Virus classification - where do you draw the line?

Abstract

High-throughput sequencing (HTS) and its use in recovering and assembling novel virus sequences from environmental, human clinical, veterinary and plant samples has unearthed a vast new catalogue of viruses. Their classification, known by their sequences alone, sets a major challenge to traditional virus taxonomy, especially at the family and species levels, which have been historically based largely on descriptive taxon definitions. These typically entail some knowledge of their phenotypic properties, including replication strategies, virion structure and clinical and epidemiological features, such as host range, geographical distribution and disease outcomes. Little to no information on these attributes is available, however, for viruses identified in metagenomic datasets. If such viruses are to be included in virus taxonomy, their assignments will have to be guided largely or entirely by metrics of genetic relatedness. The immediate problem here is that the International Committee on Taxonomy of Viruses (ICTV), an organisation that authorises the taxonomic classification of viruses, provides little or no guidance on how similar or how divergent viruses must be in order to be considered members of new species or new families. We have recently developed a method for scoring genomic (dis)similarity between viruses (Genome Relationships Applied to Virus Taxonomy - GRAViTy) among the eukaryotic and prokaryotic viruses currently classified by the ICTV. At the family and genus levels, we found large-scale consistency between genetic relationships and their taxonomic assignments for eukaryotic viruses of all genome configurations and genome sizes. Family assignments of prokaryotic viruses have, however, been made at a quite different genetic level, and groupings currently classified as sub-families are a much better match to the eukaryotic virus family level. These findings support the ongoing reorganisation of bacteriophage taxonomy by the ICTV Phage Study Group. A rapid and objective means to explore metagenomic viral diversity and make evidence-based assignments for such viruses at each taxonomic layer is essential. Analysis of sequences by GRAViTy provides evidence that family (and genus) assignments of currently classified viruses are largely underpinned by genomic relatedness, and these features could serve as a guide towards an evidence-based classification of metagenomic viruses in the future.

Fig. 1

Heat maps of CGJ distances between dsDNA. Pairwise CGJ distances were computed between each sequence plotted as a heat map using colour-coded points (see scale on the left of the figure). The light grey solid lines indicate boundaries between each virus taxonomic group. Annotations for each virus family and order are shown on the axes. This figure has been reproduced from reference [5]

Fig. 2

Virus dendrograms based on CGJ distances. UPGMA dendrograms were constructed from pairwise distance matrices, and the tips are labelled with current ICTV family and genus assignments. Virus taxonomy at the order level is also shown to the right of the dendrograms. The scale bar for the CGJ distance is shown at the bottom. Bootstrap clade support values of ≥ 30% are shown on the branches. Values ≥ 70% are in black, otherwise are in grey. Bootstrap support values for collapsed clades are shown regardless of the values. This figure has been modified from Fig. 2 in reference [5]

Fig. 3

UPGMA dendrogram of classified and unclassified dsDNA viruses (Baltimore group I) based on CGJ distances. The dendrogram is divided into six separate lines to represent the 139 clades present in the dataset. Tips are labelled with genus for members of the Caudovirales; abbreviated as S: Siphoviridae, M: Myoviridae, and P: Podoviridae, and by family/genus for other bacterial, eukaryotic and archaeal viruses, or by accession number codes for unclassified viruses. The scale bar for CGJ distance is shown at the left of each line, and the 0.8 threshold that corresponds to eukaryotic family groupings is shown as a grey dotted line. Bootstrap re-sampling was performed with pruned signature tables as described previously [5]. Clades were coloured based on host origin according to the key; those containing both classified and unclassified (U) sequences are shown in a lighter shade. The 39 new candidate unassigned taxonomic units (UTUs) arising from the inclusion of current unclassified viruses are shaded in light blue. This figure has been reproduced from reference [4]

Fig. 4

Dendrogram of members of the order Caudovirales that have subfamily assignments. Taxa are colour coded for family (see key). Minor discrepancies between genus assignments and phylogeny are shown in red circles. This figure has been reproduced from reference [4]