Genes with readily detectable homologs in cellular life forms

1. Genes with closely related homologs in cellular organisms (typically, the host of the given virus) present in a narrow group of viruses.

2. Genes that are conserved within a major group of viruses or even several groups and have relatively distant cellular homologs.

Virus-specific genes

3. ORFans, i.e., genes without detectable homologs except, possibly, in closely related viruses.

4. Virus-specific genes that are conserved in a (relatively) broad group of viruses but have no detectable homologs in cellular life forms.

Viral hallmark genes

5. Genes shared by many diverse groups of viruses, with only distant homologs in cellular organisms, and with strong indications of monophyly of all viral members of the respective gene families, – we would like to coin the phrase "viral hallmark genes" to denote these genes that can be viewed as distinguishing characters of the "virus state".

The relative contributions of each of these classes of genes to the gene sets of different viruses strongly depend on the viral genome size which differs by more than three orders of magnitude. Viruses with small genomes, such as most of the RNA viruses, often have only a few genes, the majority of which belong to the hallmark class. By contrast, in viruses with large genomes, e.g., poxviruses, all 5 classes are broadly represented. In order to illustrate the diversity of viral "genomescapes" more concretely, we show in Table ​Table22 the decomposition of the gene sets of three viruses with a small, an intermediate-sized, and a large genomes, respectively, into the 5 classes. Notably, moderate-sized and large genomes of bacteriophages and archaeal viruses are dominated by ORFans that often comprise >80% of the genes in these viruses. Rapidly evolving phage ORFans are thought to supply many, if not most, of the ORFans found in prokaryotic genomes (the lack of detectable sequence conservation notwithstanding), hence playing an important role in evolution of prokaryotes [12].

Replication of positive-strand RNA viruses, dsRNA viruses, negative-strand RNA viruses, and retroid viruses/elements is catalyzed by another idiosyncratic class of viral enzymes, RNA-dependent RNA polymerases (RdRp) and reverse transcriptases (RT). The positive-strand RNA virus RdRp and the RT form a monophyletic cluster within the vast class of the so-called palm-domains that are characteristic of numerous polymerases [21-24]. The RdRps of dsRNA viruses and negative-strand RNA viruses are likely to be highly diverged derivatives of the same polymerase domain, an old conjecture [24-26] that has been vindicated by the recent determination of the structure of a dsRNA bacteriophage RdRp [27,28].

The palm domain is likely to be the primordial protein polymerase that emerged from the RNA world where nucleotide polymerization was catalyzed by ribozymes [29]. This is supported not only by the wide spread of this domain in modern life forms but also by the structural and, by inference, evolutionary link between the palm domain and the RNA-recognition-motif (RRM) domain, an ancient RNA-binding domain that might have, initially, facilitated replication of ribozymes [30]. The palm-domain RdRps and RTs are excluded from the regular life cycles of cellular life forms, although most eukaryotic genomes encompass numerous copies of RT-containing retroelements, and prokaryotes have some such elements as well [31,32]. These elements, however, are selfish and, from the evolutionary standpoint, virus-like. The most notable incursion of an RT into the cellular domain is the catalytic subunit of the eukaryotic telomerase, the essential enzyme that is involved in the replication of chromosome ends [33,34].

The list of viral hallmark genes given in Table ​Table33 is a conservative one. There well might be other genes that merit the hallmark status but for which clear evidence is hard to come up with. An important case in point is the B-family DNA polymerase that is the main replication enzyme of numerous dsDNA viruses of bacteria and eukaryotes. Homologs of these DNA polymerases are found in all archaeal and eukaryotic genomes, so that monophyly of all viral polymerases does not seem to be demonstrable in phylogenetic analyses [35,36]. However, this potentially could be explained by relatively (with respect to cellular homologs) fast evolution of the polymerases in various viral lineages, which would obscure their common origin. Furthermore, monophyly of the polymerases of all viruses that employ a protein-primed mechanism of dsDNA replication (animal adenoviruses and tailed phages like PRD1 or 29) has been claimed [37]. Thus, although this currently cannot be shown convincingly, it seems possible (and, as discussed below, even likely) that the DNA polymerase is a viral hallmark gene in disguise. More generally, further sequencing of viral genomes, combined with comprehensive comparative analysis, might reveal additional genes that, despite relatively limited spread among viral lineages, will qualify as hallmark genes.

The combination of features of viral hallmark proteins is highly unusual and demands an evolutionary explanation. Indeed, the hallmark genes are, without exception, responsible for essential, central aspects of the viral life cycles, including genome replication, virion formation, and packaging of the genome DNA into the virion (Table ​(Table3).3). These genes span sets of extremely diverse classes of viruses, often possessing different reproduction strategies and differing by three orders of magnitude in genome size. Finally, all viral hallmark genes have remote homologs in cellular life forms (Table ​(Table3)3) but the viral versions appear to be monophyletic.

