The virome of the last eukaryotic common ancestor and eukaryogenesis

Abstract

All extant eukaryotes descend from the last eukaryotic common ancestor (LECA), which is thought to have featured complex cellular organization. To gain insight into LECA biology and eukaryogenesis—the origin of the eukaryotic cell, which remains poorly understood—we reconstructed the LECA virus repertoire. We compiled an inventory of eukaryotic hosts of all major virus taxa and reconstructed the LECA virome by inferring the origins of these groups of viruses. The origin of the LECA virome can be traced back to a small set of bacterial—not archaeal—viruses. This provenance of the LECA virome is probably due to the bacterial origin of eukaryotic membranes, which is most compatible with two endosymbiosis events in a syntrophic model of eukaryogenesis. In the first endosymbiosis, a bacterial host engulfed an Asgard archaeon, preventing archaeal viruses from entry owing to a lack of archaeal virus receptors on the external membranes.

Eukaryotes differ from archaea and bacteria due to their complex cellular organization. This includes endomembranes (in particular, the nuclear compartment), a complex cytoskeleton and the mitochondrion, which itself evolved from an alphaproteobacterial endosymbiont^1–3. All of these features seem to be traceable to the last eukaryotic common ancestor (LECA)^2,4. Several models for the origin of eukaryotes (eukaryogenesis) have been proposed, but each differs with respect to the timing of the origin of the typical eukaryotic cellular organization^5–7. Phylogenomic analyses indicate that eukaryotes possess a mix of genes originating from archaea (in particular, Asgardarchaeota) and genes of apparent bacterial origin^8–12. This dichotomy among eukaryotic genes is thought to reflect the symbiotic origin of eukaryotes. However, whereas the origin of the mitochondria from an alphaproteobacterium appears indisputable, the nature of the host of the proto-mitochondrial endosymbiont remains uncertain. The most straightforward models suggest an Asgard archaeon as the host^2,5 (Fig. 1). However, such scenarios of eukaryogenesis are incompatible with the chemistry of cell membranes and the enzymology of membrane biosynthesis, which are unrelated in archaea and bacteria, as membranes in eukaryotes are of the bacterial type¹³. Thus, any eukaryogenesis scenario with an archaeal host would require a membrane replacement step. Alternatively, more complex models of eukaryogenesis propose that an Asgard archaeon was engulfed by a bacterium, followed by the loss of the archaeal membrane, then a second endosymbiosis that gave rise to mitochondria^6,14,15 (Fig. 1).

Fig. 1 |. The LECA virome.

Left, schematic of two alternative scenarios of eukaryogenesis that include either one endosymbiotic event (with an Asgard archaeon engulfing an alphaproteobacterium) or two such events (with a deltaproteobacterium engulfing an Asgard archaeon first and the resulting chimera then engulfing an alphaproteobacterium). Right, schematic phylogenetic tree of eukaryotes¹⁹, with major clades of eukaryotes indicated at the tree leaves and the broadly used names of the informal supergroups shown at the bottom of the figure. The predominant types of organisms in each clade are depicted with pictograms. Only Chloroplastida (green plants), Stramenopila (brown algae), Rhodophyta (red algae) and Opisthokonta (animals and some fungi) include multicellular eukaryotes, whereas the rest consist of unicellular forms. The phyla of eukaryote viruses are shown as a grid next to the corresponding cellular taxa. Genome types of the corresponding viruses are indicated above the taxon names, which are also colour coded according to the virus realm. The known virus–host associations are shown with coloured circles for cultivated viruses (blue), associations predicted from metagenomics and metatranscriptomics studies (grey) and endogenous viruses integrated in the host genomes (green). References exemplifying the depicted associations are provided in Supplementary Table 1. The composition of the LECA virome was inferred from the distribution via an informal parsimony approach whereby a group was assigned to the LECA if it was represented in at least three of the six supergroups of eukaryotes. Virus phyla mapped to the LECA are indicated by the coloured bars shown at the top of the grid. The height and intensity of the colour of the bars indicate the confidence of the inference. CRuMs, collodictyonids (syn. diphylleids) + Rigifilida + Mantamonas; EVE, endogenous viral element; RT, reverse-transcribing viruses; TSAR, telonemids, stramenopiles, alveolates and Rhizaria.

