Early approaches to potyvirus taxonomy

Plant virus taxonomy received a boost in 1959 when Brandes and Wetter [5] proposed that viruses with elongated particles be assigned to six groups based on the size and morphology of their particles. Discrimination between each of these groups, one of which was the potyvirus group, was not much of a problem. Neither was discrimination between members within the potyvirus group as long as there were only a few members. However, over the next three decades, as more and more viruses were assigned to the potyvirus group, it became more and more difficult to determine whether a new isolate was an undescribed distinct virus or a strain of an already described one, and the concept of a ‘continuum’ of variants that bridged sets of strains of distinct viruses was proposed [6, 7] and widely accepted [11, 12]. Indeed Professor Robert Milne summed up the situation when he called on the ghost of Linnaeus to come and resolve the ‘poty problem’ [18].

Thus, by the late 1980s, an effective taxonomy of the potyvirus group did not seem possible owing to its size, complexity and immense variation. Conventional approaches based on physical properties, biological properties and serology had failed to resolve these issues [10, 18]. As Barnett [4] pointed out in his closing chapter to the proceedings of the 1990 Braunschweig Potyvirus Taxonomy Workshop: ‘much of the failure to reach a consensus was due to the inability of the scientists to separate taxonomic characteristics from diagnostic ones’.

Sequence data and the bimodal distribution

In their landmark paper of 1988, Shukla and Ward [21] provided the solution when they used potyvirus coat protein sequence data to discriminate between species and strains. Following on from their studies of the amino acid sequences of a few potyviruses, they analysed the 136 possible pairings of the coat protein amino acid sequences from 17 strains of eight distinct potyviruses in the literature and found a clear-cut bimodal distribution of sequence identity. Distinct members of the group exhibited sequence identity ranging from 38 to 71 (average 54%) with major differences in the length and sequence of their N-termini (which were subsequently shown to be surface located [23] and immunodominant [24–26]) and high sequence identity values in the remainder of the coat protein. In contrast, strains of individual viruses exhibited sequence identity of 90 to 99% (average 95%) and had very similar N-terminal sequences. These findings cast doubt on the currently held ‘continuum’ hypothesis proposed to explain the unsatisfactory taxonomy of the potyvirus group.

The 1990 potyvirus taxonomy workshop in Braunschweig, Germany

Immediately after the 1990 Berlin International Congress of Virology, the Potyvirus Study Group of the Plant Virus Subcommittee of the International Committee on Taxonomy of Viruses held a workshop on potyvirus taxonomy at the Biologische Bundesanstalt in Braunschweig, Federal Republic of Germany. The meeting was attended by 65 participants from 14 countries and ran from 2-4 September 1990. The papers presented at the meeting were published in a 450-page special issue of Archives of Virology [3].

At the end of the meeting, the participants voted to establish the plant virus family Potyviridae and approved three genera and one possible genus. The approved genera were the genus Potyvirus composed of the aphid-borne viruses, with Potato virus Y as the type species; the genus Bymovirus, composed of the fungus-transmitted viruses, with Barley yellow mosaic virus as the type species, and the genus Rymovirus, composed of the mite-borne viruses, with Ryegrass mosaic virus as the type species. The possible genus Ipomovirus was proposed for the whitefly-borne viruses (pending the availability of sequence data), with sweet potato mild mottle virus as the type species. This proposal was approved by the International Committee on Taxonomy of Viruses (ICTV) at its meeting in Glasgow in August 1993.

The major criterion for the assignment of potyvirus isolates to the above taxonomic categories of genus, species and strains was the level of sequence identity of their coat proteins. The sequence data for barley yellow mosaic virus (BaYMV), the type member of the genus Bymovirus, and wheat streak mosaic virus (WSMV), the only member of the genus Rymovirus for which sequence data were available, showed that, when extended to include between-genus comparisons, the bimodal distribution of sequence identity values (species 38-71%, average 54%; strains 90-99%, average 95%) became trimodal, with the genus/genus comparisons providing a more distantly related grouping [1, 30]. For BaYMV (genus Bymovirus), the level of amino acid sequence identity of the whole coat protein to those of 18 members of the genus Potyvirus ranged from 18 to 22%, and for the protease-derived cores, it was 21-25%. The corresponding values for wheat streak mosaic virus (at the time it was assigned to the Rymovirus genus) were 18-24% and 21-27%, respectively [29]. The Bymovirus/Rymovirus comparison was 16% and 19%, respectively [29].

Unfortunately, the observation that these three genera correlated with three different modes of vector transmission subsequently led to the incorrect assumption that vector transmission was the character that defined each genus. If, at the time of the Potyvirus Workshop, vector transmission had been accepted to be the criterion for assignment of a virus to a genus, the proposal to assign the whitefly-transmitted potyviruses to a fourth genus, Ipomovirus, would have been definite, not tentative (as recommended by the workshop members), as there was no doubt about the vector being whiteflies. It only became a definite assignment once sequence data for the type member sweet potato mild mottle virus became available three years later, revealing a low level of sequence identity similar to that observed in pairwise comparisons of members of different genera [2, 8].

