16S rRNA phylogeny and clustering is not a reliable proxy for genome-based taxonomy in Streptomyces

Abstract

Streptomyces is among the most extensively studied genera of bacteria but its complex taxonomy remains contested and is suspected to contain significant species-level misclassification. Resolving the classification of Streptomyces would benefit many areas of applied microbiology that rely on an accurate ground truth for grouping of related organisms, including comparative genomics-based searches for novel antimicrobials. We survey taxonomic conflicts between 16S rRNA and whole genome-based Streptomyces classifications using 2276 publicly available Streptomyces genome assemblies and 48 981 publicly available full-length 16S rRNA Streptomyces sequences from silva, Greengenes, Ribosomal Database Project (RDP), and NCBI (National Centre for Biotechnology Information) databases. We construct a full-length 16S gene tree for 14 239 distinct Streptomyces sequences that resolves three major lineages of Streptomyces, but whose topology is not consistent with existing taxonomic assignments. We use these sequence data to delineate 16S and whole genome landscapes for Streptomyces, demonstrating that 16S and whole-genome classifications are frequently in disagreement, and that 16S zero-radius Operational Taxonomic Units (zOTUs) are often inconsistent with Average Nucleotide Identity (ANI)-based taxonomy. Our results strongly imply that 16S rRNA sequence data does not map to taxonomy sufficiently well to delineate Streptomyces species routinely. We propose that alternative marker sequences should be adopted by the community for classification and metabarcoding. Insofar as Streptomyces taxonomy has been determined or supported by 16S sequence data and may in parts be in error, we also propose that reclassification of the genus by alternative approaches may benefit the Streptomyces community.

Keywords: 16S; evolution; phylogeny; taxonomy; Actinomycetales; Streptomyces.

Fig. 1.. Empirical cumulative distribution of taxonomic composition (the number of unique, identifiable species-level taxIDs) at a range of pairwise percentage sequence identity clustering thresholds for full-length 16S sequences. The y-axis begins at the ninety-eighth percentile and, even for zOTUs (100% identity clustering threshold), at this level no zOTU is associated with only a single species-level taxID. At less stringent pairwise identity thresholds, the number of taxIDs observed at a given percentile is larger. The result demonstrates that a Streptomyces zOTU cannot in general be assumed to map to a single species-level taxonomic assignment, and that such assignments are progressively less secure as a lower sequence identity threshold is considered.

Fig. 2.. Maximum-likelihood tree of the genus Streptomyces constructed from 9049 full-length 16S rRNA sequences. Clades containing a single assigned species-level taxon were collapsed to single leaf nodes. Three major clades (Clade 1–Clade 3) are highlighted in distinct colours. Squares indicate 16S sequences assigned to the same species names in the source database(s): S. griseus sequences in blue, S. clavuligerus in red, S. lydicus in green, and S. scabei in purple. Sequences with these species assignments tend not to be monophyletic, indicating incongruence between taxonomy and the 16S gene tree. An equivalent rectangular phylogram is provided in File S31.

Fig. 3.. Network graph of 1369 Streptomyces genomes, linked by common 16S sequence. There are 709 discrete subgraphs. Each node represents a distinct genome assembly. Each edge corresponds to at least one identical 16S sequence being shared between that pair of genomes. Blue connected components form cliques in which every genome shares at least one identical 16S sequence with all other genomes in the same connected component. Green connected components do not have this property.

Fig. 4.. Heatmaps of ANIm coverage (left), and ANIm identity (right) for three example connected components from Fig. 3. Heatmaps in the same row correspond to comparisons for the same connected component. The left column represents percentage genome coverage, the right column %ANI. Red cells in coverage plots correspond to genome coverage of 50% or above, interpreted as common membership of the same genus; blue cells correspond to coverage below 50% and imply distinct genus assignments. In ANI plots, red cells correspond to genome identity of 95% or above, interpreted as membership of the same species; blue cells represent imply distinct species. Here, connected components correspond to: (a, b) genomes from a single genus and species (solid red heatmap for both coverage and identity), (c, d) distinct species belonging to the same genus (solid red heatmap for coverage, some pairwise identities below 95%), or (e, f) distinct genera and species (some pairwise coverage and some pairwise identities below 50%). ANIm coverage and identity plots for the remaining connected components are provided in Supplementary File 20, and the numbers of subgraphs falling into each category is summarized in Table S1.

Fig. 5.. Connected components describing assemblies from distinct candidate genera (a) and species (b) Nodes represent distinct genome assemblies, while edges indicate the presence of at least one identical 16S sequence shared between the corresponding pair of genomes. Each unique candidate genus (a) or species (b) is represented as a distinct node colour.

Fig. 6.. Scatterplots showing genome coverage for pairwise ANI comparisons of genomes sharing identical full-length and ambiguity symbol-free 16S sequences. The number of unique NCBI taxonomy-derived species names per cluster is displayed at the top of each subgrouping, and the red horizontal line at 50% indicates the whole-genome genus circumscription threshold. Within-species pairwise comparisons (>≈95% genome identity) are shown in blue, and between-species comparisons (<95% genome identity) are shown in red. The cluster uniting genomes with the lowest genome coverage is outlined in the red box.

Fig. 7.. Scatterplots showing genome identity for pairwise ANI comparisons of genomes sharing identical full-length and ambiguity symbol free 16S sequences. The number of unique NCBI taxonomy-derived species names per cluster is displayed at the top of each plot, and the red horizontal line at 95% indicates the whole-genome species circumscription threshold. Within-genus comparisons (>≈50% genome coverage) are shown in orange, and between-genus comparisons (<50% genome coverage) are shown in purple. The cluster uniting genomes with the lowest genome identity is outlined in the red box.