Known double-stranded DNA viruses infecting hyperthermophilic members of the Archaea, the third domain of life next to the domains Bacteria and Eukarya, form a unique group in the viral world. Due to the unique properties of these viruses, eight novel families have been introduced by the International Committee on Taxonomy of Viruses (ICTV) for their classification: Lipothrixviridae, Rudiviridae, Globuloviridae, Guttaviridae, Fuselloviridae, Bicaudaviridae, Clavaviridae, and Ampullaviridae (for reviews, see refs. 16, 17). Virion morphologies of members of the latter four families are unique: viruses similar to the spindle-shaped fuselloviruses, droplet-shaped guttaviruses, two-tailed bicaudaviruses or bottle-shaped ampullaviruses have never been observed among bacteria or eukarya. Unique for the Archaea are also linear viruses with double-stranded (ds) DNA genomes, which have been assigned to the families Lipothrixviridae and Rudiviridae.

The members of the family Lipothixviridae (from the Greek lipos, “fat” and thrix, “thread”) were the first viruses of hyperthermophilic archaea to be described [10]. Their flexible, linear virions consist of the outer envelope, inner helical core and linear dsDNA genome (Fig. 1A). Following the characterization of lipothrixviruses, an additional group of linear viruses that differed considerably from the members of the family Lipothrixviridae in their virion characteristics was discovered. Unlike lipothrixviruses, these viruses did not contain a lipid envelope and were stiff, rather than flexible (Fig. 1B). Due to these differences, the latter group of viruses was classified into the novel family Rudiviridae (from the Latin rudis, “small rod”) [19]. At present, the family Lipothrixviridae is comprised of nine virus species [2, 4, 8, 10, 22], while the family Rudiviridae includes 4 virus species [19, 23, 24]. All of the members of these families were isolated from extreme geothermal environments in Europe, North America and Asia. Their known host range is restricted to hyperthermophilic archaea from the genera Sulfolobus, Acidianus, Stygiolobus and Thermoproteus (all members of the phylum Crenarchaeota).

Fig. 1
figure 1

Negative-contrast electron micrographs of the representatives of the families Lipothrixviridae and Rudiviridae. (A) Acidianus filamentous virus 1, AFV1; (B) Sulfolobus islandicus rod-shaped virus 2, SIRV2. Scale bars, 200 nm

Besides the apparent differences in virion organization and complexity, members of the two families also differ in the mechanisms they exploit for genome replication. Although genomes of both rudiviruses and lipothrixviruses consist of linear dsDNA molecules, only those of rudiviruses have hairpin ends produced by two terminally linked DNA strands [5]. A rolling hairpin model for the genome replication of rudiviruses was recently proposed based on the structural and biochemical characterization of the REP protein, which is absolutely conserved in all members of the family [13]. The replication appears to be initiated by introducing a nick on one of the genome strands, 11 nucleotides away from the terminus. A protein implicated in this process displays remarkable structural similarity to the rolling-circle replication initiation proteins and was shown to introduce a nick at the predicted replication initiation site, forming a covalent adduct with the newly created 5′ end, with the concomitant release of the 3′-hydroxyl terminus, which is suitable for priming DNA replication [13]. The replication proceeds by strand displacement and results in the generation of head-to-head concatemers [15]. A four-way junction between the concatemers is likely to be subsequently resolved by the viral Holliday-junction-resolving enzyme (Hjr) encoded by all members of the family [3, 6]. The two key enzymes in rudiviral genome replication, REP and Hjr, do not have homologues in any of the isolated lipothrixviruses. Moreover, the genome topology of viruses in the family Lipothrixviridae, with a protein covalently attached to the termini of their linear genomes [16], apparently excludes the rolling hairpin model. Unusual traits of genome replication of the lipothrixvirus AFV1 strongly suggest a novel replication mechanism, distinct from that operating in rudiviruses [16].

