuhrmann (2000), was used. The ITS region sequence analysis was conducted with a total of 79 strains, including three type strains, 60 Korean isolates, nine strains of van Berkum & Fuhrmann (2000) and seven strains of Willems et al. (2001b). The 16S rRNA gene sequence similarities among the strains were more than 93·3 % and the ITS region sequence similarities were more than 44·4 %. Both trees produced two main groups: group I, including Bradyrhizobium elkanii strains, and group II, including Bradyrhizobium japonicum and Bradyrhizobium liaoningense strains (Fig. 1a, b). According to the 16S rRNA tree, group I was divided into two subgroups (groups Ia and Ib) (Fig. 1a). Group Ia consisted of B. elkanii strains, including B. elkanii USDA 76^T, and 11 Korean strains. Group Ib included only Korean strains (39 strains) isolated from different legume hosts. The phylogenetic positions of ten Korean isolates and strains of B. japonicum, B. liaoningense and Bradyrhizobium yuanmingense, which were clustered into group II, were not clear due to low bootstrap values (Fig. 1a). Based on phylogenetic analysis of the ITS region sequences, three groups with high bootstrap values were identified (Fig. 1b). Within group I, group Ia included 41 strains, all of which, except for LMG 8293, were isolated from Korean legume hosts. Among them, three strains (LMG 8293, Mes-1 and Vga-1) clustered independently. Group Ib included four B. elkanii strains, LMG 10676 and nine Korean isolates. Within group II, group IIa was composed of four B. japonicum strains (including LMG 6138^T), B. liaoningense LMG 18230^T and four Korean isolates, and group IIb included four B. japonicum strains, LMG 11950, LMG 10697 and six Korean isolates. Group III was composed of two strains (LMG 15404-c4 and LMG 10300). The members of subclusters (groups Ia and Ib) within group I of the 16S rRNA tree were firmly reflected in those of subclusters (groups Ia and Ib) within group I of the ITS tree, except for a new branch of LMG 8293, Vga-1 and Mes-1 that was not recognized in the 16S rRNA tree (Fig. 1a, b). Two branches (groups IIa and IIb) of group II, which were supported by relatively high bootstrap values, were determined in the ITS tree; this branching pattern was not evident in the 16S rRNA tree.

van Berkum & Fuhrmann (2000) conducted a phylogenetic study using the ITS region sequences of Bradyrhizobium strains isolated from soybean. Willems et al. (2001b) studied extensively the genus Bradyrhizobium on the basis of 16S rRNA gene and ITS sequences, amplified fragment length polymorphisms and DNADNA hybridization. Using DNADNA hybridization data and ITS region sequence grouping of the Bradyrhizobium strains from several host plants, mainly Glycine max, Aeschynomene species and Faidherbia albida, seven genospecies were delineated (Willems et al., 2001b). A subsequent DNADNA hybridization study revealed four more genospecies (Willems et al., 2001c). Our ITS region sequence analysis, including representative strains of van Berkum & Fuhrmann (2000) and Willems et al. (2001b), yielded more than two new clusters discernable from the seven groups of Willems et al. (2001b) (Fig. 1b). One distinct cluster consisted of only Korean isolates from various host plants within group Ia. The other cluster was composed of three strains (LMG 8293, Vga-1 and Mes-1) within group Ia. Although, according to Willems et al. (2001b), LMG 8293 was supposed to be related to genospecies II, which was represented by B. elkanii LMG 6134^T, our results showed that this strain formed an independent cluster with Vga-1 and Mes-1. Furthermore, the ITS region sequences of two strains (LMG 15404-c4 and LMG 10300), which were classified as genospecies VI of Willems et al. (2001a), formed a distinct cluster separable from known species and the Korean isolates. The close relationship between the type strains of B. japonicum and B. liaoningense was confirmed by high sequence similarities (16S rRNA gene sequence similarity of 99·8 % and ITS region sequence similarity of 88·2 %) and a high DNADNA hybridization value (56 %) (Willems et al., 2001a, b).

Phylogenetic analysis of the Rhizobium group
For 16S rRNA gene sequence analysis, the type strains of 18 Rhizobium species and 45 Korean isolates were included. For the ITS region sequence analysis, 45 Korean isolates and 15 reference strains were used. Among the strains analysed, the 16S rRNA gene sequence similarities were shown to be above 89·4 % and the ITS region sequence similarities were more than 31·4 %. The 16S rRNA gene sequences of members of the genus Rhizobium (group I) formed one large cluster separable from the Mesorhizobium (group III) and Sinorhizobium (group II) groups (Fig. 2). Each of these main groups comprised subclusters that were supported by high bootstrap values. Within group I, four subgroups (groups IaId) were produced. Thirty-two Korean isolates formed one large cluster with several Rhizobium species, including Rhizobium leguminosarum USDA 2370^T and Rhizobium etli CFN 42^T. Gls-3 clustered in group Ib with Rhizobium galegae ATCC 43677^T, Rhizobium huautlense SO2^T and Rhizobium loessense CCBAU 7190B^T. Group Ic comprised two Korean isolates (Kusp-1 and Aei-5) and Rhizobium giardinii H152^T. Ten Korean isolates were grouped into group Id with Rhizobium radiobacter NCPPB 2437^T, Rhizobium rubi IFO 13261^T, Rhizobium vitis NCPPB 3554^T and Rhizobium undicola LMG 11875^T.

