Introduction

Soybean (Glycine max L.) is an important protein source used worldwide. In Brazil, the second largest world soybean producer, this crop occupies about 56% of the planted area, playing a significant economic and social role. A factor that decisively contributes to the success of soybean production in Brazil and its market competitiveness is the exploitation of biological nitrogen fixation (BNF) by the inoculation with strains of the Bradyrhizobium genus—SEMIA 5079 (B. japonicum), SEMIA 5080 (B. diazoefficiens), SEMIA 587 (B. elkanii), and SEMIA 5019 (B. elkanii).

Soybean has been a major host from which species of the Bradyrhizobium genus are isolated. Nine out of the 54 species of this genus have been isolated from soybean: B. japonicum USDA 6T, from Japan [1]; B. elkanii 76T, and B. diazoefficiens USDA 110T, from the United States [2, 3]; B. liaoningense LMG 18230T, B. yuanmingense CCBAU 10071T, B. huanghuaihaiense CCBAU23303T, and B. daqingense CCBAU 15774T from China [4,5,6,7]; B. ottawaense OO99T and B. amphicarpaeae 39S1MBT, from Canada [8, 9]. Besides, only 8 out of 23 Bradyrhizobium species isolated from other hosts and tested in soybean had their ability to nodulate this species confirmed, of which 2 formed efficient symbiosis: B. ferriligni CCBAU 51502T and B. shewense ERR11T [10, 11]. Despite the economic importance of soybean throughout Brazil, no new Bradyrhizobium species have been described based on native isolates nodulating this crop in Brazilian soils.

The use of integrated phenotypic, genotypic, and phylogenetic information is necessary to define the true taxonomic position of a strain or group of strains. In recent years, housekeeping genes (atpD, dnaK, glnII, gyrB, recA, and rpoB) as phylogenetic markers for bacteria have been widely used for the separation of genetically close strains into different species of the Bradyrhizobium genus [4,5,6, 8, 12, 13]. This analysis, together with new methods for genome-to-genome comparisons, such as average nucleotide identity (ANI) and digital DNA–DNA hybridization (DDH) analyses [14, 15], has contributed significantly to improving the taxonomy of Bradyrhizobium and delineate new species [12, 13, 16].

In a previous study [17], 46 strains isolated from nodules of soybean inoculated with soils from different Brazilian regions (Midwest, Northeast, Southeast and South) were classified within the Bradyrhizobium genus, based on the partial sequencing of the 16S rRNA gene, and allocated within two new phylogenetic groups, based on the concatenated sequence analysis of the five housekeeping genes (atpD, dnaK, gyrB, recA and rpoB). In the present study, five strains of one of these groups (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22), efficient in nitrogen fixation in symbiosis with soybean [17], were selected for the definition of their taxonomic position.

Materials and Methods

Origin of the Strains

The five strains were isolated from effective nodules of soybean inoculated with soil from the semi-arid region, Northeast Brazil, collected in Bom Jesus (9° 19″ 21″ S and 44° 48″ 55″ W), in a previous study [17]. The soil used as an inoculum was, during previous soybean cultivation, inoculated with the following soybean-inoculant strains: SEMIA 5079 (B. japonicum), SEMIA 5080 (B. diazoefficiens), SEMIA 587 (B. elkanii), and/or SEMIA 5019 (B. elkanii). Isolation was carried out on plates with 79 culture medium [18], also known as YMA (Yeast Mannitol Agar) [19].

Phylogenetic Analysis of the 16S rRNA Gene and the Housekeeping Genes

Sequences of 16S rRNA (1288 to 1331 bp), atpD (510 pb), dnaK (280 pb), gyrB (669 pb), recA (474 to 559 pb), and rpoB (903 to 957 pb) genes of the five strains were obtained in a previous study [17]. Sequences of each gene were aligned using the ClustalW Multiple Alignment algorithms in the BioEdit software. For comparison, the alignment included the sequences of type strains of the Bradyrhizobium species available in the GenBank (National Center for Biotechnology Information, NCBI). The sequences of four strains currently used as soybean inoculants in Brazil (SEMIA 5079 B. japonicum, SEMIA 5080 B. diazoefficiens, SEMIA 587 B. elkanii, and SEMIA 5019 B. elkanii) were also included in the alignment of dnaK and recA genes.

