Reclassification of Rhizobium tropici type A strains as Rhizobium leucaenae sp. nov.

A group of bacterial species, collectively known as rhizobia, can induce the formation of specific structures, named nodules, on the roots of legumes eliciting a symbiotic process in which the rhizobia fix atmospheric nitrogen and supply it to the plant. Interest in the use of rhizobia as biofertilizers in agriculture has promoted studies on their diversity and the description of a large number of rhizobial species. Rhizobium tropici is a broad host range rhizobial species that was isolated from Leucaena spp. nodules in Brazil (Hungria et al., 2000; Martínez-Romero et al., 1991), and from Gliricidia sepium (Acosta-Durán & Martínez-Romero, 2002) and Acaciella angustissima (Rincón-Rosales et al., 2009) in Mexico. The species has also been isolated from common bean (Phaseolus vulgaris) nodules in several countries (Amarger et al., 1994; Anyango et al., 1995; Diouf et al., 2000; Grange & Hungria, 2004; Pinto et al., 2007). R. tropici has been used as an efficient inoculant for beans in the tropics, due to its high tolerance to environmentally stressful conditions and its high genetic stability (Hungria et al., 2000, 2003).

R. tropici, reported in 1991 (Martínez-Romero et al., 1991), was the first description of a rhizobial species to include 16S rRNA gene sequence analysis. At that time, in spite of evidence that supported the proposal of two species from the nodule isolates analysed, e.g. differences in 16S rRNA gene sequences and low DNA–DNA hybridization values, only one novel species was accepted. However, two types, A and B, were recognized to account for these differences. Other species are now recognized to be close relatives of R. tropici, such as Rhizobium lusitanum (Valverde et al., 2006), Rhizobium multihospitium (Han et al., 2008), Rhizobium miluonense (Gu et al., 2008) and Rhizobium rhizogenes (Hernández-Lucas et al., 2004; Velázquez et al., 2010). Some of these species are either more closely related to type A or to type B strains in phylogenetic trees (Han et al., 2008; Valverde et al., 2006; Velázquez et al., 2010). Recently, it was recommended that species containing several distinct genotype clusters should be subdivided into multiple species, each corresponding to a single genotype cluster (Achtman & Wagner, 2008). In this study, we review reported differences between the two R. tropici groups and present new evidence to suggest that R. tropici type A strains belong to a novel species that is distinct from R. tropici.

The rhizobial strains used in this study are listed in Table S2 (available in IJSEM Online). Four Brazilian strains (BR 828, BR 10043, CPAO 29.8, 77) were chosen as representatives of a large collection of strains identified as R. tropici type A in previous studies (Grange & Hungria, 2004; Martínez-Romero et al., 1991; Mercante et al., 1998; Pinto et al., 2007). The well-studied strain CFN 299^T from the type A subgroup was also included and proposed as the type strain of the novel species. Three previously unreported Mexican isolates (CCGE 521, CCGE 522, CCGE 523) obtained from Zacatecas, the largest common-bean-growing area in Mexico, were also included. All strains were deposited at the ‘Diazotrophic and Plant Growth Promoting Bacteria Culture Collection’ of Embrapa Soja (Londrina, Brazil), and at the Center for Genomic Sciences Culture Collection (Cuernavaca, Mexico). Except where specified, strains were grown on yeast extract-mannitol (YM) broth [as Vincent (1970), except with a mannitol concentration of 5 g l⁻¹] in the dark, at 28 °C. Source cultures were maintained in YM agar (YMA) at 4 °C. Stocks were prepared on YM and kept at −80 °C (in 30 % glycerol) for long-term storage.

Total genomic DNA of each strain was extracted as described before (Kaschuk et al., 2006). Repetitive extragenic palindromic-PCR genomic fingerprints with the BOX-A1R primer were generated and analysed as described by Pinto et al. (2007). Genomic fingerprints revealed a group of four Brazilian and three Mexican strains that clustered with CFN 299^T, the reference type A strain. This cluster was clearly separated from R. tropici type B (Fig. S1, available in IJSEM Online) as well as from other rhizobial type or reference strains (data not shown). In a previous study, Brazilian type A strains CFN 299^T, CPAO 29.8 and 77 also clustered separately from R. tropici type B in a combined analysis of REP, ERIC and BOX-A1R fingerprints (Pinto et al., 2007).

