Identification and classification of the genus Bacteroides by multilocus sequence analysis.

Multilocus sequence analysis (MLSA) was performed on representative species of the genus Bacteroides. Internal fragments of the genes selected, dnaJ, gyrB, hsp60, recA, rpoB and 16S rRNA, were amplified by direct PCR and then sequenced from 38 Bacteroides strains representing 35 species. Neighbour-joining (NJ), maximum-likelihood (ML) and maximum-parsimony (MP) phylogenies of the individual genes were compared. The data confirm that the potential for discrimination of Bacteroides species is greater using MLSA of housekeeping genes than 16S rRNA genes. Among the housekeeping genes analysed, gyrB was the most informative, followed by dnaJ. Analyses of concatenated sequences (4816 bp) of all six genes revealed robust phylogenetic relationships among different Bacteroides species when compared with the single-gene trees. The NJ, ML and MP trees were very similar, and almost fully resolved relationships of Bacteroides species were obtained, to our knowledge for the first time. In addition, analysis of a concatenation (2457 bp) of the dnaJ, gyrB and hsp60 genes produced essentially the same result. Ten distinct clades were recognized using the SplitsTree4 program. For the genus Bacteroides, we can define species as a group of strains that share at least 97.5% gene sequence similarity based on the fragments of five protein-coding housekeeping genes and the 16S rRNA gene. This study demonstrates that MLSA of housekeeping genes is a valuable alternative technique for the identification and classification of species of the genus Bacteroides.