The first multilocus sequence analysis (MLSA) was conducted using the sequences of the dnaK and recA genes since their sequences are the only available, among the five housekeeping genes we studied, for the four strains currently used as soybean inoculants in Brazil. An MLSA with atpD, dnaK, gyrB, recA, and rpoB genes, without the inoculant strains, was also performed. Phylogenetic trees were constructed by the maximum likelihood method (ML) [20], using the Kimura 2 parameter model [21]. The MEGA 6 software package [22] was used in the construction of trees, with bootstrap values based on 1000 replications. The accession numbers of the gene sequences used in this study are indicated in the phylogenetic trees. Bosea thiooxidans DSM 9653T was used as an outgroup in the phylogenetic analysis of the 16S rRNA gene.

Genome Sequencing, Average Nucleotide Identity, and Digital DNA–DNA Hybridization

Genomic comparisons were used in this study to support the taxonomic position of the new group. The average nucleotide identity (ANI) and digital DNA–DNA hybridization were estimated with the genomic sequences of a representative strain of the group (UFLA06-13) and of the phylogenetically closest type strains (Bradyrhizobium brasilense UFLA03-321T, Bradyrhizobium pachyrhizi PAC48T, and Bradyrhizobium elkanii USDA 76T) in MLSA with atpD, dnaK, gyrB, recA, and rpoB genes.

Strain UFLA06-13 was grown in liquid 79 medium for genome sequencing. Genomic DNA was purified from 109 bacterial cells using the phenol–chloroform extraction protocol. The DNA library was constructed from 1 ng of total DNA with the Nextera XT kit (Illumina), following the conditions suggested by the manufacturer. Pair-end reads (2 × 250 bases) were sequenced with the MiSeq Reagent kit 500v2 (Illumina) on the MiSeq platform (Illumina). The estimated sequencing depth was 30×, and the raw reads were subject to de novo assembly in SPAdes 3.12, with default parameters [23]. The resulting assembly consisted of 8.59 Mb distributed in 561 contigs, with N50 of 188,493. Using checkM 1.0.11 [24], the completeness of the assembled genome, based on 824 lineage-specific marker genes, was estimated as 100%, whereas the contamination was estimated as only 0.52%. The authenticity of the sequenced genome was checked by verifying that the five sequences of housekeeping genes (atpD, dnaK, gyrB, recA, and rpoB) reported by Ribeiro et al. [17] for this strain were identical to those extracted from the genome.

Average nucleotide identity (ANI) values were calculated using an online calculator (available at https://enve-omics.ce.gatech.edu/ani/index) [14] to compare the strain UFLA06-13 (accession number SNUC00000000), obtained in this study, with the strains UFLA03-321T (accession number MPVQ00000000), PAC48T (accession number SAMN03782120), and USDA 76T (accession number NZ_ARAG00000000).

Phenotypic Characterization

The five strains were phenotypically characterized based on parameters previously used to differentiate Bradyrhizobium species. Two soybean inoculants strains classified as Bradyrhizobium elkanii (SEMIA 587 and SEMIA 5019) were also included in the analyses. The growth of these strains in 79 medium was evaluated under different conditions of pH (4, 5.5, 6.8, 8, 9 and 10), NaCl (w/v) (0.01, 0.25, 0.5, 0.75 and 1%), and temperature (5, 15, 20, 28, 34, 37 and 40 °C), according to the methodology previously described by Florentino et al. [25].

Their ability to assimilate different carbon sources (d-arabinose, l-asparagine, citric acid, d-fructose, glycerol, glycine, d-glucose, l-glutamine, l-glutamic acid, lactose, malic acid, maltose, mannitol, l-methionine, sodium lactate, and sucrose) and nitrogen sources (l-arginine, l-asparagine, casein hydrolyzed, l-cysteine, glycine, l-glutamic acid, and l-methionine and tryptophan) was evaluated in modified 79 medium, based on the composition described in the study of Costa et al. [12].

