Members of the family Hepeviridae are positive-stranded RNA viruses with genomes of 6.6 to 7.3 kb. The longest ORF (ORF1) encodes a non-structural protein with several distinct domains: methyltransferase, Y-domain, papain-like protease, polyproline region (also known as the hypervariable region; HVR), macro domain, helicase and RNA-dependent RNA polymerase. ORF1 is followed by ORF2, which encodes the capsid protein, and ORF3, which overlaps with ORF2 and encodes a phosphoprotein that modulates cellular activities.

Despite this conserved genome structure, phylogenetic analysis within the family is complicated by difficulty in aligning the genome sequences of the most divergent variants. In particular, amino acid sequence identities between cutthroat trout virus and other members of the family are only 26–27 % (ORF1), 18–21 % (ORF2) and 13–16 % (ORF3). In contrast, identities between avian, rat and human HEV stand at 42–49 %, 42–55 % and 20–29 %, respectively (Batts et al., 2011). These average figures mask the fact that, in some genomic regions, no credible amino acid sequence alignment is achievable (Drexler et al., 2012; Huang et al., 2004; Johne et al., 2010b; Smith et al., 2013). Hence, phylogenetic comparisons simply based on pairwise (p)-distances between complete genome nucleotide sequences (Fig. 1a) may give a distorted impression of the relationships between variants. In addition, phylogenetic analysis must take into account the fact that substitutions at synonymous sites are saturated even in comparisons between the different genotypes of human HEV (Smith et al., 2013).

To recover more accurately sequence relationships between the most divergent variants in the Hepeviridae, we screened for regions of the genome that are clearly homologous using the Motif Scan program (http://myhits.isb-sib.ch). This identified three subgenomic regions in ORF1 comprising the methyltransferase [ORF1 residues 28 (ORF1-28) to ORF1-389 numbered relative to the sequence of the HEV Burma isolate, GenBank accession M73218], helicase (ORF1-971 to ORF1-1185) and RNA-dependent RNA polymerase (ORF1-1249 to ORF1-1671). Maximum-likelihood analysis of alignments of each region reproduced phylogenetic trees with a similar topology but with much shorter terminal branches (Fig. 1b–d) than obtained from complete genome sequences. Branch lengths in these trees likely represent better reconstructions of evolutionary depth and demonstrate that cutthroat trout virus is substantially more divergent from other members of the family Hepeviridae than indicated by nucleotide sequence comparisons. Similarly, p-distances amongst these subgenomic amino acid sequences form non-overlapping distributions, with the greatest distances observed in comparisons including cutthroat trout virus (Fig. 2). In addition, cutthroat trout virus has an aberrant genome organization (ORF3 is displaced towards the middle of ORF2) and a distinct host range (fish rather than mammals or birds). For these reasons, we recommend that cutthroat trout virus should be assigned to the new genus Piscihepevirus (Table 1), as proposed previously (Meng, 2013), with the single species Piscihepevirus A. We prefer these genus and species names to Cutrovirus (Batts et al., 2011), since the virus appears to be confined to fish, having also been detected in five other species of trout (Batts et al., 2011).

We propose that the remaining HEV variants should be considered as members of the genus Orthohepevirus (Table 1), extending the usage of a previous proposal (Meng, 2013). This change of genus name from Hepevirus is preferred because it makes it clear that not all members of the Hepeviridae belong to the same genus.

Analysis of part of the RdRP domain from an incomplete virus genome sequence (Hepelivirus) obtained from raw sewage suggests that this may represent an additional genus within the family Hepeviridae (Ng et al., 2012). However, a complete genome sequence would be required to confirm the taxonomic position of this virus, for which the host range remains unknown.

The next obvious level of sequence diversity amongst members of the genus Orthohepevirus is that which separates variants originally isolated from different host species (Fig. 3). For example, all isolates from chickens belong to one clade, as do those from rats and ferrets, those from humans, deer, mongooses, pigs, wild boar and camels, while a single complete genome sequence isolated from a bat forms a fourth group. These four groupings are observed regardless, whether the analysis is performed on complete genome nucleotide sequences or subgenomic amino acid sequences (Figs 1 and 3). We propose that these groupings correspond to four species and that they be designated Orthohepevirus A to Orthohepevirus D (Table 1).