All life forms are hosts to viruses and other mobile genetic elements (MGEs), which can be either parasites or mutualists¹⁶. Recent phylogenomic efforts yielded an evolutionary taxonomy that encompasses most known viruses¹⁶. This taxonomic system comprises six realms: Riboviria, Monodnaviria, Duplodnaviria, Varidnaviria, Adnaviria and Ribozyviria^16,17. Unifying features in each group include hallmark proteins involved in genome replication (such as homologous RNA-dependent RNA polymerases (RdRPs) and reverse transcriptases) or virion formation (namely, distinct varieties of major capsid proteins and enzymes involved in genome encapsidation). Notably, there are major differences between the virome compositions of bacteria, archaea and eukaryotes (Box 1 and Fig. 2).

Box 1. Viromes of bacteria, archaea and eukaryotes.

Three of the four major virus realms, Duplodnaviria, Varidnaviria and Monodnaviria, are represented in each of the domain-specific viromes (Riboviria are so far missing in archaea), but with major differences in abundance, diversity and representation of kingdoms and phyla, where the lower taxa are confined to individual domains, as are the two smaller realms, Adnaviria and Ribozyviria. The RNA viruses in the kingdom Orthornavirae (Riboviria) are far more broadly represented in eukaryotes than they are in bacteria. Although the latest findings indicate that riboviruses are more prominent contributors to the bacterial virome than previously suspected^45,67,68, Riboviria remains dominated by viruses of eukaryotes. Even more strikingly, the kingdom Pararnavirae, which consists of reverse-transcribing viruses, is exclusively associated with eukaryotes, although bacteria and archaea harbour many non-viral retroelements.

The ssDNA viruses of the realm Monodnaviria are abundantly represented in both prokaryotic and eukaryotic viromes (Fig. 2), but the host ranges of the ssDNA viruses do not overlap already at the kingdom level⁶⁹.

In the vast realm Varidnaviria, the small kingdom Helvetiavirae is restricted to archaea and bacteria, whereas the expansive kingdom Bamfordvirae includes viruses from all three domains of life. Within Bamfordvirae, the phylum Nucleocytoviricota encompasses diverse large and giant viruses that are fully eukaryote specific. The second phylum, Preplasmiviricota, is a rare case of viruses infecting each of the three domains of life mixing at this taxonomy level.

The realm Duplodnaviria consists mostly of tailed bacteriophages and the related viruses of archaea (both within the class Caudoviricetes). Until recently, herpesviruses (order Herpesvirales) that are presently confined to animals were the only group of eukaryotic viruses within Duplodnaviria. However, the recent discovery of protist-infecting mirusviruses³⁴ suggests that duplodnaviruses could be far more widespread among eukaryotes than was previously suspected.

The small realm Adnaviria is widespread in archaea⁷⁰, but there is no detectable connection to viruses of bacteria or eukaryotes. The realm Ribozyviria includes hepatitis delta virus and hepatitis delta virus-like viruses discovered in other vertebrates, as well as insects⁷¹.

All in all, comparison of the viromes of prokaryotes and eukaryotes reveals distinct compositions, with all classes of viruses, most phyla and even some kingdoms and realms being domain specific. A major distinction between the viromes of prokaryotes and eukaryotes is the dominance of dsDNA viruses (both Duplodnaviria and Varidnaviria) in prokaryotes and the contrasting preponderance of Riboviria in eukaryotes (Fig. 2a). The elaborate endomembrane system of the eukaryotic cell apparently provides a fertile ground for RNA virus reproduction, whereas the nucleus presents a barrier for DNA viruses that few of them managed to clear or circumvent.

An orthogonal view from the vantage point of the diversity of the virus realms (Fig. 2b) shows that Duplodnaviria are heavily dominated by bacterial viruses, with small fractions of viruses infecting archaea and eukaryotes; among the Riboviria, the representation of viruses of bacteria and eukaryotes is comparable, with a slight excess of eukaryotic ones; Adnaviria is an exclusively archaeal realm; and the remaining three realms are heavily (or completely, in the case of Ribozyviria) dominated by viruses of eukaryotes (Fig. 2b).