Antibiotic resistance was also tested in 79 medium on plates with bio-disks containing the following antibiotics: ampicillin (10 μg mL−1), cefuroxime (30 μg mL−1), ciprofloxacin (5 μg mL−1), chloramphenicol (30 μg mL−1), doxycycline (30 μg mL−1), erythromycin (15 μg mL−1), gentamicin (10 μg mL−1), kanamycin (30 μg mL−1), and neomycin (30 μg mL−1), according to the protocol previously described by Guimarães et al. [26].

In previous studies, the growth of the type strains UFLA03-321T (B. brasilense), PAC48T (B. pachyrhizi), and USDA 76T (B. elkanii) in 79 medium was evaluated under the same conditions of pH, NaCl, temperature, carbon, and nitrogen sources, as well their resistance to those nine antibiotics above, using the same protocols for strain characterization adopted in this study [12, 26]. These strains were included in the phenotypic characterization table for comparison.

Phylogenetic Analysis of the Symbiotic Genes (nodC and nifH)

For the phylogenetic analysis of the symbiotic genes (nodC and nifH), the DNA of five strains was extracted by the alkaline lysis method [27]. The protocol of Sarita et al. [28], modified by De Meyer et al. [29], was used for amplification and sequencing of the nodC gene. The amplification and sequencing of the nifH gene were performed according to Gaby and Buckley [30]. For strains UFLA06-21 and UFLA06-19, the amplification of the nodC and nifH genes, respectively, was not possible. The type strains of the Bradyrhizobium species available in the GenBank and the four strains used as soybean inoculants were included in the alignment of sequences of the nifH and nodC genes (SEMIA 5079, SEMIA 5080, SEMIA 587 and SEMIA 5019). Phylogenetic trees were constructed as described above.

Results

Phylogenetic Analysis of the 16S rRNA Gene and the Housekeeping Genes

The five strains showed 16S rRNA gene sequences identical to SEMIA 587 (B. elkanii), SEMIA 5019 (B. elkanii), UFLA03-321T (B. brasilense), PAC48T (B. pachyrhizi), and WR4T (B. ripae) and shared 99.91% similarity with USDA 76T (B. elkanii) and BR 10303T (B. macuxiense) (Fig. 1). In the first MLSA, performed with the sequences of the housekeeping genes dnaK and recA, the five strains had identical sequences and formed a separate group, which shared 97.41% of similarity with UFLA03-321T (B. brasilense), 97.28% of similarity with USDA 76T (B. elkanii), 97.27% of similarity with PAC48T (B. pachyrhizi), 97.08% of similarity with WR4T (B. ripae), and 97.84% of similarity with SEMIA 587 (B. elkanii) and SEMIA 5019 (B. elkanii) (Supplementary Fig. 1). However, concatenated sequence analysis of the five housekeeping genes (atpD, dnaK, gyrB, recA, and rpoB) revealed that the five strains share greater similarity with B. brasilense UFLA03-321T (98.95%), followed by B. pachyrhizi PAC48T (97.63) and B. elkanii USDA 76T (97.62) (Fig. 2).

Fig. 1
figure 1

Maximum Likelihood phylogeny based on 16S rRNA gene sequences (1226 bp) showing the relationships between strains UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22 (in bold) and type strains of the Bradyrhizobium species. Bootstrap values higher than 50% are indicated at nodes. GenBank accession numbers are provided in parentheses

Full size image
Fig. 2
figure 2

Maximum Likelihood phylogeny based on partial concatenated sequences (2036 bp) of housekeeping genes (atpD, dnaK, gyrB, recA, and rpoB) showing the relationships between strains UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22 (in bold) and type strains of the Bradyrhizobium species. Bootstrap values higher than 50% are indicated at nodes. GenBank accession numbers are provided in parentheses

Full size image

Genome Sequencing, Average Nucleotide Identity, and Digital DNA–DNA Hybridization

The ANI values estimated from the genome sequences of the representative strain of the group (UFLA06-13) and of the phylogenetically closest type strains UFLA03-321T (B. brasilense), PAC48T (B. pachyrhizi), and USDA 76T (B. elkanii) were 98.28, 95.50, and 95.70%, respectively.