Subgenomic sequences of variants isolated from other mammalian species have recently become available but phylogenetic analyses are complicated because they overlap with only relatively short regions. The most complete sequence (5081 nt) is derived from a variant isolated from a moose (Lin et al., 2014). Phylogenetic analysis of the amino acid sequence of this isolate for part of ORF1 and all of ORF2 (Fig. 3) shows that it is distinct from all other variants, although it shares a branch with species Orthohepevirus A in the same way that the ferret variants group with rat-derived species Orthohepevirus C. Similar analysis of virus sequences isolated from mink (Krog et al., 2013) shows that these group with sequences isolated from ferrets. A sequence isolated from a fox (Bodewes et al., 2013) appears to be more distinct (Fig. 3d) and could represent an additional species. However, this last analysis was based on a relatively short subgenomic region (ORF1-1420 to ORF1-1505) in which the grouping of species Orthohepevirus C isolates from rat and ferret, and of the moose variant with Orthohepevirus A variants from human, pig and wild boar, were less distinct than for comparisons based on longer coding regions (Fig. 3). We recommend that assignment of the moose and fox variants to particular Orthohepevirus species should not be made until comparisons based upon complete genome sequences are available.

An advantage of using species names of the form Orthohepevirus A, etc., rather than names based on avian HEV, rat HEV or bat HEV, is that they will not be compromised by current or future discoveries on the extent of host range. For example, variants of species Orthohepevirus C have been isolated from Rattus (Li et al., 2013) and Bandicota species (the greater bandicoot rat) (Li et al., 2013), which are members of the order Rodentia, but also from Asian musk shrews (Guan et al., 2013), which are members of the order Soricomorpha, and from ferrets (Li et al., 2014; Raj et al., 2012) and mink (Krog et al., 2013), which are members of the order Carnivora. Wider screening for members of this species, and also for members of species Orthohepevirus B and Orthohepevirus D, may reveal wider host ranges than those recognized at present.

Although the International Committee on Taxonomy of Viruses does not provide official designations for taxonomic entities below the species level, it is useful for researchers to have an agreed designation of genotypes within each species grouping.

Several studies have attempted to define subgenotypes of genotypes HEV-1, HEV-3 and HEV-4 in the species Orthohepevirus A, in some instances based on the analysis of subgenomic regions (Lu et al., 2006; Zhu et al., 2011). Such categories may form useful labels for epidemiological studies (Dai et al., 2013), but more recent analysis of complete genome sequences suggests that it is not possible to define discrete boundaries that distinguish subgenotypes with consistency (Okamoto, 2007; Oliveira-Filho et al., 2013; Smith et al., 2013). We recommend the approach, commonly adopted in several recent publications, of labelling clades apparent within sequence sets (Dai et al., 2013; Ijaz et al., 2014; Oliveira-Filho et al., 2013) without defining them as permanent classification assignments.

A recurring difficulty in the literature, relating to molecular studies of members of the family Hepeviridae, is that of comparing different studies for which there is no explicit standard sequence with reference to which nucleotides or amino acid residues are numbered. The presence of numerous insertions or deletions and regions of low similarity in alignments of Hepeviridae sequences precludes a unified numbering system that is applicable across all species or genera. We recommend that genome sequences be numbered with reference to the first nucleotide of the prototype complete genome sequence available for each species within the genus Orthohepevirus (Table 1). Nucleotide sites in variants that contain insertions relative to the prototype sequence should be identified with additional letters, beginning at the site of insertion. For example, a three-nucleotide insertion at position 1788 of the prototype sequence would be numbered 1788a, 1788b, 1788c. Insertions of more than 26 nt would be numbered from the twenty-seventh position as 1788aa, 1788ab, etc. and then as 1788ba, 1788bb, etc. as required. This mirrors the system adopted for HCV (Kuiken et al., 2006). Similarly, amino acid residues should be numbered with reference to the first residue of the appropriate ORF from the reference sequence, for example ORF1-929, with an insertion at this site indicated by suffix letters such as ORF1-929a, ORF1-929b, etc., followed if necessary by ORF1-929 aa, ORF1-929ab, etc.

The proposed classification, which assigns a separate hierarchy of genus and species, respects the different levels of divergence between the cutthroat trout virus and all other hepeviruses. The degrees of sequence divergence and conservation of genomic features associated with these categories closely match those attributed to genera and species in other virus families (e.g. Picornaviridae, Caliciviridae and Flaviviridae). This description of relationships among members of the family Hepeviridae will help to resolve current confusion in the literature and, by providing a rational basis for taxonomic assignments, help to reduce the number of future conflicts as more members of this family are discovered.