Fig. 2 |. Viromes of prokaryotes and eukaryotes.

a, Representation of the six virus realms in bacteria, archaea and eukaryotes. b, The host ranges, at the domains of life level, of the six realms of viruses. Virus diversity in each realm is illustrated by images of the corresponding virions. The fractions of each realm were calculated as the fractions of virus genera recognized by the International Committee on Taxonomy of Viruses⁶⁶. Virions constructed from structural proteins with distinct folds are coloured differently.

We previously reconstructed the complex virome of the last universal cellular ancestor (LUCA) and found that Varidnaviria and Duplodnaviria (and possibly Riboviria and Monodnaviria) evolved at early stages of life predating the LUCA¹⁸. Given that viruses are obligate intracellular parasites that intimately interact with various components of the host cell—in particular, with cell membranes—analysis of virome composition can inform our understanding of host cell biology.

We sought to gain insight into the virome of the LECA and its evolutionary origin. We reconstructed the LECA virome and traced its origins to bacterial viruses, but did not find any links to archaeal viruses. We also address the implications of our reconstructed LECA virome for eukaryogenesis.

Reconstruction of the LECA virome

Using the reported host ranges of different virus groups across the branches of the eukaryotic phylogenetic tree¹⁹, we reconstructed the LECA virome (Supplementary Table 1). In this reconstruction, we faced two major obstacles. First, viromes are sparsely sampled across the diversity of eukaryotes, especially among unicellular organisms that encompass most of that diversity and are themselves not uniformly sampled¹⁹ (Fig. 2). Nevertheless, there have been considerable advances in characterizing the viromes of unicellular eukaryotes through metagenomics and metatranscriptomics^20–26. Second, ancestral reconstruction is confounded by horizontal virus transfer among diverse host lineages, whereby viruses change hosts (for example, between animals and plants via vectors such as insects or nematodes, or between plants and fungi via direct interaction)²⁷. Given these obstacles, we did not attempt formal, maximum likelihood reconstruction approaches^28,29, but rather, applied a semi-formal, parsimony-based approach, as we did previously for the LUCA virome¹⁸. We assumed that a group of viruses could be assigned to the LECA virome if it was represented in three or more of the six supergroups of eukaryotes (Fig. 1). Information on virus host range was extracted from the published literature using keyword searches (Supplementary Table 1). Considering the uncertainty of the topology in the deepest branchings of the eukaryotic tree¹⁹, we surmised that this simple majority rule approach would result in the most realistic approximation of the LECA virome.

The realm (the top rank in virus taxonomy) Riboviria is broadly represented in the LECA virome. Within Riboviria, all five phyla in the Orhtornavirae kingdom are widely spread across the tree and map back to the LECA. Moreover, the diversification of some ribovirus phyla seems to pre-date the LECA (Supplementary Table 1). One potential caveat is the possibility of horizontal virus transfer over long evolutionary distances, such as between plants and animals, particularly in the case of the ribovirus phylum Negarnaviricota²⁷. In the kingdom Pararnavirae, two virus families, Metaviridae and Pseudoviridae (also known as Ty3/Gypsy and Ty1/Copia retrotransposons, respectively), are nearly ubiquitous among eukaryotes and confidently map back to the LECA (Fig. 1 and Supplementary Table 1). The remaining pararnaviruses, however, appear to have evolved later, in animals and plants (Supplementary Table 1).

In the realm Monodnaviria, two phyla, Cressdnaviricota³⁰ and Cossaviricota, are represented in eukaryotes. The Cressdnaviricota are broadly distributed and trace back to the LECA (Fig. 1 and Supplementary Table 1). In contrast, the phylum Cossaviricota consists of several groups of single-stranded DNA (ssDNA) viruses and viruses with small double-stranded DNA (dsDNA) genomes with relatively narrow host ranges confined to animals, and thus appears to have evolved post-LECA (Fig. 1 and Supplementary Table 1).

In the realm Varidnaviria, the phylum Nucleocytoviricota is exclusive to eukaryotes and widespread across the eukaryote diversity, tracing back to the LECA (Fig. 1 and Supplementary Table 1). In the phylum Preplasmiviricota, eukaryotic viruses are represented by a diverse group of endogenous viruses known as polintons or polintoviruses³¹, virophages³² and the currently unclassified polinton-like viruses³³. These viruses are widespread in eukaryotes and probably belong to the LECA heritage (Fig. 1 and Supplementary Table 1). In contrast, adenoviruses that belong to the same phylum are limited in their host range to animals and seem to be a late (that is, postdating the origin of animals) derivative of polintons³¹.