Phenotypic Characterization

All strains grow at a wide range of pH levels (4–10) and temperatures (15–37 °C); however, they do not grow at 1% NaCl in 79 culture medium (Table 1). Strains PAC 48T (B. pachyrhizi), USDA 76T, and SEMIA 587 (B. elkanii) show weak growth at 40 °C, while B. brasilense strains (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, UFLA06-22, and UFLA03-321T) and the strain SEMIA 5019 (B. elkanii) does not grow at this temperature (Table 1). All strains use d-arabinose, l-arabinose, glycerol, and mannitol but do not use citric acid and glycine as a carbon source (Table 1).

Table 1 Phenotypic characteristics of the Bradyrhizobium brasilense strains isolated from Glycine max nodules (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22) and Vigna unguiculata nodule (UFLA03-321T) from Brazilian soils and phylogenetically close Bradyrhizobium strains
Full size table

B. brasilense strains (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, UFLA06-22, and UFLA03-321T) weakly use l-asparagine and do not use malic acid, l-methionine, sodium lactate, and sucrose. Conversely, strains PAC48T (B. pachyrhizi), USDA 76T, SEMIA 587, and SEMIA 5019 (B. elkanii) use l-asparagine, sodium lactate, and sucrose and weakly use l-methionine as a carbon source. The use of d-glucose, l-glutamic acid, l-glutamic acid, and lactose as a carbon source varied among B. brasilens strains.

In relation to the nitrogen sources, all strains assimilate l-asparagine and l-glutamic acid, but none of them use glycine as a nitrogen source (Table 1). Likewise, B. brasilense strains (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, UFLA06-22, and UFLA03-321T) do not use l-cysteine, l-methionine, and tryptophan as N source. All strains are tolerant to the antibiotics ampicillin, cefuroxime, ciprofloxacin, chloramphenicol, doxycycline, and erythromycin, except for UFLA06-15, which showed weak tolerance to chloramphenicol, doxycycline, and erythromycin. Strains UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, UFLA06-22, and PAC 48T are sensitive to the antibiotics gentamycin, kanamycin, and neomycin.

Phylogenetic Analysis of the Symbiotic Genes (nodC and nifH)

Strains UFLA06-13, UFLA06-15, UFLA06-19, and UFLA06-22 had nodC gene sequences identical to USDA 76T and SEMIA 5019 (B. elkanii) and shared 99.45% similarity with SEMIA 587 (B. elkanii), forming a separate group, close to B. ferriligni CCBAU 51502T (96.33% similarity) (Supplementary Fig. 2). In the phylogenetic analysis of the nifH gene, strains UFLA06-13, UFLA06-15, UFLA06-21, and UFLA06-22 showed identical sequences and shared 99.43% of similarity with USDA 76T (B. elkanii), 98.84% of similarity with SEMIA 5019 and SEMIA 587 (B. elkanii), and 98.25% of similarity with B. ferriligni CCBAU 51502T) (Supplementary Fig. 3). The new B. brasilense strains shared only 78.39 and 89.74% of similarity with the type strain of B. brasilense (UFLA03-321T) in the phylogenetic analysis of the nodC and nifH genes, respectively.

Discussion

This study investigated the taxonomic position of a group of Bradyrhizobium strains (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22) isolated from nodules of soybean inoculated with soils from the semi-arid region, Northeast Brazil. The fact that the five strains show 16S rRNA gene sequences identical to five Bradyrhizobium species confirms that the 16S rRNA gene does not have sufficient discriminatory power to distinguish members of the Bradyrhizobium genus, corroborating previous studies [10, 12, 13, 17, 26, 31].

Multilocus sequence analysis of housekeeping genes (atpD, dnaK, glnII, gyrB, recA, and rpoB) has been successfully employed for better discrimination among closely related strains within the Bradyrhizobium genus [6,7,8, 16, 32]. In the present study, the MLSA performed with the sequences of the housekeeping genes dnaK and recA was not sufficient to define the taxonomic position of the five strains (Supplementary Fig. 1). Conversely, MLSA with five genes (atpD, dnaK, gyrB, recA, and rpoB) indicated that these strains belong to the Bradyrhizobium brasilense species (Fig. 2). This result reveals the possibility of getting a better taxonomic resolution by employing a larger number of housekeeping genes in the MLSA.