Despite the pronounced differences in virion morphology and genome replication mechanisms, accumulation and thorough comparative analysis of complete genome sequences from both rudiviruses and lipothrixviruses has revealed an unexpected evolutionary link between the two families [12, 17, 18]. It has become apparent that, at the genome level, some lipothrixviruses are no more similar to other members of the family Lipothrixviridae than they are to the rudiviruses (Fig. 2). An extreme example is provided by the lipothrixvirus SIFV, which shares the same number of homologous genes (ten) with the lipothrixvirus AFV1 and the rudivirus SIRV1. The two sets of genes (SIFV–AFV1 and SIFV–SIRV1) are not identical, but do overlap (Fig. 2; [12]). The overlapping set of genes includes genes encoding glycosyl transferases, transcription factors and small genes of unknown function. Such a relatively close genetic relationship between viruses belonging to different families is unusual in the archaeal virosphere, where viruses from different families typically share only a few (if any) homologous genes [18]. Rudiviruses and lipothrixviruses do not encode recognizable integrases and are not known to lysogenize their hosts, a phenomenon which could potentially favour gene shuffling between unrelated viruses [12]. Consequently, there is no reason to believe that horizontal gene exchange, which could explain the common gene content of rudiviruses and lipothrixviruses, is more vigorous between these linear viruses than between other hyperthermophilic archaeal viruses. We therefore conclude that the genomic relationship between the two groups of viruses most likely reflects their common ancestry.

Fig. 2
figure 2

Genomic relationship between linear archaeal viruses of the families Rudiviridae and Lipothrixviridae. Genes shared by Sulfolobus islandicus rod-shaped virus 1 (SIRV1; Rudiviridae) and lipothrixviruses Sulfolobus islandicus filamentous virus (SIFV) and Acidianus filamentous virus 1 (AFV1) are shaded blue. Genes restricted to virus pairs SIRV1-SIFV, SIRV1-AFV1 and SIFV-AFV1 are shown in red, yellow and green, respectively. The figure is adapted from ref. [12] (color figure online)

Even more convincing evidence for the evolutionary link between rudiviruses and lipothrixviruses has come from structural analysis of their major capsid proteins (MCPs), which turned out to be highly similar despite their very low sequence similarity (Fig. 3A). The filamentous, enveloped virions of lipothrixviruses are composed of two MCPs, MCP1 and MCP2. The structures of both proteins from the virus AFV1 have been determined by X-ray crystallography [7], revealing the same unique four-helix-bundle fold at the C-termini of both proteins (Fig. 3B). However, due to the differences in their N-terminal regions, the two AFV1 MCPs display a distinct hydrophobicity profile, which allowed the topological model to be proposed. According to this model, the basic MCP1 forms a core around which the genomic dsDNA is wrapped, whereas the MCP2 interacts with the genome with its basic N-terminal region, and the hydrophilic C-terminal domain is embedded in the lipid envelope [7]. The rod-shaped, non-enveloped virions of rudiviruses are composed of one major protein: the highly basic MCP accounts for 99% of virion proteins and associates with the genomic DNA to form the helical body of the virion. Strikingly, the N-terminally truncated MCP from the rudivirus SIRV was found to have the same four-helix bundle topology as the MCPs of the lipothrixvirus AFV1 [20] (Fig. 3). This fact is remarkable considering that there is only 17% sequence identity between the MCP1 of AFV1 and MCP of SIRV (Fig. 3). In maximum-likelihood phylogenetic reconstruction, the MCPs of rudiviruses and lipothrixviruses are robustly segregated into separate clusters (Fig. 3C), suggesting that these MCP genes have evolved within their respective viral genomes for an appreciable period of time without being transferred between members of the two families.