The realm Duplodnaviria is dominated by bacterial and archaeal viruses. Until recently, the phylum Peploviricota, which includes herpesviruses, was the only group of eukaryotic viruses in this realm and appeared to be limited to animal hosts. However, the recent discovery of mirusviruses that are expected to be assigned to Duplodnaviria³⁴, given the presence of the corresponding hallmark structural proteins, changed the picture. The host range of mirusviruses has not been directly characterized but probably includes unicellular eukaryotes, suggesting that duplodnaviruses were represented in the LECA virome (Fig. 1 and Supplementary Table 1).

Similarly, the recent expansion of the previously tiny realm Ribozyviria^35,36, to include viruses probably infecting diverse protists, suggests the possibility that the diversity of this realm has been largely overlooked and brings into question its origin in animals. We cannot rule out the presence of ribozyviruses in LECA, although under the criteria adopted here they were not included (Fig. 1).

There are similarities between the LECA and LUCA viromes. The viromes of both common ancestors are complex and include representatives of most viral phyla. This finding leads us to propose that the time between the origin of eukaryotes and the advent of the LECA involved extensive diversification of the virosphere, concomitant with the evolution of distinct architecture of the eukaryotic cell³⁷.

Next, we discuss potential origins of the eukaryotic virome, through a process that we name eukaryovirogenesis, in the context of specific models of eukaryogenesis.

Bacterial origins of the LECA virome

The information-processing systems of the eukaryotic cell—replication, transcription and translation machineries—evolved from cognate systems in archaea³⁸. Given that viruses and other MGEs are informational parasites, it seems plausible that the eukaryotic virome might have evolved from the archaeal virome. Recently, the origins of eukaryotic information systems, along with many cytoskeleton and endomembrane components, were traced to the archaeal phylum Asgardarchaeota^10–12, which includes the closest archaeal relatives to eukaryotes. In the best-supported phylogenies of universal genes, eukaryotes branched from within Asgardarchaeota. Owing to this evolutionary relationship, viruses of Asgard archaea may be possible ancestors of the viruses of eukaryotes. However, analyses of several families of viruses associated with Asgardarchaeota did not find support for any of these viruses being candidates for ancestors of known eukaryotic viruses^39–41. The proposed relationship between some of the Asgard viruses and Nucleocytoviricota⁴¹ stems entirely from the generic homology among proteins involved in DNA replication and nucleotide metabolism. These proteins are common in diverse viruses with large DNA genomes, particularly in other archaeal Caudoviricetes⁴², as well as cell-based organisms. Indeed, the fraction of Nucleocytoviricota gene homologues in Asgard archaeal viruses is not higher than in bacteriophages or non-Asgard archaeal viruses⁴¹. The shared presence of these widespread genes does not reflect a common origin of the Asgard viruses and eukaryotic Nucleocytoviricota, and similarly, there is no specific relationship traceable between any viruses of eukaryotes and viruses of other archaea. Nevertheless, the Asgard archaeal virome is as-yet sparsely sampled, so it cannot be ruled out that uncharacterized archaeal viruses seeded some part of the eukaryotic virome; furthermore, extensive study of the Asgard virome will be needed to address this possibility.

In contrast, bacterial roots were detected for the eukaryote-infecting viruses from all four virus realms, as reported by previous studies on the evolution of each of the realms (Fig. 3). In the phylogenetic trees of the RdRPs of the kingdom Orthornavirae, the deepest branch is the phylum Lenarviricota, which consists of bacterial leviviruses and their direct descendants infecting a broad range of eukaryotes^43–45. The evolutionary scenario for this phylum can be readily reconstructed (Fig. 3): initially, an ancestral levivirus lost its capsid protein gene, giving rise to eukaryotic capsidless RNA replicators in the classes Amabiliviricetes and Howeltoviricetes, with the latter replicating in the mitochondria. The Amabiliviricetes subsequently gave rise to Miaviricetes, the largest group within Orthornavirae⁴⁵, by capturing the single jelly-roll (SJR) capsid protein gene (the most common capsid protein among viruses of eukaryotes thought to originally derive from a host sugar-binding protein⁴⁶, probably from an RNA virus of the phylum Kitrinoviricota). The progenitor of the remaining four phyla of Orthornavirae apparently originated from a common ancestor with Lenarviricota and followed a similar evolutionary path whereby the RdRP was inherited from a bacterial ancestor but the levivirus capsid protein was replaced with structurally unrelated proteins. In line with the general trend in virus evolution, the origin of eukaryote-infecting orthornaviruses seems to have involved preservation of the ancestral replication machinery that ultimately combined with eukaryote-specific virion proteins. It should be emphasized that, albeit with incomplete sampling, no RNA viruses of archaea have been discovered so far, leaving the bacterial origin of the eukaryotic RNA virome as the only viable scenario at the time of writing.

