Main

Viruses were discovered at the end of the nineteenth century as filterable agents causing infectious diseases of plants and animals1,2,3,4,5. Subsequently, their pathogenicity and ability to undergo rapid evolutionary change6 has sparked a large body of research, often connected to the so-called ‘microevolution’ of relatively closely related viruses7,8. However, over the last decade, our appreciation of the importance and distribution of viruses has expanded beyond the original parasitic–pathogen model, and now virologists recognize the role of viruses in host regulation and the maintenance of natural ecosystems9. Shotgun metagenomic sequencing has also revealed the presence of a vast variety of viruses in diverse environmental samples and in apparently healthy organisms from all divisions of life10,11,12,13.

To understand the true extent of virus genomic diversity—which may be significantly broader than that of their hosts—and the origins and forces that shape this diversity, virologists will have to systematically rationalize the more distant relationships between viruses, ideally reflecting their ‘macroevolution’, and virus taxonomy should provide an inclusive yet dynamic classification framework to reflect these relationships. In contrast to the taxonomies of cellular organisms, this new virus taxonomic framework will have to accommodate the current view that viruses have multiple origins (polyphyly) and that their diversity cannot be represented by a single virosphere-wide tree14.

The traditional five-rank structure of virus taxonomy

The International Committee on Taxonomy of Viruses (ICTV) oversees the official classification of viruses and nomenclature of taxa, that is, taxonomy (Box 1)15. In its earliest versions, the ICTV classification of viruses into taxa formally recognized only genera and families but, over time, this classification scheme developed into a five-rank hierarchy of species, genus, subfamily (used rarely), family and order16,17. This five-rank structure matched a section of the Linnaean hierarchical structure used in the taxonomies of cellular organisms and remained in place until 2017 (Fig. 1, left). In addition to changes in the rank hierarchy, the recognition of virus taxa has also evolved over time, from a traditional phenotype-based characterization process to a multistage process that increasingly, but not exclusively, includes genomic properties and sequences (Box 2)15,18. Nowadays, formal virus classification emphasizes comparative sequence analyses of conserved genes and proteins, including gene phylogeny, gene synteny and shared gene content. Other molecular traits are also considered when appropriate19,20.

Fig. 1: A comparison of the ICTV taxonomic rank hierarchy in 1991–2017 and 2019.
figure 1

Taxonomic ranks are shown in relation to the distribution pattern of taxa. The number of taxa assigned to each rank (as recorded in the current ICTV Master Species List, release 2018b, MSL34 (ref. 47)) are shown in white font on the 15-rank structure. When the ranks are described as a hierarchy, the species rank is often referred to as the lowest rank and the realm rank as the highest rank. However, when the ranks are used as phylogenetic terms, the realm rank can be described as basal and the species rank as apical or terminal. Both conventions are used in this Consensus Statement. Black arrows, ranks common to the five- and 15-rank structure; pink arrows, ranks introduced in the 15-rank structure.

Classifications outside of the ICTV taxonomic remit

Until recently, the evolutionary relationships between viruses of different families or orders were considered by the ICTV, and by many in the virology community, as being too distant to be resolved in a credible classification. Thus, there was little impetus to extend the taxonomy rank structure. The result was a taxonomy that, in profound contrast to its cellular counterparts, included many disjointed taxa, the number of which increases with the accelerating discovery of novel viruses (exceeding 100 families in 2018). However, classification efforts continued outside of the official taxonomic framework and, over the last few decades, several informal groupings such as ‘supergroups’ or ‘superfamilies’ were proposed for subsets of RNA viruses21,22,23 and DNA viruses24,25,26,27,28. These groupings relate to otherwise seemingly disparate viruses belonging to different families and have a variety of different hosts, genome types and organizations, and replication mechanisms. Importantly, these groupings have relied on distant relationships often associated with structure–function hypotheses (for example, an essential virus protein involved in virus replication or virion morphogenesis), which were then validated in subsequent experimental studies29,30,31; these provided independent support for the inferred classifications.

