Argument 1

General. The virus host imposes major constraints on virus divergence which effectively links virus and host diversities. Due to the high mutation rate of viruses and the apparent lack of virus-free hosts, it is likely that virus diversity may not be smaller, and is more likely larger than host diversity. Consequently, it could be argued that the number of ranks used to classify viruses should parallel the ranks used to classify hosts, if sufficient virus diversity is available for analysis.

Argument 2

Domain and Class. Virologists recognize the Baltimore classification of viruses into six (or seven occasionally) classes (Fig. 1) at its foundation. However, the term “class” itself is not a formally recognized rank in virus taxonomy. This discrepancy is repeatedly highlighted by many virologists, and it calls for action. It would be worth considering recognizing the Baltimore groupings as taxa of the “domain” rather than class rank, which is basal level of the Linnaean taxonomy that encompasses most distant relationships.

Unlike currently recognized taxa, Baltimore classes were established using purely functional considerations concerning genome type and its expression. Its broad albeit informal use in taxonomy is due to overall (perceived) good agreement between these classes and taxa that are recognized phylogenetically. However, this correspondence is not universal, as, for instance, was demonstrated for the dsRNA Birnaviridae and the ssRNA + Permutotetraviridae, which form an inter-class monophyletic group [9, 30]. There are few other examples of complex relationship between Baltimore classes and phylogenetic groups.

The formal recognition of Baltimore classes at the basal rank of virus taxonomy would open these newly formed taxa, as any other taxa, for oversight and revision by practitioners under the ICTV auspices. Using the conventional taxonomy proposal framework of creating, dissolving, and moving taxa, and coupling it with evolutionary reasoning and public debate, the placement and composition of the original taxa including Baltimore classes could be refined, and a public record of the revision and its reasoning created. The formal recognition of Baltimore classes as taxa at the basal rank of virus taxonomy would also (re-)define the scale of the entire taxonomy, which is informative for defining scales of other newly created ranks. Combined these changes will contribute to the advancement of virus taxonomy and its use in teaching, research, and practical applications.

Argument 3

Order and above. While the basal rank of virus taxonomy is order, several “super-order” groupings of viruses have been described on phylogenetic grounds. For ssRNA + viruses, they are known as supergroups or superfamilies [7, 8, 24]. For instance, the Picornavirus-like and Alphavirus-like supergroups each include a single order as a subset, Picornavirales and Tymovirales, respectively, as well as many other more distantly related families. In the case of Picornavirus-like supergroup [16, 20], these families, e.g. Caliciviridae and Potyviridae, share different number of conserved domains and separated by different distances from the Picornavirales, indicating that more than a single additional level above order would be required to accommodate them fully in an hierarchical structure.

This example is not an exception, as was shown by our extensive analysis of the order Nidovirales, which comprises a distinct supergroup of ssRNA + viruses. This order includes two large monophyletic sets of viruses above the family level (“sub-order”), which are known as small and large nidoviruses, respectively. They are recognized using phylogeny of the most conserved proteins and the presence/absence of the replicative riboexonuclease, ExoN, that serves as a marker domain whose presence correlates with genome size. An extra rank between family and order would reflect better the relationship between the phylogeny and the taxonomy of nidoviruses. An alternative solution would be the introduction of an extra rank above the taxa order and elevation of the current nidovirus order to this rank. Since Nidovirales distantly resemble Astroviridae and a subset of Luteoviridae and all these together have sequence affinity to the Picornavirus-like supergroup, further additional ranks could be filled to reflect these relationships.

Clearly, the availability of extra levels above the family rank will facilitate the taxonomy development of other, currently less structured supergroups of ssRNA + viruses of plants and animals, and other classes of viruses, which have highly diverse monophyletic groups, e.g. Reoviridae, Mononegavirales, etc.

Argument 4

Family. Using a rigorous method for quantifying the statistical support for clusters and ranks called DEmARC, we partitioned the genomic diversity of several RNA virus families into hierarchical classifications. We observed that these classifications included extra level(s) (ranks), whose support was comparable with those of other ranks currently recognized in taxonomy. It could be argued that, at the moment, the extra level(s) in these genetics-based classification of, for example, arteriviruses and filoviruses might have been observed due to the relatively small sampling of viruses in these families (< 700 genomes). However, this explanation seems unlikely for the Coronaviridae and, particularly, Picornaviridae families, which are amongst the most well sampled groups of viruses studied. In fact, an extra level supported by DEmARC, below the current genus level and called “subgroup”, is used by practicing coronavirologists because of its biological relevance. Although the Linneaen taxonomic structure does not offer extra major levels below the family rank, the availability extra levels above the family rank could be used to improve the correspondence between taxonomy and genome-base classification.

Conclusions

The current taxonomic rank structure of five levels was developed to accommodate an ever-increasing but still relatively limited virus diversity known to us, at a time when virus discovery was mostly an annual event to celebrate. Now viruses are discovered daily if not hourly. The number of known virus species is expected increase from the current thousands to zillions in the future and their classification will be driven by comparative genomics. To accommodate the complexity of phylogenetic relationships apparent within this fast growing diversity, the virus taxonomic rank structure must be adjusted accordingly, as has already become evident from bioinformatics analysis of few better characterized groups of viruses. Using an expanded version of the Linnaean taxonomic structure of ten ranks (eight canonical plus two optional) will contribute to a better description of virus diversity and improve cross-talk between the taxonomies of viruses and their hosts. The number of these ranks could be revisited and, if necessarily, expanded further in the future. Besides the obvious changes to the Virus Code, the formal recognition of the Baltimore classes at the basal level of virus taxonomy could be the first practical steps in this direction.