Fig. 3 |. Bacterial roots of the LECA virome.

The hypothetical scenarios of the origin of the major components of the LECA virome from bacterial viruses and non-viral MGEs. The major changes accompanying the evolutionary transitions between the corresponding bacterial and eukaryotic viruses and MGEs and subsequent evolution in eukaryotes are explained within text boxes over the arrows. Viruses and MGEs are depicted with the schematics of the corresponding genomes and virions. DNA and RNA genomes are indicated with red and green wavy lines, respectively. The capsid protein genes are shown in blue, yellow, pink and grey, with the corresponding capsids depicted with the matching colours. DJR, double jelly-roll; MCP, major capsid protein; RCRE, rolling circle replication endonuclease; HEART, hepadnavirus-like retroelement⁷².

Pararnavirae is the only possible exception to the bacterial origin of the eukaryotic virome because archaeal origin cannot be ruled out. The reverse transcriptase of the pararnaviruses was reported to be inherited from prokaryotic group II introns (retrotransposons), which are broadly represented in both bacteria and archaea. Indeed, phylogenetic analyses do not unequivocally link pararnaviruses with either bacterial or archaeal ancestors⁴⁷. The general pathway of evolution, though, seems to be the same, whether from a bacterial or archaeal retrotransposon, whereby the ancestral reverse-transcribing virus evolved by recruiting cellular proteins for the functions of capsid proteins, nucleocapsids and virus proteases, at a pre-LECA stage of eukaryogenesis^48,49 (Fig. 3).

The ancestral eukaryotic group in Monodnaviria, the cressdna-viruses, as in the case for pararnaviruses, evolved from non-viral bacterial MGEs—in this case, small bacterial (but not archaeal) plasmids replicating via the rolling circle mechanism⁵⁰. These plasmids provided the genome scaffold and the gene encoding the endonuclease and superfamily 3 helicase (a signature of eukaryotic cressdnaviruses) required for replication initiation, whereas the capsid protein was apparently acquired via recombination with complementary DNA copies of ribovirus SJR capsid protein genes (Fig. 3).

The phylum Preplasmiviricota and specifically polintoviruses in the realm Varidnaviria appear to be direct descendants of tailless bacteriophages belonging to this phylum³¹. The origin of the phylum Nucleocytoviricota probably involved recombination between a mirusvirus-like duplodnavirus (from which the replication machinery of the nucleocytoviruses was inherited) and a polintovirus that donated the structural module replacing the duplodnavirus morphogenetic genes (Fig. 3).

Finally, the structural gene module of mirusviruses, which probably comprise the ancestral group from the realm Duplodnaviria infecting eukaryotic hosts, clearly derives from the homologous genes of tailed bacteriophages or archaeal viruses, which comprise the class Caudoviricetes within this realm³⁴.

The above findings led us to propose two key features of eukaryovirogenesis that seem to have occurred concomitantly with eukaryogenesis itself (Fig. 2). First, all diverse groups of viruses comprising the LECA virome evolved from ancestral bacterial viruses, or non-viral MGEs, with the only possible exception being pararnaviruses. Second, the provenances of the genes encoding the components of the replication apparatus and those encoding virion components are markedly different, recapitulating the key trend of primordial virogenesis, where the replication machinery is thought to descend from the primordial pool of replicators whereas the structural proteins were apparently captured from the host at early stages of cellular evolution¹⁸.