Fig. 3
figure 3

Major capsid proteins (MCP) of linear dsDNA viruses infecting archaea. A. Sequence alignment of the MCP of the rudivirus SIRV (ORF134) with the two MCPs of the lipothrixvirus AFV1. The alignment is coloured according to sequence conservation (BLOSUM62 matrix). B. Comparison of the N-terminally truncated MCP of SIRV with the two MCPs of AFV1. The structures are colored using a rainbow colour gradient from the N-terminus (blue) to the C-terminus (red). C. Maximum-likelihood phylogeny of the MCPs of rudiviruses and lipothrixviruses. For phylogenetic analysis, homologous MCP sequences were collected using PSI-BLAST [1] and aligned using PROMALS3D [14]. MCP sequences of lipothrixviruses AFV2 and TTV1 could not be retrieved using sequence-similarity-based approaches and are therefore not included in this analysis. A maximum-likelihood phylogeny was constructed in MEGA5 [21] using the JTT matrix model (+I, +G [4 categories]). All positions containing gaps were eliminated, and the final dataset consisted of 120 positions. Numbers at the branch points represent bootstrap values (100 replicates). The scale bar represents the number of substitutions per site. The tree was rooted on the branch between rudiviral and lipothrixviral MCPs. GenBank accession numbers: Rudiviridae MCPs (SIRV1 ORF134, NP_666607; SIRV2 ORF134, NP_666560; SRV ORF134, CAQ58456; ARV gp24, YP_001542641), Lipothrixviridae MCP1 (AFV1 ORF132, YP_003749; AFV3 gp34, YP_001604376; AFV6 gp35, YP_001604193; AFV7 gp28, YP_001604252; AFV8 gp30, YP_001604311; AFV9 gp32, YP_001798550; SIFV ORF35, NP_445700) and MCP2 (AFV1 ORF140, YP_003750; AFV3 gp35, YP_001604377; AFV6 gp36, YP_001604194; AFV7 gp29, YP_001604253; AFV8 gp31, YP_001604312; AFV9 gp33, YP_001798551; SIFV ORF36, NP_445701) (color figure online)

Phylogenetic analysis of the MCP sequences supports the current classification of flexible enveloped and rigid non-enveloped linear viruses of archaea into two distinct families. However, the structural similarity between the MCPs of rudiviruses and lipothrixviruses (Fig. 3B) unequivocally points towards the common ancestry of these viruses. Indeed, based on the structural data, it has been envisioned that lipothrixviruses might have evolved from a “simpler” non-enveloped rudivirus-like ancestor [7]. One can suggest a sequence of evolutionary events in which the gene for the single MCP of such putative ancestor has been duplicated and evolved so as to facilitate interactions with a hydrophobic envelope, producing the more complex, lipothrixvirus-like virions.

Importantly, the four-helix bundle topology of the rudiviral and lipothrixviral MCPs has not been observed previously in the MCPs of other known dsDNA viruses [11]. Indeed, when rudiviral and lipothrixviral MCPs were compared to all currently available protein structures using the DALI server [9], the AFV1 MCP2 and SIRV MCP were reciprocally found to be the closest structural relatives, with highly significant Z scores (13.3 and 13.4, respectively). No structural virion proteins from other viruses were identified as similar. Notably, the arrangement of the four helices in the MCPs of rudiviruses and lipothrixviruses is different from that found in the capsid protein of the ssRNA-genome-containing tobacco mosaic virus [7] (Fig. 4), suggesting an independent origin for these archaeal and plant virus capsid proteins. Consequently, structural information not only illuminates the relationship between the two groups of linear archaeal viruses but also highlights their distinctiveness from any other known virus group.

Fig. 4
figure 4

Different arrangement of α-helixes in the four-helix-bundle major capsid proteins of tobacco mosaic virus (TMV; PDB ID:1EI7) and Acidianus filamentous virus 1 (AFV1; PDB ID:3FBL). The structures are colored using a rainbow color gradient from the N-terminus (blue) to the C-terminus (red). The insertion (ins) between the second and the third α-helixes in the TMV protein was omitted for more convenient comparison. The figure is modified from ref. [11] (color figure online)

In summary, recent structural and genomic studies have led to the accumulation of compelling evidence that points to a common ancestry of rudiviruses and lipothrixviruses. Accordingly, in order to acknowledge the evolutionary relationship between linear viruses of the two families, we propose to unify them in the new taxonomic order “Ligamenvirales” (from the Latin ligamen, for “string”, “thread”). The proposal is under consideration by the Executive Committee of the International Committee on Taxonomy of Viruses.