Main

In the past few years the importance of virology for understanding fundamental aspects of biological evolution has grown. In particular, RNA viruses might hold clues to the origin of genetic systems, being, possibly, the living relics of the ancient RNA world that is widely believed to predate the extant DNA-based genetic cycle of cellular organisms1,2. On a more practical level, knowledge of RNA virus evolution is indispensable for unravelling the origins of devastating emergent diseases such as AIDS, severe acute respiratory syndrome and haemorrhagic Ebola fever3. In addition, metagenomic research has revealed an enormous diversity of DNA and RNA viruses in the environment and has shown that, at least in marine habitats, viruses are the most abundant biological entities, with as many as 10 virus particles per cell4,5,6,7. Because most marine viruses kill host cells, they substantially contribute to the global carbon cycle7.

Several complementary developments have led to a dramatic expansion of the explored part of the 'virosphere'. The most conspicuous discoveries include many unusual archaeal viruses8,9,10, large phycodnaviruses infecting green algae and stramenopiles11,12, insect polydnaviruses13 and the giant mimivirus14,15. In addition, bacteriophage genomics has uncovered enormous, unanticipated diversity of this part of the viral world16,17,18,19,20,21. Evolutionary genomic analysis of the rapidly growing collection of viral genomes has revealed both deep unity, as exemplified by the demonstration of the common ancestry of diverse families of large DNA viruses of eukaryotes22, and the enormous variability of genome content, for example, in the case of archaeal viruses for which common origins are typically not traceable8,10.

In parallel, there has been a resurgence of interest in viruses and virus-like selfish genetic elements as major players in the origin and evolution of cellular life23,24,25,26,27,28,29. Two concepts of ancient origin and early evolution of viruses have been proposed, both emphasizing the tight connections between the evolution of viruses and cells25,28. One concept expounds the 'three RNA cells' scenario, according to which RNA viruses 'invented' DNA and introduced it, complete with the replication machinery, into putative primordial RNA cells that are envisaged to have been ancestors of each of the three domains of extant life25. The second, 'virus world' concept, based primarily on the mounting evidence from comparative genomics, posits that both RNA and DNA viruses evolved from primordial genetic systems that existed before the emergence of fully fledged cells, and that the large DNA genomes of the first cellular life forms evolved by accretion of virus-like and plasmid-like DNA replicons28. The virus world model also suggests that the major classes of viruses of eukaryotes evolved through mixing and matching the genes that were derived from prokaryotic viruses, plasmids and chromosomes at the time of eukaryogenesis. The collective result of these developments is a new landscape of data, models and ideas that calls for rewriting the fundamentals of virology27,28,30.

A long-standing enigma in virology is the non-uniform distribution of the major classes of RNA, DNA and retroid viruses among the branches of host organisms28. For instance, vertebrates can be infected by all classes of viruses, whereas green plants do not seem to be infected with retroid RNA viruses or true (non-pararetro) double-stranded (ds) DNA viruses31,32. Even more intriguing are the disparities between the abundance and diversity of positive-strand RNA viruses in plants and animals33, the extreme paucity of such viruses in bacteria34,35, and their apparent absence in archaea8,10 (M. Young, personal communication). These striking but largely unexplained patterns of virus distribution suggest that tight connections exist between major evolutionary transitions in the history of life and the global ecology of viruses. Understanding these connections is essential for the development of a general picture of the evolution of viruses and cells.

The current view of evolution of viruses and their host ranges derives primarily from studies on a few model organisms, such as mammals, birds, green plants (mostly cultured), and, to a lesser extent, insects, fungi and several groups of well-characterized bacteria. Until recently, there has been almost no research on viruses that infect the diverse groups of unicellular eukaryotes. However, this has changed as viruses have recently been isolated from a variety of marine eukaryotes such as algae and dinoflagellates7. These studies have resulted in the identification and sequencing of many positive-strand RNA viruses, which has dramatically increased the size and diversity of this virus class36,37,38,39 (see Supplementary information S1 (table)). In addition, several RNA viruses have been identified and sequenced as a result of metagenomic studies40,41,42,43.

In this Analysis article, we exploit the growing collection of diverse viral genome sequences that infect a wide range of eukaryotes to carry out a genomic comparison and phylogenetic analysis of a major division of eukaryotic positive-strand RNA viruses, the picorna-like superfamily, in an attempt to shed light on the early stages of its evolution. We conclude that the diverse groups of picorna-like viruses probably evolved in a Big Bang that antedated the radiation of the five supergroups of eukaryotes. Our analysis provides independent evidence in support of the concept of the major transitions in the history of life as explosive, non-linear events44 and suggests that the Big Bangs of host organism evolution trigger concomitant bursts of viral evolution.