Similarly, the validity of other accepted genera, Macluravirus, Brambyvirus and the recently established Poacevirus, has been confirmed by pairwise comparisons of coat protein or whole genome sequences, revealing low sequence identity (15-23%) between members of different genera [2, 9, 28].

Doubts about Rymovirus

Doubts about the taxonomic status of ryegrass mosaic virus (RGMV) arose with the publication of its coat protein sequence [20], which showed low sequence similarity (13% identity) to that of wheat streak mosaic virus, the virus whose sequence had been used to establish the genus Rymovirus, representing the mite-transmitted viruses, but high sequence similarity to potato virus Y (PVY), the type member of the genus Potyvirus (Fig. 1). These [20] and other authors [19] recommended that the taxonomic status of RGMV be reviewed.

Fig. 1
figure 1

Trimodal distribution of pairwise amino acid sequence identities between members of the proposed genera of the family Potyviridae. The background figure for the bimodal distribution of species and strain comparisons in the genus Potyvirus is from reference 21. When viruses of the genera Ipomovirus, Macluravirus, Tritimovirus, Brambyvirus, Bymovirus and Poacevirus are compared with each other and with Potyvirus representatives, a new, more distantly related peak (15-23%) is seen [1, 2, 8, 19, 20, 27] corresponding to distinct genera. As indicated, the Rymovirus-Potyvirus comparisons do not fall into this most distantly related peak, but fall within the limits seen for other Potyvirus-Potyvirus comparisons

Full size image

As a consequence, a new genus, Tritimovirus, was approved to accommodate the WSMV, BrSMV and possibly SCSMV, but the remaining viruses (AgMV, HoMV, ONMV, RGMV and SpMV) were retained as members of a distinct sixth genus (Rymovirus) by the ICTV Potyviridae Study Group. This decision was questioned by Shukla et al. [27]. As shown in Fig. 2, sequence analysis revealed that the members of the family Potyviridae for which sequence data was available distribute into only five genus groupings, not six, with the members of the proposed genus Rymovirus sharing strong sequence similarity with members of the genus Potyvirus. The authors [27] pointed out that sequence identity values no longer correlated with the then division of the family Potyviridae into six genera and suggested that the Rymovirus genus classification is in error. They suggested that AgMV, RGMV, ONMV, HoMV and SpMV should be transferred to the genus Potyvirus and that the genus Rymovirus be removed from the family Potyviridae [27].

Fig. 2
figure 2

Phylogenetic analysis of coat protein amino acid sequences of members of the family Potyviridae [27]. At the time this Figure was prepared sequence data was only available for the six genera shown namely Potyvirus, Macluravirus, Ipomovirus, Tritimovirus, Rymovirus and Bymovirus. There was no data for the Brambyvirus genus or the Poacevirus genus. As shown here, the Abacarus mite-transmitted rymoviruses AgMV, HoMV and RGMV (enclosed in a box) cluster with the aphid-transmitted viruses of the genus Potyvirus, not with the Aceria mite-transmitted tritimoviruses

Full size image

Adams et al. [1] reached the same conclusion from their phylogenetic analysis of the entire open reading frame and each gene set of 187 members of the family Potyviridae, including viruses of the genera Bymovirus, Ipomovirus, Potyvirus, Rymovirus and Tritimovirus.

Revised taxonomy

Based on sequence comparisons between members of the different genera in the family Potyviridae, it is suggested that the taxonomy of the Potyviridae currently described in the 9th Report of the International Committee on Taxonomy of Viruses [13] be revised to include the following genera:

Genus Potyvirus, including the rymovirus subgroup, whose members are transmitted by Abacarus mites [1]

Genus Ipomovirus,

Genus Macluravirus,

Genus Tritimovirus, whose members are transmitted by Aceria mites [1]

Genus Brambivirus

Genus Bymovirus

Genus Poacevirus (a recently established genus not in the 9th Report but on the ICTV website).

Taxonomy reflects genetic relatedness, and until the genetic basis of a biological property such as mode of transmission is understood, the value of that property as a criterion for taxonomic assignments as opposed to identification and diagnosis is questionable. In the past, diagnostic markers such as host range and symptomatology were considered to be valuable taxonomic markers and were used to assign distinct virus species within the genus Potyvirus. However, subsequent comparative sequence data have shown that this conclusion was wrong and that these properties can be affected by accumulated point mutations, revealing that supposedly distinct viruses are in fact closely related strains [15, 16, 31, 32]. Sequence data have also revealed the converse, that some previously assigned strains are in fact members of distinct virus species [14, 17, 21, 22].

Interestingly, the patterns of host reactions of potyviruses on panels of differential plant genotypes resemble the reactivity patterns seen with monoclonal antibodies, suggesting that the former also reflects the impact of point mutations on protein-protein and protein-macromolecule interactions [16]. The potyvirus coat protein is multifunctional and is involved in infection, cross-protection, host specificity, symptomatology and vector transmission. All of these properties can be affected by accumulated point mutations, precluding their use as taxonomic criteria.

The major criterion used by the 1990 Potyvirus Taxonomy Workshop to assign viruses to different genera was relative sequence identity, not the type of transmission vector. This important point seems to have been lost over time.