Taxonomic groups separated by MLSA are usually evaluated by DNA–DNA hybridization (DDH) analysis to confirm their taxonomic position. DDH analysis has been used as the standard method for species delineation, considering a cutoff score of 70% genomic relatedness [33]. Instead of using the traditional DDH, this study used a new method to compare bacterial genomes, the average nucleotide identity (ANI). This method has been extensively used to replace the traditional DDH analysis, avoiding the high costs, labor, and specialized infrastructure. The recommended cutoff score of 70% based on traditional DDH for species delineation [33] has been found to correlate to 95–96% of the ANI values [14, 34].

A recent study applied a threshold of 96% ANI to delineate a new species of Bradyrhizobium (B. brasilense UFLA03-321T), which showed ANI values of 95.5% with B. pachyrhizi (PAC48T) and 94.0% with B. elkanii (USDA 76T) [12]. The cutoff score of 96% has also been previously used to delineate species in other genera [35]. According to this cutoff score, the ANI values of 95.50 and 95.70% between the strain UFLA06-13 and strains PAC48T (B. pachyrhizi) and USDA 76T (B. elkanii), respectively, in this study, indicate that they represent distinct species. Conversely, the ANI value of 98.28% between the strain UFLA06-13 and the strain UFLA03-321T supports the classification of the group as Bradyrhizobium brasilense, confirming the taxonomic position indicated by MLSA of the housekeeping genes atpD, dnaK, gyrB, recA, and rpoB. The present study reports for the first time the classification of strains isolated from nodules of soybean within a new Bradyrhizobium species (B. brasilense) native from Brazilian soils.

Phenotypic characteristics usually have little taxonomic value since they vary within the same bacterial species [36]. However, the information on phenotypic variability can be useful to verify that strains within a given species are not merely clones. Although the B. brasilense strains (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, UFLA06-22, and UFLA03-321T) showed similar phenotypes in relation to most of the evaluated characteristics, they differed regarding the growth at 0.75% NaCl; the use of d-Glucose, l-Glutamine, l-Glutamic acid, and Lactose as a carbon source; the use of l-Arginine as a nitrogen source; and the tolerance to the antibiotics Chloramphenicol (30 μg mL−1), Doxycycline (30 μg mL−1), Erythromycin (15 μg mL−1), Gentamycin (10 μg mL−1), and Neomycin (30 μg mL−1). The main differential phenotypic characteristics between B. brasilense strains and phylogenetically closest strains were the absence of growth at 40 °C; weak use of l-asparagine and no use of malic acid; l-methionine, sodium lactate, and sucrose as a carbon source; and l-cysteine, l-methionine, and tryptophan as a nitrogen source.

The tolerance of the strains to a wide range of pH levels and temperatures and their resistance to several antibiotics corroborate previous studies on Bradyrhizobium strains [13, 16, 26, 37]. These characteristics are desirable in the selection of rhizobia since the inoculant strains are recommended for use under different soil and climate conditions. These strains also exhibited good symbiotic efficiency with soybean in a previous study, showing their potential for selection and use as inoculants in this crop [17].

Genes involved in nodulation and nitrogen fixation processes usually do not provide relevant taxonomic information for species classification. However, phylogenetic analyses of symbiotic genes have been used for the delineation of symbiovars, based on similar symbiotic behaviors of rhizobia groups with a determined host range [38]. The symbiovars within the Bradyrhizobium genus have been mainly defined according to the analysis of the nodC gene [32, 39,40,41]. However, other genes, such as nodA and nifH, have also been recommended for this purpose [41, 42].

In this study, phylogenetic analyses of nodC and nifH genes clustered the strains UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22 with B. elkanii (strains USDA 76T, SEMIA 587, and SEMIA 5019) and B. ferriligni (strain CCBAU 51502T). These B. elkanii strains, isolated from Glycine max nodules in soils from the United States (USDA 76T) [2], from South (SEMIA 587) and Southeast (SEMIA 5019) regions of Brazil [43], and the strain of B. ferriligni (CCBAU 51502T), isolated from Erythrophleum fordii nodule in soils from China [10], have been recently classified within a new symbiovar denominated sojae, based on the phylogenetic analysis of the genes nodC and nifH [41]. Thus, this study proposes that the new B. brasilense strains (UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22) must be allocated within the symbiovar sojae.