Besides the 16S rRNA gene phylogeny, a multilocus sequence analysis approach with five protein-coding genes was used to establish the relationships between type A strains and related taxa. Near full-length 16S rRNA gene and fragments of recA, gltA, rpoA, glnII and rpoB were amplified and sequenced as described previously (Ribeiro et al., 2009; Rincón-Rosales et al., 2009). All these genes have been used previously in studies of rhizobial diversity (Martens et al., 2007; Ribeiro et al., 2009; Rincón-Rosales et al., 2009). Multiple sequence alignments were performed with clustal_x version 1.83 (Thompson et al., 1997) and manually checked with BioEdit (Hall, 1999). Best-fit models of sequence evolution were selected for each gene, and for the concatenated set of five protein-coding genes with JModelTest 0.1.1, using the Akaike information criterion (Posada, 2008). Maximum-likelihood phylogenies were constructed with PhyML 3 using subtree pruning and regrafting moves to improve tree topology (Guindon et al., 2010). Support for tree nodes was evaluated by the Shimodaira–Hasegawa-like approximate likelihood-ratio test implemented in PhyML.

R. tropici type A strains formed a well-supported clade in the 16S rRNA gene phylogeny and, together with R. lusitanum, R. rhizogenes, R. tropici type B, R. multihospitium and R. miluonense, constituted a group of closely related species, hereby designated the ‘R. tropici group’ (Fig. 1). The 16S rRNA gene sequences of type A strains showed ≥99.8 % identity to each other and ≤99.4 % identity with other strains in the ‘R. tropici group’ (Table S3, available in IJSEM Online). Willems & Collins (1993) described an insertion of 72 nucleotides in the V1 region of the 16S rRNA gene of strain CFN 299^T. Later, it was confirmed that the insertion was present in several type A but not in type B strains (Hungria et al., 2000; van Berkum et al., 1994). In this study, presence of the insertion was confirmed in the Brazilian type A strains CPAO 29.8, 77, BR 10043 and BR 828, and also in the Mexican strains CCGE521, CCGE522 and CCGE523. The insertion was absent in all available sequences from the Rhizobium type strains with a complete V1 region sequenced. Thus, the 72 bp insertion seems to be a characteristic of type A strains. Another genetic difference in the ribosomal operon between type A and type B strains was highlighted in a study by Pinto et al. (2007), in which strains CFN 299^T, CPAO 29.8 and 77 showed similar profiles in the amplified rRNA gene restriction analysis of the 23S rRNA and were separate from the cluster including R. tropici CIAT 899^T.

Nucleotide similarities between type A strains and other species of the ‘R. tropici group’ were less than 94.0, 98.7, 97.2, 93.9 and 94.4 % for the recA, gltA, rpoA, glnII and rpoB genes, respectively. Except for R. multihospitium (recA) and R. lusitanum (gltA), similarities among type A strains were higher than between this group and the other species (Table S3). Type A sequences formed well-supported clades that differed from those containing type B and other species sequences in all single gene phylogenies (Figs S3–S7). In the phylogeny constructed with the concatenated alignments (Fig. 2), the type A strains formed a sister clade to a highly supported group containing R. tropici type B, R. multihospitium and R. miluonense. R. lusitanum occupied an intermediate position between the above species and R. rhizogenes, with the latter being the basal species in the ‘R. tropici group’.

DNA–DNA hybridizations were used to determine the DNA relatedness among type A strains and between this group and the type species of related taxa. A previously described filter hybridization methodology (Martínez-Romero et al., 1991) using ³²P-labelled DNA of strain CFN 299^T as a probe was used. Type A strains shared high levels of DNA–DNA relatedness (≥78.8 %). Strain CFN 299^T showed 45.5, 37.5, 38.2, 33.2 and 28.7 % hybridization values with R. lusitanum P1-7^T, R. miluonense CCBAU 41251^T, R. tropici CIAT 899^T, R. multihospitium CCBAU 83401^T and R. rhizogenes IAM 13570^T, respectively. These data are consistent with the previously reported 39 % DNA–DNA hybridization value between CFN 299^T and CIAT 899^T (Martínez-Romero et al., 1991), which is below the threshold for species definition (Coenye et al., 2005; Vandamme et al., 1996). Other studies have also shown low DNA–DNA hybridization values between CFN 299^T and strains belonging to the ‘R. tropici group’ (Gu et al., 2008; Han et al., 2008; Valverde et al., 2006). All these data support the differentiation of type A strains from all the species in the ‘R. tropici group’.