The LECA virome and eukaryogenesis

A bacterial origin for the LECA virome demands explanation, given the archaeal origin of the eukaryotic information-processing systems. A key feature that is probably relevant for eukaryovirogenesis and that links eukaryotes to bacteria rather than archaea is the bacterial provenance of eukaryotic cell membranes¹³. Bacterial and eukaryotic membranes are based on glycerol-3-phosphate ester linked to fatty acids, whereas archaeal membranes comprise glycerol-1-phosphate ether bound to isoprenoids⁵¹. Although eukaryotes inherited archaeal pathways for the biosynthesis of isoprenoids, these are not major structural components of eukaryotic membranes⁵².

For any model of eukaryogenesis, the disparity between bacterial and eukaryotic membranes on the one hand and archaeal membranes on the other is a major challenge. The simplest symbiogenetic scenarios⁵, which involve an archaeal host and an alphaproteobacterial proto-mitochondrial endosymbiont as the only two partners in eukaryogenesis, face the difficulty of replacing the archaeal membrane with the bacterial one, for which there is no known precedent. The bacterial provenance of the LECA virome seems more compatible with alternative eukaryogenesis models, in which the emerging protoeukaryote never had an archaeal plasma membrane^14,15,53. Initially motivated by the plausibility of a metabolic symbiosis between a hydrogen-producing bacterium (possibly, a deltaproteobacterium) and a methanogenic archaeon, these syntrophy scenarios received a major boost with the recent demonstration of the syntrophic relationship between an Asgard archaeon and a deltaproteobacterium^54–56. In an updated syntrophy model¹⁵, eukaryogenesis started as a metabolic ectosymbiosis between a sulfate-reducing deltaproteobacterium and a hydrogen-producing Asgard archaeon, which was subsequently internalized and lost its membrane, probably after the emergence of bacterial endomembranes that surrounded the engulfed archaeon. The internalized archaeon became the progenitor of the eukaryotic nucleus (Figs. 1 and 4). This model implies two-stage eukaryogenesis, in which the merger between a (deltaproteo)bacterial host and an archaeal endosymbiont gave rise to an intermediate—the first eukaryotic common ancestor (FECA). This first endosymbiotic event was followed by the secondary endosymbiosis, whereby the FECA gave rise to the second eukaryotic common ancestor (SECA) by capturing a versatile sulfur-oxidizing and facultatively aerobic alphaproteobacterium that became the mitochondrion (Fig. 4). This eukaryogenesis scenario is buttressed by the wide spread of serial endosymbiosis in the subsequent evolution of eukaryotes, of which the evolution of chloroplasts from cyanobacteria in the ancestor of Archaeplastida is only one example^57,58. Furthermore, this scenario is compatible with phylogenomic analysis indicating that alphap roteobacterial proteins were acquired relatively late in the evolution of eukaryotes^59,60. Notably, the enzymes of the biosynthetic pathway for steroids, which are essential components of eukaryotic membranes, have clear deltaproteobacterial origin⁶¹.

Fig. 4 |. Eukaryogenesis and eukaryovirogenesis.

