Rhizobial 16S rRNA and dnaK genes: mosaicism and the uncertain phylogenetic placement of Rhizobium galegae

Abstract

The phylogenetic relatedness among 12 agriculturally important species in the order Rhizobiales was estimated by comparative 16S rRNA and dnaK sequence analyses. Two groups of related species were identified by neighbor-joining and maximum-parsimony analysis. One group consisted of Mesorhizobium loti and Mesorhizobium ciceri, and the other group consisted of Agrobacterium rhizogenes, Rhizobium tropici, Rhizobium etli, and Rhizobium leguminosarum. Although bootstrap support for the placement of the remaining six species varied, A. tumefaciens, Agrobacterium rubi, and Agrobacterium vitis were consistently associated in the same subcluster. The three other species included Rhizobium galegae, Sinorhizobium meliloti, and Brucella ovis. Among these, the placement of R. galegae was the least consistent, in that it was placed flanking the A. rhizogenes-Rhizobium cluster in the dnaK nucleotide sequence trees, while it was placed with the other three Agrobacterium species in the 16S rRNA and the DnaK amino acid trees. In an effort to explain the inconsistent placement of R. galegae, we examined polymorphic site distribution patterns among the various species. Localized runs of nucleotide sequence similarity were evident between R. galegae and certain other species, suggesting that the R. galegae genes are chimeric. These results provide a tenable explanation for the weak statistical support often associated with the phylogenetic placement of R. galegae, and they also illustrate a potential pitfall in the use of partial sequences for species identification.

FIG. 1.

Unrooted neighbor-joining trees based on single-gene sequences from 12 species representing five genera (Agrobacterium, Brucella, Mesorhizobium, Rhizobium, and Sinorhizobium) within the order Rhizobiales. Percentage bootstrap support at each internal node is based on 1,000 replicate trees. (A) 16S rRNA nucleotide sequence tree. The total alignment length for the analysis was 1,401 bp. Positions with gaps were omitted, and the Kimura two-parameter distance correction was applied. (B) dnaK nucleotide sequence tree. The total alignment length for the analysis was 1,917 bp. Positions with gaps were omitted, and the Kimura two-parameter distance correction was applied. (C) Inferred DnaK amino acid sequence tree. The total alignment included 642 amino acid residues. Positions with gaps were omitted, and rates of amino acid substitution were assumed to follow a Poisson distribution.

FIG. 2.

Unrooted maximum-parsimony trees based on single-gene sequences from 12 species representing five genera (Agrobacterium, Brucella, Mesorhizobium, Rhizobium, and Sinorhizobium) within the order Rhizobiales. Trees were generated by a heuristic mini-min tree search option with a search factor of 2. The percentage bootstrap support at each internal node is based on 1,000 replicate trees. (A) 16S rRNA nucleotide sequence tree. The total alignment length for the analysis was 1,401 bp. (B) dnaK nucleotide sequence tree. The total alignment length for the analysis was 1,917 bp. (C) Inferred DnaK amino acid sequence tree. The total alignment included 642 amino acid residues.

FIG. 3.

Linear distribution of 182 polymorphic nucleotide positions in a multiple alignment of 16S rRNA sequences representing five genera (Agrobacterium, Brucella, Mesorhizobium, Rhizobium, and Sinorhizobium) within the order Rhizobiales. Each vertical line represents a deviation from the consensus sequence. The locations of hypervariable expansion segments (V1 through V10) described by Raué et al. (15) are shown at the bottom. Brackets define endpoints of segments that were identified by the Stephens test as containing nonrandom clusters of partition-specific nucleotide sequences. Shaded segments indicate runs of sequence similarity that were identified by Sawyer's Geneconv method.

FIG. 4.

Linear distribution of 193 polymorphic amino acid positions in a multiple alignment of DnaK protein sequences representing five genera (Agrobacterium, Brucella, Mesorhizobium, Rhizobium, and Sinorhizobium) within the order Rhizobiales. Each vertical line represents a deviation from the consensus sequence. Regions corresponding to the three domains described by Mogk et al. (13) are shown at the bottom. Shaded segments indicate statistically significant runs of nucleotide sequence similarity identified by Sawyer's method.