Next-generation systematics: An innovative approach to resolve the structure of complex prokaryotic taxa

Abstract

Prokaryotic systematics provides the fundamental framework for microbiological research but remains a discipline that relies on a labour- and time-intensive polyphasic taxonomic approach, including DNA-DNA hybridization, variation in 16S rRNA gene sequence and phenotypic characteristics. These techniques suffer from poor resolution in distinguishing between closely related species and often result in misclassification and misidentification of strains. Moreover, guidelines are unclear for the delineation of bacterial genera. Here, we have applied an innovative phylogenetic and taxogenomic approach to a heterogeneous actinobacterial taxon, Rhodococcus, to identify boundaries for intrageneric and supraspecific classification. Seven species-groups were identified within the genus Rhodococcus that are as distantly related to one another as they are to representatives of other mycolic acid containing actinobacteria and can thus be equated with the rank of genus. It was also evident that strains assigned to rhodococcal species-groups are underspeciated with many misclassified using conventional taxonomic criteria. The phylogenetic and taxogenomic methods used in this study provide data of theoretical value for the circumscription of generic and species boundaries and are also of practical significance as they provide a robust basis for the classification and identification of rhodococci of agricultural, industrial and medical/veterinary significance.

Figure 1

Un-rooted radial maximum-likelihood phylogenetic trees derived from (A) concatenated codon alignment of the core genome (scale bar represents nucleotide substitutions per codon site) and (B) a subset of amino acids from 400 broadly conserved prokaryotic proteins. The scale bar shows normalized fraction of total branch lengths as described by Segata et al..

Figure 2

3D graphical representation of pairwise similarity matrices obtained by (A) fragmented BLAST searches (FBS values), (B) genomic average nucleotide identities (ANIb-G values), (C) average nucleotide identities among core genes (ANIb-C values) and (D) average amino-acid identities from the core genes (AAI values). Rhodococcus species-groups A-G are labelled whilst the reference genera are plotted at the lower right hand corner.

Figure 3

Average taxogenomic values (filled diamonds), (A) fragmented BLAST similarities (FBS), (B) genomic average nucleotide identities (ANIb-G), (C) average nucleotide identities among core genes (ANIb-C) and (D) average amino-acid identities from the core genes (AAI) with standard deviations. The median values are shown with filled circles. Average pairwise similarities with standard deviations within species, within groups of species (excluding similarity among members assigned to the same species), and between different groups are marked in light, intermediate and dark grey colour, respectively. The average diversity between different individual groups is plotted at the bottom for each group against species-groups (A–G), Nocardia, Gordonia, Corynebacterium diphtheriae, Williamsia and Segniliparus, respectively (excluding self-values).

Figure 4. A 3D distribution of genome size, GC content and fraction of shared genes within each species-group (Supplementary Table 1).

The three axes are shown in blue in the centre of the plot and are labelled. The individuals belonging to seven species-groups are shown in different colours.