The scenario of eukaryovirogenesis is based on the updated syntrophy model of the eukaryogenesis scenario with a two-stage endosymbiosis¹⁵. The main stages of eukaryogenesis and eukaryovirogenesis are indicated with numbers. (1) Formation of a syntrophic metabolic consortium consisting of a deltaproteobacterium and an Asgard archaeon, where each organism is associated with a specific virome. (2) Internalization of the Asgard archaeal symbiont by the deltaproteobacterium results in the emergence of the FECA and in exclusion of the archaeal virome by the bacterial membrane of the FECA. At this stage, the Asgard archaeon is still bound by the archaeal-type membrane, but periplasmic space starts to develop around it. The symbiosis is stabilized through fusion of the bacterial and archaeal genomes, entailing horizontal exchange of genes and chromosomal retroelements such as group II introns. RNA viruses of the deltaproteobacterium undergo diversification accompanied by replacement of the ancestral capsid protein gene with the SJR capsid protein. The deltaproteobacterium also carries rolling circle plasmids and is infected with tailed dsDNA bacteriophages. A distinct consortium between the FECA and an alphaproteobacterium carrying its own specific virome is formed. The alphaproteobacterial virome includes a T7-like virus that integrates into the genome and persists as a prophage, as well as distinct RNA viruses (leviviruses). (3) Internalization of the alphaproteobacterium by the FECA results in shedding of most of the alphaproteobacterial viruses, except for tectiviruses, leviviruses and the T7-like prophage. An endomembrane system develops from the cytoplasmic membrane of the deltaproteobacterium in proximity of the Asgard archaeal membrane. The alphaproteobacterial genome migrates to the shaping nucleus, leading to the emergence of the SECA. Loss of the levivirus capsid protein gene leads to the emergence of capsidless narnaviruses and mitoviruses (narna/mito) replicating within the alphaproteobacterium. Escape of the tectivirus genome into the cytoplasm of the deltaproteobacterium by losing the capsid protein genes yields linear cytoplasmic plasmids, whereas integration of the tectivirus into the emerging nucleus gives rise to polintons and polintoviruses. Capture of the SJR capsid protein gene by rolling circle plasmids gives rise to Cressdnaviricota. (4) The alphaproteobacterial endosymbiont undergoes final transformation into the mitochondrion, retaining capsidless narnaviruses and linear tectivirus-derived plasmids. The archaeal membrane of the Asgard endosymbiont is replaced with the endomembrane-derived nuclear envelope with nuclear pores, surrounded by the endoplasmic reticulum network, yielding the LECA. Re-acquisition of the capsid protein gene by narna-like viruses yields members of Miaviricetes; other members of the Orthornavirae undergo extensive diversification. Polintons in the nuclear genome give rise to polinton-like viruses, virophages and other groups of viruses with double jelly-roll capsid proteins. Recombination between polintons and duplodnaviruses of the mirusvirus group gives rise to Nucleocytoviricota. Reverse-transcribing pararnaviruses emerge from retroelements originating either from Asgard archaea or bacterial group II introns. Question marks signify uncertainty regarding the nature of the ancestral deltaproteobacterial RNA virus. CP, capsid protein; ER, endoplasmic reticulum; RC, rolling circle.

Under the syntrophy scenario, the LECA virome was shaped by two waves of adaptation of bacterial viruses: first from the deltaproteobacterial virome and then from the virome of the alphaproteobacterial proto-mitochondrion (Fig. 4). Crucially, in this scenario, the emerging eukaryotic cell maintained the bacterial membranes through all stages of eukaryogenesis, whereas the archaeal membrane of the primary endosymbiont was lost. Thus, the viruses of the Asgard archaeon were excluded during the first stage of eukaryogenesis, primarily due to the inaccessibility of the archaeal virus receptors following the internalization of the archaeal symbiont. The escape from viruses would facilitate the endogenization of the archaeal symbiont en route to the FECA, jumpstarting eukaryogenesis. The bacterial provenance of the eukaryote virome buttresses models of eukaryogenesis that postulate the evolutionary continuity of bacterial membranes, such as the syntrophy scenario.

Possible stages in eukaryovirogenesis

Although it is difficult to assign the origins of specific virus groups in the LECA virome to one of the two bacterial partners, there are clues for this assignment (Figs. 3 and 4). Starting with Orthornavirae, the evolution of Lenarviricota from alphaproteobacterial leviviruses seems most likely, given the mitochondrial replication site of mitoviruses, which are direct eukaryotic descendants of leviviruses^62,63. Thus, the eukaryotic members of Lenarviricota apparently evolved along the path from SECA to LECA (Fig. 4). The origins of the rest of the eukaryotic riboviruses are much less clear, but could be more ancient considering the topology of the phylogenetic tree of the RdRP where the first split is between Lenarviricota and the rest of Orthornavirae⁴⁵. The common ancestor of the four phyla of Orthornavirae, which consist primarily of eukaryotic viruses, emerged en route from the FECA to the SECA, from an RNA virus of deltaproteobacteria—either a levivirus or an unknown ancestral virus (Fig. 4). Additionally, the origin of the eukaryotic riboviruses from a deltaproteobacterial rather than alphaproteobacterial ancestor appears likely because this scenario does not require virus escape from the endosymbiont. The virome of deltaproteobacteria has barely been sampled, suggesting that ancestors of riboviruses might not yet be discovered. The origin of eukaryotic riboviruses was precipitated by the exaptation of a cellular protein as SJR capsid protein.