The DNA G+C contents of three type A strains were estimated according to Moreira et al. (2011). Vibrio furnissii CAIM 518^T was included as a control; the reported DNA G+C content for this strain is 50.4 mol% (Brenner et al., 1983). The analyses were performed at least twice using six replicates in each experiment. The DNA G+C contents of strains CFN299^T, CPAO 29.8 and CCGE 522 were 64.1, 63.5 and 63.6 mol%, respectively, which are within the range reported for Rhizobium species. The fatty acid profile of the representative type A strain CFN 299^T, grown for 48 h on YMA (Vincent, 1970) plates, was determined with the MIDI system using the TSBA6 database. The results obtained (Table S5, available in IJSEM Online) further confirmed the affiliation to the Rhizobium genus as common fatty acids found in this genus (Tighe et al., 2000) were also observed in CFN 299^T.

Morpho-physiological characterization of the strains included evaluation of colony morphology, acid/alkaline production in YMA, tolerance to various pH and temperature conditions, growth in Luria–Bertani (LB) and peptone-yeast extract (PY) without Ca media, as described previously (Hungria et al., 2000; Martínez-Romero et al., 1991). Growth on selected carbon and nitrogen sources was evaluated as described previously (Martínez-Romero et al., 1991). In addition, bacteria were evaluated for the capacity to utilize the 49 carbohydrates included in the API 50CH kit (bioMérieux) as specified by the manufacturer, using YM-minus-mannitol as the basal medium. Resistance to chloramphenicol (50 µg ml⁻¹) and kanamycin sulfate (5 µg ml⁻¹) in YMA, and to captan (50 µg ml⁻¹) and thiram (25 µg ml⁻¹) in PY agar plates was also evaluated. Captan and thiram are seed-dressing fungicides commonly used in agriculture and may influence the survival of seed-applied rhizobial inoculants.

All type A strains analysed in this study showed similar reactions in the morpho-physiological tests. Known traits distinguishing type A from type B strains first reported by Martínez-Romero et al. (1991) and further described by Hungria et al. (2000) and Pinto et al. (2007), such as colony morphology, antibiotic-resistance patterns, growth on LB or PY without Ca were confirmed (Table 1). R. tropici type B strains can tolerate 40 °C, whereas type A strains grow very poorly at this temperature. Novel differences between the two groups were found in this study. Both strain types produce acid in YMA medium but only colonies from type A strains acquire a yellow colour when the medium is supplemented with bromothymol blue. Besides being more sensitive to several antibiotics than type B strains (Martínez-Romero et al., 1991), type A strains are also more sensitive to the fungicides captan and thiram (Table S4, available in IJSEM Online). Additionally, 15 out of the 49 tests in the API 50CH kit differentiated the strain types (Table S4).

In the description of R. tropici as a novel species, type A and B strains formed well separated groups from data derived from multilocus enzyme electrophoresis analysis with eight metabolic enzymes (Martínez-Romero et al., 1991). The groups were separated at a genetic distance over 0.5, which was used as the limit to distinguish species (Musser et al., 1987; Selander et al., 1985). Further analysis with glutamine synthetase II isoenzymes clearly distinguished type A and type B strains (Taboada et al., 1996). Megaplasmids of similar size (over 1700 kb) were observed in both type A and type B strains but they were found to be subgroup-specific, indicating that type A and B strains belonged to different taxa (Geniaux et al., 1995). Such megaplasmids may correspond to chromids (E. Ormeño-Orrillo and others, unpublished results) as defined by Harrison et al. (2010), and should have taxonomic value in contrast to conjugative plasmids. Other differences, reported elsewhere, include a Hup (uptake hydrogenase)-positive phenotype of type A strains, whereas only a few type B strains showed this characteristic (van Berkum et al., 1994).

All differences in genetic and phenotypic properties between the type A and B strains of R. tropici reported in this and previous studies suggest that type A strains are members of a distinct species, for which the name Rhizobium leucaenae is proposed. This novel species can also be differentiated from other species of the ‘R. tropici group’ by sequence analysis, as mentioned previously, and by the phenotypic traits presented in Table 1.

Based on data reported in previous works, R. leucaenae is present in several regions of Brazil (Grange & Hungria, 2004; Martínez-Romero et al., 1991; Mercante et al., 1998; Pinto et al., 2007). Although these regions encompass a variety of ecosystems, R. leucaenae occurs abundantly in the Cerrados savannah, which occupies about 25 % of Brazil. In that region, R. leucaenae represented 79 % and 15 % of the rhizobial population obtained using Leucaena leucocephala and common bean as trap hosts, respectively (Mercante et al., 1998). R. leucaenae has also been found in the state of Veracruz, Mexico, in nodules of Gliricidia sepium (Acosta-Durán & Martínez-Romero, 2002). In Zacatecas, the largest bean-growing area in Mexico, a low proportion (less than 10 %) of the nodule bacterial isolates from beans were identified as R. leucaenae based on analysis of 16S rRNA or rpoB gene sequences (our own unpublished results).