The extended picornavirus-like superfamily

There seems to be an inherent paradox about the evolution of RNA viruses in general and picornaviruses in particular. RNA replication is extremely error-prone, especially in picornaviruses, with a mutation rate that is high enough to maintain a broad quasispecies distribution of RNA sequences and push the viruses to the brink of a mutational meltdown or error catastrophe45,46,47,48. Moreover, it has been shown that the distribution of variants in a quasispecies is not a biologically irrelevant consequence of error-prone replication but rather a crucial factor of viral evolution. The interaction of variants within a quasispecies ensures the adaptability of viruses in changing environments and, in particular, substantially contributes to viral pathogenesis48,49,50. Nevertheless, there is readily detectable conservation of protein domain sequences among viruses that infect diverse hosts and have widely different structures and reproduction strategies. As pointed out by Biebricher and Eigen, RNA viruses “operate close to the error threshold that allows maximum exploration of sequence space while conserving the information content of the genotype”51. However, it seems that the functional constraints on the viral proteins that have key functions in reproduction are strong enough to maintain the alignment of the sequences of the respective domains over a broad range of viral groups, in spite of the mutational pressure. This allows deep phylogenetic analyses52.

In early comparative genomic analyses, positive-strand RNA viruses of eukaryotes were classified into three superfamilies: picorna-like, alpha-like and flavi-like33,53,54. These three superfamilies include most known positive-strand RNA viruses, although the classification of nidoviruses and RNA bacteriophages remained uncertain. The superfamilies were delineated through a combination of phylogenetic analysis of conserved protein sequences, primarily those of RNA-dependent RNA polymerases (RdRps)55, and comparison of diagnostic features of genome organization that are linked to replication and expression strategies. Phylogenetic analysis of RNA viruses at the level of superfamilies is difficult owing to their deep divergence and the high rate of sequence evolution, so it has been argued that the phylogenetic signal contained in the RdRp sequences might be insufficient to define the superfamilies56. Nevertheless, the core subsets of each superfamily were readily identified by straightforward sequence comparison and phylogenetic analyses, and the existence of signature arrangements of conserved genes clinches the case for the objective existence of the superfamilies57,58.

The picornavirus-like superfamily, in particular, is characterized by a partially conserved set of genes that consists of the RdRp, a chymotrypsin-like protease (3CPro, named after the picornavirus 3C protease), a superfamily 3 helicase (S3H) and a genome-linked protein (viral protein, genome-linked, VPg) (Fig. 1; Supplementary information S1 (table)). This set of four genes can be considered to be a signature of the picorna-like superfamily because these genes are not found in other characterized RNA viruses (with the exception of the distinct 3CPro-like proteases of nidoviruses59). Furthermore, most of the viruses in the picorna-like superfamily have icosahedral virions that are composed of capsid proteins with the characteristic jelly-roll fold (jelly-roll capsid protein, JRC). It has to be emphasized that the presence of all four signature genes is not an absolute requirement for classifying a virus as a member of the picorna-like superfamily. In some of the viruses included in the superfamily this genomic layout (bauplan) is incomplete or substantially altered (Fig. 1) but there is additional, strong evidence of their evolutionary relationship to picorna-like viruses. For example, astroviruses have no helicase, whereas nodaviruses lack the helicase, the protease and the VPg (Fig. 1). However, even in the case of the nodaviruses, a connection to the picornavirus superfamily seems convincing thanks to the presence of characteristic motifs and the overall sequence conservation of the RdRp33,55,60.

Figure 1: The genome layouts in the main evolutionary lineages (clades) of picorna-like viruses.
figure 1

The boxes and lines represent open reading frames (ORFs) and non-coding sequences, respectively, roughly to scale. The signature proteins of the picorna-like superfamily are RdRp (RNA-dependent RNA polymerase, S3H (superfamily 3 helicase), CPro or SPro (chymotrypsin-like cysteine or serine proteases), JRC (jelly-roll capsid protein, three structural subsets of which are P for picornavirus-like (P)68, sobemovirus-like (S)119 and nodavirus-like (N)120), and VPg (viral protein, genome-linked; denoted g). −1FS, minus one frameshift; 2APro, 2A protease; 3CPro, 3C protease; 3DRdRp, 3D RdRp; An, poly(A) sequence; BDRM, Bryopsis cinicola dsRNA replicon from mitochondria; CI, cylindrical inclusion protein; CPF, capsid protein, filamentous capsid; CPMV, cowpea mosaic virus; CPU, capsid protein, unknown evolutionary origin; CPV, Cryptosporidium parvum virus; CrPV, cricket paralysis virus; FHV, flock house virus; GLV, Giardia lamblia virus; HAstV1, human astrovirus 1; HCPro, helper component proteinase; IRES, internal ribosome entry site; MP, movement protein; NIb, nuclear inclusion b protein; NoroV, norovirus; ORF, open reading frame; P1Pro, protein 1 proteinase; P3, protein 3; p6, 6 kDa protein; PV, poliovirus; S2H, superfamily 2 helicase; sg, subgenomic RNA promoter; SBMV, southern bean mosaic virus; SssRNAV, Schizochytrium single-stranded RNA virus; TEV, tobacco etch virus; VP, virion protein.