The provenance of Pararnavirae remains uncertain, with possibilities being an origin in group II introns of the Asgard archaeal endosymbiont or any of the bacterial partners. Pararnavirae are the only major group of viruses of eukaryotes to which the exclusion of archaeal ancestor due to membrane incompatibility does not apply, because these viruses apparently originated from non-viral MGEs at a relatively late stage of eukaryogenesis. Cressdnaviricota, the dominant phylum of Monodnaviria in eukaryotes³⁰, probably evolved from bacterial, potentially deltaproteobacterial, plasmids through the acquisition of capsid protein genes (possibly from RNA viruses) during the transition from the FECA to the SECA (Fig. 4).

The topology of the phylogenetic tree of the protein-primed DNA polymerase (the hallmark protein of Preplasmiviricota), where the first split in the eukaryotic portion of the tree is between mitochondrial linear dsDNA plasmids and eukaryotic viruses³¹, implies an alphaproteobacterial origin of the eukaryotic members of this phylum en route from the SECA to the LECA. Notably, the host range of contemporary tectiviruses includes alphaproteobacteria⁶⁴. Some of the descendants of an alphaproteobacterial tectivirus lost the capsid protein genes and became the mitochondrial plasmids, whereas others migrated to the proto-nucleus as polintons and then gave rise to the other eukaryotic lineages of Preplasmiviricota (polinton-like viruses, adenoviruses, virophages and linear cytoplasmic plasmids).

The ancestor of the mirusviruses was probably a deltaproteobacterial phage that gave rise to ‘Mirusviricota’ (and eventually, to Peploviricota in animals) in protoeukaryotes en route from the FECA to the SECA and then to Nucleocytoviricota through recombination with Preplasmiviricota that donated the structural gene module replacing that of duplodnaviruses. Given the apparent alphaproteobacterial origin of the eukaryotic members of Preplasmiviricota, the origin of Nucleocytoviricota should be associated with the SECA to LECA stage (Fig. 4). Along a similar yet opposite route, the replication module of preplasmiviruses recombined with the morphogenetic module of parvoviruses giving rise to the Mouviricetes within the phylum Cossaviricota⁶⁵, underscoring the importance of module shuffling during diversification of the eukaryotic virome.

Outlook

The recent expansion of characterized diversity in all realms of viruses enables far more robust reconstruction of ancestral viromes than was previously possible. By examining the distributions of viruses infecting members of different eukaryotic clades across the evolutionary tree of eukaryotes, we propose that each of the four major virus realms was probably already represented by multiple groups in the LECA virome. The principal diversification of the eukaryotic virome apparently occurred during the relatively short time separating the origin of protoeukaryotes via endosymbiosis and the advent of the LECA, which already resembled extant unicellular eukaryotes. For each major virus group, with the possible exception of reverse-transcribing viruses, an origin from bacterial viruses or non-viral bacterial MGEs such as rolling circle plasmids is readily traceable. Although some of the corresponding virus groups can be traced back to the LUCA virome¹⁸, all evidence points to eukaryotes inheriting the bacterial rather than archaeal descendants of these ancient viruses.

We propose that eukaryovirogenesis involved extensive diversification of virus genomes: in particular, the replacement of structural gene modules. One caveat to this proposal is that the archaeal virosphere, and particularly the Asgard viruses, remain undersampled. When new groups of archaeal viruses are discovered, we will need to incorporate them into our analyses.

The bacterial origin of the LECA virome is most compatible with a model of eukaryogenesis in which the emerging protoeukaryote maintained the bacterial membrane through two stages of symbiogenesis. The first stage probably involved a (deltaproteo)bacterium engulfing an Asgard archaeal endosymbiont, which gave rise to the nucleus. The second stage probably featured the capture of an alphaproteobacterium by the archaeo-bacterial chimera to form the mitochondrion. This evolutionary continuity of bacterial membranes during eukaryogenesis would have caused exclusion of viruses specific for the archaeal endosymbiont from the evolving eukaryotic cell due to lack of archaeal virus receptors on the bacterial membranes. We further propose that the origins of the major groups of eukaryotic viruses can be tentatively assigned to one of the two steps in this endosymbiotic eukaryogenesis scenario. It seems that viruses of eukaryotes have either deltaproteobacterial or alphaproteobacterial origins and the diversification of virus genomes during eukaryovirogenesis involved multiple recombination events between these two groups of viruses. Viruses of both deltaproteobacteria and alphaproteobacteria have been poorly sampled, so in-depth study of these viromes should shed new light on eukaryovirogenesis.