The dendrogram generated from the ITS region sequences of these Rhizobium strains and Korean isolates revealed tree topologies comparable to those of the 16S rRNA dendrogram and produced three groups (I, II and III), which were supported by high bootstrap values (Fig. 3a). Group I formed a large complex, including all the strains of groups Ia and Ib of the 16S rRNA tree and three strains (R. undicola LMG 11875^T, R. vitis LMG 8750^T and Kus-7) that were classified in group Id of the 16S rRNA tree. However, the phylogenetic relationships among the strains of group I on the basis of the ITS tree could not be clarified due to low bootstrap values. Group II of the ITS tree was composed of R. giardinii H152^T, Aei-5 and Kusp-1, which were found in group Ic of the 16S rRNA tree. The members of group III of the ITS tree were identical to those of group Id of the 16S rRNA tree, except for R. undicola LMG 11875^T, R. vitis LMG 8750^T and Kus-7 (Figs 2 and 3a).

The taxonomy of the genus Rhizobium remains controversial and, recently, Young et al. (2001) suggested that the former genus Agrobacterium be combined with Rhizobium based mainly on phylogenetic analysis of the 16S rRNA gene sequences without prejudice to phenotypic characters such as pathogenicity. In the phylogenetic trees based on 16S rRNA gene and ITS sequences, the Rhizobium species, including the Korean isolates, could be divided into several groups. Three major groups (I, II and III) in the ITS tree were supported by high bootstrap values and could be tentatively recognized as individual branches partially supported by the 16S rRNA tree (Fig. 2 and 3a). It could thus be suggested that the former Rhizobium and Agrobacterium be reclassified into more than two genera. However, such reclassification also might not be fully supported because of the presence of several peripheral groups and the discrepancy between the 16S rRNA and ITS trees.

Phylogenetic analysis of the Mesorhizobium and Sinorhizobium groups
For Mesorhizobium and related strains, the 16S rRNA tree was constructed for eight type strains and 22 related Korean strains and the ITS tree for 22 Korean isolates and seven type strains. Whereas the sequence similarities of the 16S rRNA gene were above 97·2 %, the ITS region sequence similarities were above 49·2 %. The 16S rRNA dendrogram grouped all Mesorhizobium and related strains into one cluster, group III (Fig. 2). Within group III, Wif-1 formed group IIIa with Mesorhizobium ciceri UPM-Ca7^T and Mesorhizobium loti LMG 6125^T. Twenty-one Korean isolates clustered with Mesorhizobium plurifarium LMG 11892^T, Mesorhizobium amorphae ACCC 19665^T and Mesorhizobium huakuii IFO 15243^T, forming group IIIb. The ITS tree of the Mesorhizobium and related strains resolved them into clearer branches than the 16S rRNA gene tree, but showed large disparities with the 16S rRNA tree (Fig. 3b). Due to the low ITS region sequence similarities (less than 80 %) to reference strains, most of the Korean isolates, except for Amf-5, Put-3 and Leb-4, could not be placed into any known species, suggesting the presence of several novel taxonomic units. Characteristically, the high degree of relatedness (92·8 %) of the ITS region sequences of Mesorhizobium ciceri LMG 14989^T and Mesorhizobium loti LMG 6125^T was consistent with the results of 16S rRNA gene sequence analysis (Tan et al., 1997) and high DNADNA hybridization (44 %) between these strains (Nour et al., 1994).

For phylogenetic analysis of Sinorhizobium and related strains, the 16S rRNA gene sequences for nine type strains and one Korean isolate were used. For the ITS region sequences, seven type strains, one reference strain and one Korean isolate were used. The similarities of the 16S rRNA gene and ITS region sequences were over 97·1 and 45·7 %, respectively. According to the 16S rRNA dendrogram, one Korean isolate, Mes-2, was closely related to Sinorhizobium meliloti LMG 6133^T (Fig. 2). Based on ITS region sequences, the Sinorhizobium strains were divided into two clear clades with relatively high bootstrap values. Mes-2 formed one clade with S. meliloti LMG 6133^T, Sinorhizobium medicae LMG 18864, Sinorhizobium fredii LMG 6217^T and Sinorhizobium xinjiangense LMG 17930^T. Another clade included Sinorhizobium arboris LMG 14919^T and two clones of S. kostiense LMG 19227^T (data not shown). The close similarity between S. fredii LMG 6217^T and S. xinjiangense LMG 17930^T, previously noted on the basis of 16S rRNA gene sequence similarity of 99·8 % (Tan et al., 1997), was also confirmed by the ITS region sequence similarity (94·9 %).

Phylogenetic analysis of the Korean legume isolates on the basis of 16S rRNA gene and ITS region sequences showed their remarkable heterogeneity. To clarify the exact taxonomic positions of these isolates, more taxonomic methods, including other genetic methods and DNADNA hybridization, will be needed.