Full size image

We carried out additional sequence analysis in order to validate and update the roster of viruses in the picorna-like superfamily. To this end, we defined the core of the superfamily to include all viruses that contain the 'picorna-like' RdRp and one (3CPro) or two (3CPro and S3H) of the additional signature genes. The amino acid sequence alignment of the RdRps of the viruses that comprise this core was used to generate a position-specific scoring matrix (PSSM), which was screened against the National Center for Biotechnology Information's RefSeq database in order to identify potential additional members of the picorna-like superfamily. This analysis confirmed that the RdRps of nodaviruses had highly significant and specific similarity to those of the picorna-like viruses (Supplementary information S1, S2 (table,figure); the original outputs of the PSSM searches are available on request). Notably, and in accord with the previous conclusions on the multiple originations of dsRNA viruses from positive-strand RNA viruses61,62,63,64, we found that the RdRps of two distinct families of dsRNA viruses, Partitiviridae and Totiviridae, also seemed to be related to the picorna-like superfamily (Fig. 1; Supplementary information S1 (table)).

Genome analysis of the recently isolated positive-strand RNA viruses of unicellular eukaryotes yielded an unexpected result. All four of these viruses, which infect taxonomically diverse hosts, belong to the picorna-like superfamily according to the criteria outlined above, namely, the (partial) conservation of the picorna-type set of signature genes and specific sequence conservation of at least some of the proteins encoded by these signature genes36,37,38,39, for example, Schizochytrium ssRNA virus (Fig. 1). Metagenomic analyses also revealed an apparent prevalence of picorna-like viruses among marine RNA viruses (hosts are unknown)41,42. The current sampling of the diversity of eukaryotic viruses is not sufficient to conclude whether this is a true reflection of the host ranges of the superfamilies of eukaryotic ssRNA viruses or an unrecognized bias in sequencing studies. This uncertainty notwithstanding, identification of RNA viruses in unicellular eukaryotes has led to a notable expansion of the picorna-like superfamily. Remarkably, this superfamily is now represented in four of the five supergroups of eukaryotes65,66, namely Unikonta (including animals, fungi and Amoebozoa), Plantae (land plants, and green and red algae), Chromalveolata (for example, apicomplexa, dinoflagellates, diatoms and oomycetes) and Excavata (for example, kinetoplastids, trichomonads and diplomonads such as Giardia lamblia) (Fig. 2). By contrast, the alpha-like and flavi-like superfamilies of positive-strand RNA viruses have so far only been detected in unikonts (primarily, animals) and plants, with only two known exceptions41,67.

Figure 2: The host ranges of picorna-like viruses.
figure 2

This simplified evolutionary tree of eukaryotes represents five supergroups, the relationships between which remain unresolved65,66. Black lines and names correspond to evolutionary lineages for which no picorna-like viruses have been described so far, whereas coloured lines correspond to lineages known to be infected by picorna-like viruses (named in adjacent coloured boxes). APV, Acyrthosiphon pisum virus; CPV, Cryptosporidium parvum virus; HcRNAV, Heterocapsa circularisquama RNA virus; KFV, kelp fly virus; NrV, Nora virus; RsRNA, Rhizosolenia setigera RNA virus; SIV, Solenopsis invicta virus-2; SmVA, Sclerophtora macrospora virus A; SmVB, Sclerophtora macrospora virus B; SssRNAV, Schizochytrium single-stranded RNA virus.

Full size image

The extended picorna-like superfamily of positive-strand RNA viruses identified here includes the recently proposed order Picornavirales68, which has five families and three floating genera, along with an additional nine families, one genus and 15 unclassified viruses. It includes extremely diverse viruses and virus-like elements, many of which do not closely resemble picornaviruses. As discussed previously28, the notion of monophyly has limited applicability when broad groups of viruses are considered, given the important roles of gene sampling and recombination in the evolution of viruses (as captured, in particular, in the concept of reticulate evolution of bacteriophages69). Nevertheless, we believe that the picorna-like superfamily as described here is a valid group based on current sequence resources, although changes, especially expansion, will undoubtedly result from future analyses. New developments in the taxonomy of the picorna-like viruses should also be expected (see International Committee on Taxonomy of Viruses).

Here we refrain from further discussion of taxonomy and focus on the evolution of picorna-like viruses, with the aim of clarifying the phylogenetic positions of new viruses of unicellular eukaryotes, superimposing the evolutionary trees of viruses and hosts and, hopefully, gaining new insights into the original diversification of viruses of eukaryotes.