Also, before these developments took place, Baltimore had introduced a non-hierarchical classification of viruses which groups viruses into just seven (originally six) classes according to their genome type (double-stranded DNA, single-stranded DNA, double-stranded RNA, positive-sense RNA, negative-sense RNA, reverse-transcribing RNA and reverse-transcribing DNA) and its relation to the synthesis of mRNA32,33. Because of its conceptual clarity and functional foundation, this classification system is still widely used. It complements virus taxonomy by grouping viruses into meaningful classes at a different scale of virus divergence, albeit without attempting to evaluate their evolutionary relationships.

The new Linnaean-like ranking hierarchy of virus taxonomy

In 2016, the urgency, timeliness and logistics of introducing additional ranks to the virus taxonomy hierarchy were discussed at length by the ICTV Executive Committee (EC). The discussion addressed how best to mirror the complete Linnaean taxonomy system (based on a nest of seven principal or primary ranks: species, genus, family, order, class, phylum and kingdom), how to allow for the hierarchical clustering of virus taxa in higher ranks such as orders, and whether the Baltimore classes might be adopted as taxa, perhaps at the basal ranks of the taxonomy34. Figuratively speaking, a taxonomic hierarchy was sought that could accommodate a virosphere-wide tree (or trees) from the roots to the tips of the branches. Because of its potential impact on the practice of virus taxonomy, the EC created a Working Group to consider the matter in more detail. An account of the process undertaken by the Working Group to propose a new taxonomy is outlined in Box 3.

The Working Group concluded that an extended, formal virus classification hierarchy should provide 15 ranks, including eight principal (or primary) ranks and seven derivative (or secondary) ranks (Fig. 1). The eight principal ranks include four that were already in use (order, family, genus and species) and four that are new: realm, kingdom, phylum and class, which are all above the order rank. The class rank in this series is not to be confused with the ‘classes’ described by Baltimore, or the typological attributes of a taxonomic rank35. These new principal ranks cover the entire scale of virus divergence to include the deepest virus relationships at the basal rank of realm. The large scale of virus divergence encompassed by the 15 ranks is exemplified by the newly created Riboviria taxon (a realm) that currently includes all RNA viruses encoding an RNA-directed RNA polymerase, including viruses of three Baltimore classes (III, IV and V)36.

The seven secondary ranks include the previously used subfamily rank and six new ranks that are derivatives of most of the remaining principal ranks. The exception is the species rank, which is currently not associated with a secondary rank, as no consensus on the definition of ‘subspecies’ could be reached. This new rank hierarchy and the associated nomenclature (Fig. 1, right), including defined suffixes for taxa, follow those used in the Linnaean system with a single exception. The basal rank is called ‘realm’ in virus taxonomy, rather than ‘domain’ (as in other taxonomies), reflecting a complex interrelation between virus taxonomy and its counterparts for cellular organisms.

The new rank hierarchy and its normative basis, in the form of changes to the ICTV Code, were approved by the ICTV EC and subsequently ratified by the ICTV in two votes in 2018 and 2019 (refs. 37,38). These changes provide the virology community with the opportunity to submit taxonomic proposals that fill the new principal and secondary ranks with defined taxa.

Application and impact of the extended virus taxonomy

When developing taxonomy, virologists are only obliged to assign a (new) virus to taxa at genus and species ranks. Other ranks may be used optionally when the scientific justification is sufficient. The new 15-rank taxonomic structure applies to all viruses, although none have been assigned to all ranks thus far. To illustrate the application of some of the newly established ranks, Fig. 2 and Table 1 show the full current classification and taxa demarcation criteria of two viruses with RNA genomes, Ebola virus (EBOV) and severe acute respiratory syndrome coronavirus (SARS-CoV). Both viruses are well-known human pathogens and members of the species Zaire ebolavirus and Severe acute respiratory syndrome-related coronavirus, respectively. Although both of these viruses infect humans, they differ taxonomically in the ranks that are populated and the demarcation criteria that define the taxa. Only at the basal, realm rank are the two viruses included in the same taxon, Riboviria. Figure 2 also shows the corresponding information for another human pathogen, herpes simplex virus 1, which has a double-stranded DNA genome and is currently assigned to taxa of five ‘traditional’ ranks.