Predominance of slow-growing non-acid producing rhizobia type (now assigned to Bradyrhizobium genus) in tropical areas was hypothesized by Norris [44]. Indeed, for instance Bradyrhizobium japonicum and B. elkanii are widespread symbionts of diverse Brazilian native legume species from Amazonia and Atlantic forests [45, 46] in non-inoculated areas, many of them similar to B. elkanii USDA 76T. Thus, it is unlikely that these native Bradyrhizobium species had acquired their symbiotic genes from soybean nodulating bacteria because this exotic species was introduced in the end of the nineteen century in Brazil. This is also supported by the fact that Bradyrhizobium is being found to be the predominant rhizobia genus in evaluations by culture-independent techniques throughout the world [47] indicating its ubiquity and a high saprophytic ability. Interestingly, B. brasilense belongs to B. elkanii superclade [48]. Thus, the high affinity of nodC and nifH genes of B. brasilense sv. sojae with B. elkanii USDA 76T is coherent with the affinity of their genomes. It is possible that our strains acquired symbiotic genes by horizontal gene transfer (HGT) from soybean strains in the area as hypothesized by Ferreira and Hungria [49] and Barcellos et al. [50] to explain the variability among isolates observed in their study. However, other strains of B. brasilense isolated from non-inoculated fields were efficient, for instance, in siratro [12], which unlike soybean is a promiscuous host capable of nodulating with bradyrhizobia containing a wide range of nod genotypes. Moreover, symbiotic genes are in the chromosome (not so easily transferred, although rapid transfer has been demonstrated, for example, in Germany [51]), and the fields in our study were inoculated recently [17]. In spite of this evidence of HGT from inoculant soybean strains to local B. brasilense strains, there is a possibility that B. brasilense had already harbored the symbiotic genes, although we have yet to observe these genes in B. brasilense strains isolated from native legumes.

The ability of the five new B. brasilense strains to effectively nodulate and fix nitrogen with their original host (Glycine max) was confirmed in a previous study [17]. The strain UFLA06-13 has also been evaluated for nodulation capacity in other hosts (Vigna unguiculata, Phaseolus lunatus, Stizolobium aterrimum, and Acacia mangium) [52, Costa et al., unpublished results]. This strain formed nodules in V. unguiculata and S. aterrimum but did not nodulate P. lunatus and A. mangium. Interestingly, the four B. brasiliense strains described in a previous study (UFLA03-321T, UFLA03-320, and UFLA03-290 from V. unguiculata nodules; and UFLA04-0212, from Macroptilium atropurpureum nodules) also did not nodulate A. mangium [52, Costa et al., unpublished results]; however, they were able to nodulate V. unguiculata, S. aterrimum, and Glycine max [12, 26, 53, Costa et al., unpublished results], except for strain UFLA03-290, which was not evaluated for its nodulation capacity in soybean. Similarly, strains UFLA06-13 and UFLA03-290 do not nodulate P. lunatus, while strains UFLA03-321T, UFLA03-320, and UFLA04-0212 inefficiently nodulate this legume [52]. Although the B. brasilense strains used in this study belong to a symbiovar (sv. sojae) different from those of the B. brasilense strains previously described (sv. tropici) [9], the data regarding the ability of nodulation with legumes clearly show that these B. brasilense strains share some symbiotic similarity.

Conclusions

Results from the multilocus sequence analysis of five housekeeping genes (atpD, dnaK, gyrB, recA, and rpoB), the average nucleotide identity analysis, and the phenotypic characterization support the classification of strains UFLA06-13, UFLA06-15, UFLA06-19, UFLA06-21, and UFLA06-22 as Bradyrhizobium brasilense. Phylogenetic analyses of nodC and nifH genes show that these strains belong to the symbiovar sojae, together with B. elkanii (USDA 76T) and B. ferriligni (CCBAU 51502T).