Phylogenies of RdRps and helicases

Only two proteins that are encoded in most picorna-like viruses show sequence conservation that is sufficient to obtain resolved phylogenetic trees: the RdRp and the S3H. Multiple alignments of these proteins (Supplementary information S2, S3 (figures) for RdRp and S3H, respectively) were used for maximum-likelihood phylogenetic analysis (Fig. 3). The RdRp tree consists of six strongly or moderately supported major clades that form a star phylogeny with short, apparently unresolvable internal branches (Fig. 3). The clades are as follows, roughly in the order of the decreasing diversity of viruses and hosts.

Figure 3: The phylogenetic tree of the RNA-dependent RNA polymerases of picorna-like viruses.
figure 3

All the sequences of viral genomes and encoded proteins were from GenBank. Viral lineages are colour-coded to reflect their host range. Multiple alignments of protein sequences were constructed using the MUSCLE program121, with subsequent manual adjustment using the corresponding crystal structures. Maximum likelihood trees were constructed using the TREEFINDER program122, with the Whelan And Goldman (WAG123) evolutionary model with γ-distributed site rates. The obtained trees were then used to initialize Monte Carlo Markov Chain (MCMC) computations using the MrBayes program124. For two runs of four MCMCs, 1.1 × 106 generations were retained, with the first 105 generations discarded as burn-in. We pooled together 1,000 samples taken every 1,000 generations from each of the runs, and constructed consensus trees. Support values (fraction of sampled trees with the given tree bipartition present) are indicated for selected clades. The extremely short, apparently unresolvable internal branches of both trees indicative of star phylogeny are best compatible with the rapid diversification of picorna-like viruses in a Big Bang-type event44. AhV, Atkinsonella hypoxylon virus; ALSV, apple latent spherical virus; ANV, avian nephritis virus; APV, Acyrthosiphon pisum virus; BAYMV, barley yellow mosaic virus; BDRC, Bryopsis cinicola dsRNA replicon from chloroplasts; BDRM, Bryopsis cinicola dsRNA replicon from mitochondria; BWYV, beet western yellows virus; CHV, cryphonectria parasitica hypovirus; CPMV, cowpea mosaic virus; CPV, Cryptosporidium parvum virus; CrPV, cricket paralysis virus; DCV, Drosophila C virus; DWV, deformed wing virus; EMCV, Encephalomyocarditis virus; FCCV, Fragaria chiloensis cryptic virus; FCV, feline calicivirus; FgV, Fusarium graminearum virus; FHV, flock house virus; FMDV, foot-and-mouth disease virus; GFLV, grapevine fanleaf virus; GLV, Giardia lamblia virus; HaRNAV, Heterosigma akashiwo RNA virus; HAstV1, human astrovirus 1; HAV, hepatitis A virus; HcRNAV, Heterocapsa circularisquama RNA virus; HRV1A, human rhinovurus 1A; IFV, infectious flacherie virus; JP, Jericho pier; KFV, kelp fly virus; LRV, leishmania RNA virus 1-1; LTSV, lucerne transient streak virus; MBV, mushroom bacilliform virus; NoroV, norovirus; NoV, Nodamura virus; NrV, Nora virus; OPV, Ophiostoma himal-ulmi virus 1; PEMV-1, pea enation mosaic virus-1; PLRV, potato leafroll virus; PnPV, Perina nuda picorna-like virus; PV, poliovirus; PYFV, parsnip yellow fleck virus; RasR1, Raphanus sativus dsRNA 1; RHDV, rabbit haemorrhagic disease virus; RsRNA, Rhizosolenia setigera RNA virus; RTSV, rice tungro spherical virus; SAstV1, sheep astrovirus-1; SBMV, southern bean mosaic virus; SCPMV, southern cowpea mosaic virus; ScV, Saccharomyces cerevisiae virus L-A; SDV, satsuma dwarf virus; SIV, Solenopsis invicta virus-2; SJNNV, striped jack nervous necrosis virus; SmVA, Sclerophtora macrospora virus A; SmVB, Sclerophtora macrospora virus B; SPMMV, sweet potato mild mottle virus; SssRNAV, Schizochytrium single-stranded RNA virus; SV, Sapporo virus; TAstV1, turkey astrovirus-1; TEV, tobacco etch virus; TRSV, tobacco ringspot virus; TrV, Triatoma virus; TSV, Taura syndrome virus; TVV, Trichomonas vaginalis virus; WSMV, wheat streak mosaic virus.

Full size image

Comovirus and dicistrovirus clade (clade 1 in Fig. 3 ). This group has the greatest diversity and includes viruses that infect host organisms of three eukaryotic supergroups: Plantae, Unikonta and Chromalveolata. There are three distinct subclades: the comovirus lineage, which encompasses a variety of plant viruses; the dicistrovirus and marnavirus lineage, which is an assemblage of insect viruses70, recently isolated viruses infecting marine chromalveolates36,38,39, and closely related marine viruses with unknown hosts42; and the third lineage, consisting of iflaviruses and other insect viruses70,71,72.