Fig. 2: Classification of EBOV, SARS-CoV and herpes simplex virus 1 in the 15-rank taxonomic hierarchy.
figure 2

Intra-cluster virus divergence, which increases from the virus to the species rank to the realm rank, is represented by the increasing width of the respective rectangles, which are not drawn to scale. EBOV is most closely related to, but distinct from, Bombali, Bundibugyo, Reston, Sudan and Taï Forest viruses, which belong to separate species included in the Ebolavirus genus. SARS-CoV is one of several closely related coronaviruses isolated from humans and animals, such as palm civets and bats, and are included in the species Severe acute respiratory syndrome-related coronavirus. Herpes simplex virus 1 is one of two human herpesviruses belonging to different species in the Simplexvirus genus. Ranks that were introduced with the extended rank structure are indicated by an asterisk.

Table 1 Classification of EBOV and SARS-CoV in the 15-rank taxonomic hierarchy

A surge of activity to populate the existing and additional ranks of virus taxonomy can now be expected and is to be encouraged. This will include the consideration of numerous supergroups and superfamilies that have remained outside of virus taxonomy to date. At the time of writing (January 2020), the ranks included highly different numbers of taxa (Fig. 1). As expected, the species rank, with the lowest degree of intra-taxon virus divergence, is the most heavily populated, and ranks with higher degrees of intra-taxon divergence are the least populated. This highly uneven distribution of the number of taxa assigned at different ranks is due not only to the hierarchical relationship of the ranks, but also to the very recent introduction of the more basal ranks, and the lack of requirement to fill these ranks. The uneven distribution also reflects variations in the sampling of different virus lineages in diverse hosts as well as differences in the approaches adopted for the recognition of taxa in particular species by different ICTV Study Groups.

These variations could persist, although the currently observed differences in taxon density may be partially alleviated when researchers define the more distant taxonomic relationships among viruses and improve their resolution by involving traditional and new evolutionary methodologies, such as network analyses28. For example, increasing evidence supports an ancestral relationship of some viruses of the order Caudovirales (a group of bacterial viruses with double-stranded DNA genomes) and viruses of the order Herpesvirales (a group of animal viruses with double-stranded DNA genomes) through a shared virion morphogenesis module. This module includes the HK97-type major capsid protein, portal protein, capsid maturation protease and the genome-packaging terminase complex25,39,40,41, and appears to reflect monophyletic relationships that may warrant taxonomic recognition.

As a result of the change to the number and scope of ranks, virus taxonomy is now, for the first time, able to accommodate taxa at any level of virus divergence between the very narrow (species) and the extremely wide (realms). How these ranks are used will depend on the research community, including the ICTV. We stress that the validity of any established taxa, or those created in the future, depends on the strength of scientific evidence to support the demarcation and ranking of taxa, which is considered on a case-to-case basis.

The codified availability of a greater number of ranks in a formal virus classification that emulates a Linnaean framework may also facilitate the comparison, and possibly improve the compatibility of virus taxonomy with the taxonomies of cellular organisms. Although the switching of hosts by viruses may be a complicating factor, the availability of fossils and a defined evolutionary timescale for some virus hosts should benefit virus taxonomy. Such information will be essential for taxa demarcation and rank definition in the future, notwithstanding that all taxonomies depend on the accuracy of evolutionary reconstructions, which are most challenging for distant relationships that reflect numerous changes, including those resulting from horizontal (lateral) gene transfer.

We expect that the described changes to the hierarchical rank structure will create a new impetus for the exploration of virus macroevolution and a framework for its application to taxonomy. The changes will also stimulate research on the defining characteristics of monophyletic virus lineages and the recognition of historical events that played a decisive role in their origins and evolution. These events may be comparable to major transitions in the evolution of cellular life forms, such as the origins of eukaryotes or plants. This information could be used to define taxa and ranks, and the improved interaction with evolutionary research will facilitate the main mission of virus taxonomy, which is to serve the virology community and the public at large in a comprehensive, scientific manner.