Three hypotheses on the origin of viral hallmark genes immediately come to mind. The first possibility is that the notion of hallmark virus proteins is based on an artifact. The argument, that is commonly invoked in discussions of unexpected patterns of homologous relationships and could well be waged against the notion of viral hallmark proteins, is that genuine orthologs of these proteins (direct evolutionary counterparts, typically, with the same function) actually do exist in cellular life forms but are not detectable due to rapid sequence divergence between viral and cellular proteins. However, this reasoning does not seem to survive closer scrutiny. Firstly, the conservation of the hallmark proteins in extremely diverse classes of viruses with widely different replication/expression strategies (Table ​(Table3)3) but not in cellular life forms is hardly compatible with the rapid divergence interpretation. Indeed, for this to be the case, acceleration of evolution of the hallmark genes in diverse classes of viruses should occur in such a manner that the similarity between viral proteins survived, whereas the similarity between viral proteins and their hypothetical cellular orthologs vanished. Parallel conservation of hallmark protein sequences might be perceived in the case of structural protein, such as JRC, but is hardly imaginable for proteins involved in replication of structurally very different genomes, such as S3H that is conserved among viruses with RNA, ssDNA, and dsDNA genomes. Furthermore, for most of the hallmark proteins, distant and functionally distinct homologs from cellular organisms are detectable (S3H and viral RNA-dependent polymerases are the primary examples) which makes the existence of elusive orthologs extremely unlikely.

The other two hypotheses accept the hallmark viral proteins as reality but offer contrasting evolutionary scenarios to account for their existence and spread.

1. One hypothesis would posit that the hallmark genes comprise the heritage of a "last universal common ancestor of viruses" (LUCAV). This scenario implies that, despite all evidence to the contrary (see above) all extant viruses are monophyletic, although their subsequent evolution involved massive gene loss in some lineages as well as extensive acquisition of new genes from the hosts in others.

2. By contrast, under the hypothesis of polyphyletic origin of viruses, the spread of the hallmark genes across the range of virus groups could be explained by horizontal gene transfer (HGT).

Upon closer inspection, none of these hypotheses seems to be a viable general explanation for the existence and distribution of the viral hallmark genes. Indeed, the relatively small number and the mosaic spread of the hallmark genes (Table ​(Table3)3) do not seem to be conducive to the LUCAV notion although it is apparent that a great number of diverse viruses, if not all of them, share some common history. Conversely, the extremely distant similarity between the hallmark proteins from diverse virus groups with dramatically different replication strategies is poorly compatible with an HGT scenario.

The existence of the hallmark virus genes seems to effectively falsify both the cell degeneration and the escaped-genes concepts of viral evolution. With regard to the cell degeneration hypothesis, let us consider the NCLDV, the class of large viruses to which the cell degeneration concept might most readily apply and, indeed, was, in the wake of the discovery of the giant mimivirus [50-52]. Among the 11 signature genes that are shared by all NCLDVs ([53] and Table ​Table1),1), three crucial ones (JRC, S3H, and the packaging ATPase) are virus hallmark genes. A clear inference is that even the simplest, ancestral NCLDV would not be functional without these genes. However, cellular derivation of this ancestral NCLDV would have to invoke decidedly non-parsimonious, ad hoc scenarios, such as concerted loss of all hallmark genes from all known cellular life forms or their derivation from an extinct major lineage of cell evolution. The same line of logic essentially refutes the escaped genes concept inasmuch as the hallmark genes had no cellular "home" to escape from. Again, to save "escaped genes", an extinct cellular domain would have to be postulated.

Two recent conceptual developments in the study of origin and evolution of viruses deserve special attention (Table ​(Table4;4; see also discussion below). First, Bamford and coworkers [13-15] and, independently, Johnson and coworkers [54] capitalized on the conservation of the structure of the jelly-roll capsid protein in a wide variety of viruses to propose the idea of an ancient virus lineage spanning all three domains of cellular life (archaea, bacteria, and eukaryotes). Second, Forterre presented an elaborate scheme of virus-cell coevolution from the earliest stages of life's evolution. Under this concept, viruses emerged independently within three lineages of RNA-based cells (the progenitors of archaea, bacteria, and eukaryotes) and "invented" DNA replication that was subsequently captured from different viruses by each type of host cells, in three independent transitions to DNA genomes [43,55].