Sobemovirus and nodavirus clade (clade 2). This clade is only moderately supported. However, it consists of two definitively supported subclades, each of which combines viruses infecting hosts from three (sobemovirus lineage: Plantae, Fungi73 and Chromalveolata37,74) or two (nodavirus lineage: opisthokonts60 and Chromalveolata75) eukaryotic supergroups.

Astrovirus and potyvirus clade (clade 3). This strongly supported clade unites animal astroviruses76, plant potyviruses and dsRNA hypoviruses. dsRNA hypoviruses infect fungal pathogens of plants and have been proposed to have evolved from potyviruses77,78,79,80. Although specific sequence similarities between astrovirus and potyvirus RdRps have been noticed previously81, the recent expansion in the number of relevant sequenced viruses allows confident validation of this clade.

Calicivirus and totivirus clade (clade 4). This is an unexpected but strongly supported unification of a distinct family of animal viruses, the caliciviruses82, with the dsRNA totiviruses, which have been isolated from several diverse excavates and fungi83,84.

Partitivirus clade (clade 5). This clade contains dsRNA viruses of plants, fungi83 and an apicomplexan (which is a chromalveolate)85. Some of the partitivirus-related genetic RNA elements do not have capsids and replicate in the mitochondria or chloroplasts of green algae86.

Picornavirus clade (clade 6). This is the only monotypic group that consists entirely of the Picornaviridae family of vertebrate viruses68 allied with a solitary insect virus87.

Phylogenetic analysis

Strikingly, five of the six major clades of picorna-like virus RdRps include viruses whose hosts belong to two or three eukaryotic supergroups. Evolution of viruses cannot be reduced to the evolution of their RdRps. However, RdRp is the only universal protein in the picorna-like superfamily, so in this Analysis we use the RdRp tree as a standard against which to compare trees and distributions of other genes.

Phylogenetic analysis of RNA helicases (S3H) of picorna-like viruses is more limited in scope than the RdRp analysis because viruses in three of the six RdRp clades do not encode this protein (Figs 1, 3). The S3H tree consists of four well-supported clades (Fig. 4). The largest and most diverse clade mainly corresponds to RdRp clade 1. However, there are notable exceptions: dicistroviruses fall outside the clade and form a lineage of their own; the S3Hs of two insect viruses (kelp fly virus and Acyrthosiphon pisum virus) belong to the calicivirus clade; and the S3H of another insect virus (nora virus) belongs to the picornavirus clade (Fig. 4). Although artefacts of tree topology cannot be ruled out, the respective clades are well supported, so these limited discrepancies between the phylogenies of the RdRp and the S3H of picorna-like viruses suggest the possibility of multiple recombination events during viral evolution.

Figure 4: The phylogenetic tree of superfamily 3 helicases of picorna-like viruses.
figure 4

Note that only a subset of viruses in the picorna-like superfamily encode a superfamily 3 helicase. All the sequences of viral genomes and encoded proteins were from GenBank. Viral lineages are colour-coded to reflect their host range. Multiple alignments of protein sequences were constructed using the MUSCLE program121, with subsequent manual adjustment using the corresponding crystal structures. Maximum likelihood trees were constructed using the TREEFINDER program122, with the Whelan And Goldman (WAG123) evolutionary model with γ-distributed site rates. The obtained trees were then used to initialize Monte Carlo Markov Chain (MCMC) computations using the MrBayes program124. For two runs of four MCMCs, 1.1 × 106 generations were retained, with the first 105 generations discarded as burn-in. We pooled together 1,000 samples taken every 1,000 generations from each of the runs, and constructed consensus trees. Support values (fractions of sampled trees with the given tree bipartition present) are indicated for selected clades. The short, apparently unresolvable internal branches of both trees that are indicative of star phylogeny are best compatible with the rapid diversification of picorna-like viruses in a Big Bang-type event. ALSV, apple latent spherical virus; APV, Acyrthosiphon pisum virus; CPMV, cowpea mosaic virus; CRPV, cricket paralysis virus; DCV, Drosophila C virus; DWV, deformed wing virus; EMCV, Encephalomyocarditis virus; FCV, feline calicivirus; FMDV, foot-and-mouth disease virus; GFLV, grapevine fanleaf virus; HaRNAV, Heterosigma akashiwo RNA virus; HAV, hepatitis A virus; HRV1A, human rhinovurus 1A; IFV, infectious flacherie virus; JP, Jericho pier; KFV, kelp fly virus; NoroV, norovirus; NrV, Nora virus; PnPV, Perina nuda picorna-like virus; PV, poliovirus; PYFV, parsnip yellow fleck virus; RHDV, rabbit haemorrhagic disease virus; RsRNA, Rhizosolenia setigera RNA virus; RTSV, rice tungro spherical virus; SDV, satsuma dwarf virus; SIV, Solenopsis invicta virus-2; SssRNAV, Schizochytrium single-stranded RNA virus; SV, Sapporo virus; TRSV, tobacco ringspot virus; TrV, Triatoma virus; TSV, Taura syndrome virus.

Full size image

The third conserved protein of picorna-like viruses, 3CPro, is more common than the S3H and is present in families from all RdRp clades apart from the partitivirus clade (Fig. 1). Most viral proteases have a catalytic cysteine that replaces the active serine residue that is characteristic of the rest of trypsin-like proteases88. However, at least two groups of viruses — the sobemovirus lineage of the RdRp clade 2 and astroviruses — possess serine proteases (Fig. 1). A reliable tree of virus proteases could not be obtained owing to the relatively low information content of the multiple alignment (Supplementary information S4 (figure)). However, it is noteworthy that viral serine proteases were polyphyletic, that is, the serine proteases of astroviruses formed a strongly supported clade with the cysteine proteases of potyviruses, whereas the serine proteases of sobemoviruses, luteoviruses and related viruses of fungi and chromalveolates comprised a distinct clade (data not shown).

The Big Bang of picorna-like virus evolution

The phylogenetic analyses presented in this article show that five of the six clades in the RdRp tree encompass picorna-like viruses that infect hosts from two or three eukaryotic supergroups. Early and, presumably, rapid diversification of picorna-like viruses, antedating the divergence of eukaryotic supergroups, seems to be the most parsimonious evolutionary scenario. However, the contribution of subsequent horizontal virus transfer (HVT) could be substantial as well, in accord with the concept of the reticulate evolution of viruses69. In particular, transmission of viruses between plants and fungi seems possible given the close associations between plants and their fungal pathogens. HVT might have been particularly important in the evolution of the Partitiviridae family, in which plant and fungal viruses are intermixed in phylogenetic trees89 (Fig. 3), and is also likely to account for the evolution of the Hypoviridae77 (Fig. 3).

However, it seems that HVT only confounded the results of a Big Bang of virus diversification, a scenario that conforms to the recently proposed general model of major evolutionary transitions44. In the Big Bang scenario, major branches of picorna-like viruses had already emerged by the time the eukaryotic supergroups radiated from their common ancestor and, then, viruses from this ancestral pool explored the evolving hosts and infected those that were susceptible. One prediction of the Big Bang model is that picorna-like viruses will eventually be identified that infect hosts from all the major lineages of eukaryotic organisms, although viruses of this superfamily so far have not been isolated from Amoebozoa, red algae and Rhizaria (which are generally poorly studied organisms).

The alternative hypothesis — namely, emergence of the ancestors of the six major clades of picorna-like viruses in one of the eukaryotic supergroups, with subsequent HVT to hosts from other supergroups — seems to be substantially less parsimonious, considering that this scenario would require numerous HVT events between organisms with widely different global ecologies and lifestyles. Furthermore, none of the supergroups of eukaryotes are known to host picorna-like viruses from all of the six clades that are present in an RdRp tree, a distribution that seems to be most compatible with viruses from a pre-existing ancestral pool infecting the emerging eukaryotic supergroups (Fig. 3).

How does this scenario of picorna-like virus evolution relate to the existing notions on the evolution of their cellular hosts? The Big Bang of picorna-like viruses is consistent with the probably rapid and tumultuous nature of eukaryogenesis that, under the symbiogenetic scenarios, was initiated by the archaeo-bacterial symbiosis90,91,92,93. Under this model, eukaryogenesis would involve extensive recombination between the symbiont and host genomes and, apparently, infestation of the host genes by group II retroelements that came from the symbiont and gave rise to the spliceosomal introns91. Explosive evolution of eukaryotic viruses in general, and the Big Bang of picorna-like virus evolution in particular, would be inherent to this turbulent era28. As discussed in detail elsewhere, symbiogenesis appears to be the most parsimonious scenario for the emergence of the eukaryotic cell, considering the presence of mitochondria or related organelles in all extensively characterized modern eukaryotes and the explanatory power of this model with respect to the origin of the nucleus and other eukaryotic organelles. However, the alternative scenario, namely the origin of an amitochondrial ancestor of eukaryotes as one of the three primary domains of life, has also been strongly defended in recent theoretical studies94,95. Adopting this scenario would not affect our conclusion on the Big Bang of picorna-like virus evolution but would push this event to an early, primordial stage of the evolution of life. This stage is believed to have involved rampant recombination between diverse genetic elements, a state that would be conducive to the explosive diversification of viruses24,28.

The origins of picorna-like viruses

The picorna-like superfamily is defined by the presence of a partially conserved set of genes that includes those encoding RdRp, the S3H, the 3CPro, VPg and JRC (Fig. 1). Among sequenced genomes of viruses infecting bacteria and archaea, none contain any pair of genes from this set. Barring the unlikely possibility that such viruses of prokaryotes remain to be discovered, it follows that the ancestor(s) of the picorna-like viral superfamily was assembled from individual genes during eukaryogenesis. Can we trace the sources of these genes? Despite the rapidity of the evolutionary processes during a Big Bang and the high rate of evolution of RNA virus genes, database searches seem to provide tangible clues.