A crucial, even if fairly obvious, aspect of viral evolution is that it is inextricably linked to the evolution of the hosts which, when traced back to the earliest stages of life's evolution, attests to the necessity to consider scenarios of virus origins in conjunction with models for the origin and early evolution of cells. This dramatically ups the ante for the study of virus evolution, bringing it to the center stage of evolutionary biology [56,57]. Therein, however, seem to lie some of the major problems encountered by the current hypotheses of virus evolution (Table ​(Table4).4). The concept of an ancient virus lineage advocated by Bamford and coworkers, in addition to being based on the broad spread of a single gene (JRC), which is construed as the "self" of a virus [13], simply leaves a glaring gap as the connection between viral and cellular evolution is not considered such that it remains unclear in what kind of cellular environment the early evolution of the primordial viral lineage took place.

By contrast, Forterre's concept is embedded within a specific scenario of cellular evolution, and we believe that this is the valid approach to the analysis of the origins and evolution of viruses and cells – the only chance to achieve understanding of these difficult problems is to consider them in conjunction. However, the specific scenario of cellular evolution favored by Forterre appears to be poorly compatible with the results of comparative genomics, thus compromising the model of virus evolution associated with it. Indeed, Forterre considers three lineages of RNA-cells evolving from the Last Universal Common Ancestor (LUCA) of the known life forms and giving rise to bacteria, archaea, and eukaryotes, with the transition to DNA-cells mediated by domain-specific viruses. This scenario, while complete with respect to the evolution of both cells and viruses, encounters at least three major, probably, insurmountable problems. First, it is dubious at best that the combination of complexity and genetic stability required of an even a minimal cell – a complement of a few hundred genes that is accurately transmitted over many cellular generations – is attainable with an RNA genome. Indeed, given the inherent instability of RNA, such a genome would have to consist of several hundred RNA molecules. Accurate partitioning of this multipartite genome between daughter cells would require an RNA segregation system of unprecedented precision; thus, faithful vertical inheritance of the genome is hardly imaginable. Stochastic segregation of RNA segments in a polyploid cell might seem to be a straightforward solution to this problem but a quick calculation shows that this is unfeasible. Indeed, if the probability of a the two daughter RNA segments being segregated into the two daughter cells is 1\2, then the probability that, say, 100 RNA segments in a reasonable multipartite genome are all correctly segregated (i.e., none of the segments is missing in either of the daughter cells) is (1/2)¹⁰⁰ ≈ 10^-30. Thus, ploidy of ~10³⁰ would be required for accurate genome segregation; in other words, the RNA cell would have to contain many tons of RNA. Thus, it is much more likely that the first fully-fledged cells already had DNA genomes resembling (even if quantitatively simpler than) those of modern archaea and bacteria. Forterre's proposal is based on the well-known observation that the DNA replication systems of archaea and bacteria are, largely, unrelated, making it unlikely that LUCA had a DNA genome [58-60]. We believe that a far more plausible implication of the disparity of DNA replication machineries in archaea and eukaryotes is a non-cellular LUCA, a hypothesis that finds crucial support in the fact that the membrane lipids and membrane biogenesis systems, as well as cell walls, are also distinct and, largely, unrelated in archaea and bacteria [39,61-63].

For efforts on reconstructing the non-cellular LUCA, a key guiding principle is that, although modern-type cells, presumably, did not exist at this stage of evolution, some form of compartmentalization was required to ensure concentrations of substrates and genetic elements sufficient for effective replication and, consequently, evolution. Furthermore, as discussed in greater detail below, compartmentalization is a necessary condition of selection among evolving ensembles of genetic elements. Following this principle, a specific model has been elaborated under which early evolution, from abiogenic syntheses of complex organic molecules to the emergence of archaeal and bacterial cells, unraveled within networks of inorganic compartments that are found at hydrothermal vents and consists, primarily, of iron sulfide [39,63]. Here, in order to be concrete, we attempt a reconstruction of the earliest events in the evolution of primordial virus-like entities within the framework of this model although our general conclusions do not seem to hinge on any particular model of organization of the ancient, precellular life.

Second, the model by viral overtake meets a stumbling block: if three different viruses brought the three distinct DNA replication machineries into the three cell lineages, why have not the hallmark genes that are essential for viral DNA replication, in particular S3H and the viral-type primase, entered the genomes of any of those cellular lineages? One could argue that, in each case, an unusual virus, not carrying any of these hallmark genes, was involved, but this happening three times independently stretches credulity.