We derived PSSMs for the RdRps, S3H and 3CPro of the picorna-like superfamily and compared them with the non-redundant protein database using PSI-BLAST (position-specific iterative basic local alignment search tool)96 to identify the closest homologues outside the picorna-like superfamily that could be the ancestors of these signature genes. The RdRp PSSM produced highly significant hits to the RdRps of the other two superfamilies of eukaryotic positive-strand RNA viruses and, notably, the reverse transcriptases (RTs) of bacterial group II retroelements (Table 1 and Supplementary information S2 (figure)). The similarity between the RdRps of picorna-like viruses and the RdRps of RNA bacteriophages was substantially lower (Table 1). The conservation of several sequence motifs and the structural similarity between RdRps of positive-strand RNA viruses and RTs have been described previously97,98,99,100, and the relationship between the two classes of polymerases is complemented by biochemical evidence, for example, the ability of RdRps to efficiently use dNTPs as substrates in the presence of Mn2+ cations101,102,103.

Table 1 Homologues of the signature genes of picorna-like superfamily viruses*
Full size table

Considering the symbiotic scenario of eukaryogenesis, it is notable that the RdRps of picorna-like viruses are most similar to RTs from prokaryotic retroelements, as opposed to those from eukaryotic retroid viruses or retroelements. Given these findings and the wide spread of group II retroelements in bacteria, in a sharp contrast to the scarcity of RNA bacteriophages, it appears plausible that the RdRps of eukaryotic positive-strand RNA viruses evolved from prokaryotic RTs. Group II retroelements are widely believed to be the progenitors of eukaryotic spliceosomal introns104,105,106, as well as ancestors of the eukaryotic telomerase and retroid viruses107,108,109. So this hypothesis places the origin of the picorna-like superfamily and other eukaryotic positive-strand RNA viruses in the middle of the turbulent process of eukaryogenesis.

The roots of the 3CPros of picorna-like viruses appear even clearer. Most of the statistically significant hits observed with the 3CPro PSSM are members of a distinct family of bacterial and mitochondrial serine proteases typified by the Escherichia coli periplasmic protease HtrA110 (Table 1). This relationship is supported by the analysis of structural neighbours, in which the mitochondrial protease HTRA2 (also known as OMI)111 comes up as the closest non-viral neighbour of 3CPro (data not shown). The similarity between the serine proteases of the HtrA family and the cysteine proteases of picornaviruses has been noticed previously88 but, at the time, the sequence information was insufficient to infer the nature of the evolutionary relationship between these protein families. With the current genomic data and considering the bacterial provenance, mitochondrial localization and function of the HtrA family of proteases in eukaryotes, it can be concluded that the 3CPro descends from an HtrA-family protease, and that this protease in turn is most probably derived from the mitochondrial endosymbiont.

The case of the SF3 helicase of picorna-like viruses is more complex. The PSSM-initiated sequence searches reveal that the highest similarity is to the helicases of circoviruses, followed by bacterial AAA+ ATPases; the available bacteriophage S3H sequences are much less similar to the picorna-like virus helicases (Table 1). However, the S3Hs have several sequence and structure features pointing to their monophyly112,113,114, which suggests that the S3Hs of eukaryotic viruses evolved from their bacteriophage homologues. In this scenario, the observed hierarchy of sequence similarity could be explained by the slower evolution of AAA+ ATPases of cellular organisms compared with the related viral S3H, or by the absence in the current database of the phage group that provided the putative ancestral helicase. Conceivably, the circoviruses are derivatives of this putative phage family.

The JRCs of picorna-like viruses, similarly, might have derived from capsid proteins of DNA-containing viruses of bacteria or archaea18. It should be noted, however, that the known icosahedral capsid proteins of prokaryotic DNA viruses, such as bacteriophages PRD1 or phi29 or Sulfolobus turret icosahedral virus115, have double JRC domains, whereas the capsid proteins of picorna-like viruses contain single JRC domains116. The similarity between the picorna-like virus JRCs and the capsid proteins of bacterial and archaeal viruses can be traced only through structural comparisons and is limited in extent (Ref. 18 and E.V.K., unpublished data), attesting to a substantial modification and, possibly, partial degradation of the JRC fold that was required to encapsidate small RNAs of picorna-like viruses. Alternatively, the picorna-like viral version of the JRC might have been derived from an unknown small prokaryotic virus.

Thus, the available evidence points to the assembly of the ancestral picorna-like viruses from diverse building blocks during eukaryogenesis and before the radiation of the eukaryotic supergroups (Fig. 5). The emergence of these ancestral viruses is probably best depicted as a Big-Bang-type event, so the order of emergence of the individual clades and the specific relationships between them could be undecipherable. In accordance with the concept of reticulate evolution of viruses, it is even conceivable that a common viral ancestor of picorna-like viruses never existed, that is, that the major clades of picorna-like viruses obtained their signature genes from different prokaryotic viruses and genetic elements. However, given the consistent presence of the five signature genes in the majority of picorna-like viruses, this possibility appears to be non-parsimonious. It is more likely that the Big Bang of picorna-like virus evolution was precipitated by accidental assembly of the signature genes in an ancestral virus (Fig. 5).

Figure 5: The proposed evolutionary scenario for the picorna-like superfamily of positive-strand RNA viruses of eukaryotes.
figure 5

This scenario is based on the symbiogenetic model of the origin of eukaryotes, according to which eukaryogenesis was initiated by the engulfment of an α-proteobacterium (the future mitochondrion) by an archaeon90,91,92,93. As discussed in the text, reverse transcriptase (RT)-encoding group II retroelements originating from the bacterial symbiont are the likely ancestors of both eukaryotic retroelements and spliceosomal introns, and the RT of these elements might have given rise to the RNA-dependent RNA polymerase (RdRp) of picorna-like viruses. Under this scenario, the superfamily 3 helicase (S3H) and the jelly-roll capsid protein (JRC) of the ancestral picorna-like virus are tentatively derived from a bacteriophage of the symbiont, and the 3C-like proteinase (3CPro) is derived from a symbiont's membrane protease (see text for details). Coloured ovals and the arrows at the bottom of the diagram symbolize burst-like emergence of the five supergroups of eukaryotes and of the six clades of the picorna-like viruses, respectively. CPF, capsid protein, filamentous capsid; CPU, capsid protein, unknown evolutionary origin; g, viral protein, genome-linked; IRES, internal ribosome entry site; JRC-N, nodavirus -like JRC; JRC-P, picornavirus -like JRC; JRC-S, sobemovirus-like JRC; S2H, superfamily 2 helicase.

Full size image

The evolutionary scenario schematically depicted in Fig. 5 is predicated on the symbiogenetic model of eukaryogenesis. At least one piece of evidence, the distinct bacterial origin of 3CPro, seems to be best compatible with this model. In general, however, the scenario of picorna-like virus evolution is robust with respect to the concepts of eukaryogenesis and would fit the three-domain model as well. The main difference would be pushing the assembly of the ancestral virus back to the pre-cellular stage of virus evolution24,28. Moreover, this scenario seems to be better compatible with the current data on the diversity of the JRC18 because, in this case, the JRC of picorna-like viruses could be considered the primitive form of this fold.

Subsequent evolution of picorna-like viruses seems to have involved a variety of substantial modifications of the viral genome layout, which often occurred in parallel in different clades (Fig. 5). The apparent replacement of the S3H by a superfamily 2 helicase in potyviruses is a case in point, as is the replacement of the JRC gene with a gene for an unrelated capsid protein that forms filamentous capsids117 in the same viral family. In this case, the changes to the viral bauplan can be linked to a specific host range that would facilitate recombination between viruses: in plants — the host organisms of potyviruses — viruses of the alpha-like supergroup, which typically have a superfamily 2 helicase and a filamentous capsid, are extremely abundant and were the likely source of the respective genes acquired by potyviruses. The hypoviruses, which are probable derivatives of potyviruses (although this is not obvious from the RdRp tree), have apparently lost both the capsid protein and the 3CPro. In this case, the loss of the capsid is linked to the predominantly vertical transmission of viruses in fungi. The nodaviruses (and, apparently, Sclerophtora macrospora virus A, the related virus from a chromalveolate) present perhaps the most dramatic case of gene loss and bauplan modification in the picorna-like superfamily, with both the 3C-like protease and VPg lost. A parallel loss of 3CPro is seen in the totiviruses and, apparently, in the entire partitivirus clade.

The Big Bang model implies that the early stages of the evolution of picorna-like viruses did not involve virus–host co-evolution inasmuch as different major clades of picorna-like viruses invaded the same eukaryotic supergroups. Of course, co-evolution is common at later, less turbulent phases of evolution that involve extensive virus–host co-adaptation as has been amply documented, for example, for mammalian herpesviruses118.

Conclusions

The results of phylogenetic analysis presented here suggest that diversification of the picorna-like superfamily of eukaryotic positive-strand RNA viruses occurred in a Big Bang at an early stage of eukaryogenesis, before the divergence of the supergroups of eukaryotes. This scenario implies that viruses from the ancestral pool invaded the emerging supergroups of eukaryotes. Thus, at least at this early stage in the evolution of RNA viruses of eukaryotes, there seems to have been no virus–host co-evolution in the sense of concomitant evolution of the host and viral lineages. However, evolution of picorna-like viruses was tightly intertwined with the pivotal events of eukaryogenesis such as the emergence of mitochondria and spliceosomal introns.