The third major issue that is ignored in Forterre's hypothesis and similar, three-domain scenarios of cellular evolution [64,65], is the readily demonstrable relationship between eukaryotes and the two prokaryotic domains. This relationship follows the divide between the two principal functional categories of genes in the eukaryotic cell, the informational genes that are, almost invariably, most closely related to archaeal homologs, and the operational genes that are of bacterial provenance [66-68]. By far the simplest, most parsimonious explanation of this dichotomy is that the eukaryotic cell emerged as a result of a symbiosis between an archaeon and a bacterium. Given the overwhelming evidence of the origin of mitochondria and related organelles (hydrogenosomes and mitosomes) from α-proteobacteria, the nature of the partners and the direction of the symbiosis appear clear: an α-proteobacterium invaded an archaeal cell [69-71]. In addition to the genomic evidence, there is a clear biochemical rationale for the symbiosis to actually comprise the onset of eukaryogenesis such that the invading α-proteobacterium became an anaerobic symbiont that originally supplied hydrogen to the methanogenic, archaeal host and subsequently gave rise to aerobic mitochondria [71]. The massive transfer of the symbiont's genes to the host genome gave rise to the mosaic provenance of the eukaryotic genomes. The argument against the symbiotic models of eukaryogenesis and for three-domain models stems primarily from the existence of numerous eukaryote-specific proteins (ESPs) without readily detectable prokaryotic homologs [64,72,73]. However, the abundance of ESPs hardly can be taken as evidence of a third domain of life, distinct from archaea and bacteria, and comprising the pre-symbiosis eukaryotic line of descent. First, although many of the ESPs are proteins with important biological functions, they do not belong to the core of either the informational or the metabolic proteins repertoires of eukaryotes that are of archaeal and bacterial descent, respectively. Second, more often than not, prokaryotic homologs of an apparent ESP can be identified by using sensitive techniques of protein motif analysis and, particularly, structural comparison [74]. Thus, it seems likely that many, if not most, ESPs are cases of accelerated evolution in eukaryotes which accompanies the emergence of novel, eukaryote-specific functional systems, such as the cytoskeleton and ubiquitin signaling. The prevalence and nature of "true" eukaryotic innovations, i.e., proteins that evolved through processes other than descent with (perhaps, radical) modification from archaeal or bacterial proteins, remain to be characterized. Clearly, however, even if such novelties turn out to be numerically prominent, they are, typically, involved in ancillary functions and hardly can be construed as the heritage of a distinct primary line of cell evolution [74].

As indicated above, we present a scenario of viral origins under the recent model of the emergence of cells and genomes within networks of inorganic compartments [39,63]. These compartments are envisaged being inhabited by diverse populations of genetic elements, initially, segments of self-replicating RNA, subsequently, larger and more complex RNA molecules encompassing one or a few protein-coding genes, and later yet, also DNA segments of gradually increasing size. Thus, early life forms, including LUCA, are perceived as ensembles of genetic elements inhabiting a system of inorganic compartments. This model explains the lack of homology between the membranes, membrane biogenesis systems, and the DNA replication machineries of archaea and bacteria by delineating a LUCA that had neither a membrane nor DNA replication. The model also outlines the processes that might have enabled selection and evolutionary change in such a system. Under this scenario, a transition gradually took place from selection at the level of individual genetic elements to selection for ensembles of such elements encoding both enzymes directly involved in replication and proteins responsible for accessory functions, such as translation and nucleic acid precursor synthesis. From selection for gene ensembles, there is a direct path to selection for compartment contents such that compartments sustaining rapid replication of genetic elements would infect adjacent compartment and, effectively, propagate their "genomes" [39].

This model implies that, at the early stages of evolution, including the LUCA stage, the entire genetic system was, in a sense, "virus-like". Initially, all RNA segments in the population would be completely selfish, and there would be no distinction between parasitic, "viral" elements and those that would give rise to genomes of cellular life forms. However, this distinction would emerge as soon as the first "selfish cooperatives" – relatively stable ensembles of co-inherited genetic elements [39] – evolve. Inevitably, the selfish cooperatives would harbor parasitic genetic elements that would divert the resources of the ensemble for their own replication. The separation of parasitic elements from cooperators would be cemented by the emergence of physical linkage of the latter: genes that encode components of the replication machinery and are physically linked to genes for accessory functions would be locked into ensembles of cooperators, whereas solitary genes (or small gene arrays) encoding replication functions would occupy the parasitic niche. From a complementary standpoint, within the inorganic compartment model, the distinction between the progenitors of viruses and the ancestors of cellular life forms may be described as one between primary infectious agents, that routinely moved between compartments, and increasingly complex and "sedentary" ensembles of selfish cooperators that would spill from one compartment into another on much rarer occasions.

Like other models of the early stages of evolution of biological complexity, this scenario faces the problem of takeover by selfish elements [76-78]. If the primordial parasites become too aggressive, they would kill off their hosts within a compartment and could survive only by infecting a new compartment. Devastating "epidemics" sweeping through entire networks and eventually eliminating all their content are imaginable, and indeed, this might have been the fate of many, if not most, primordial "organisms". The evolution of those primordial life forms that survived would involve, firstly, emergence of temperate virus-like agents that do not kill the host, and secondly, early invention of primitive defense mechanisms, likely, based on RNA interference (RNAi). The ubiquity of both temperate selfish elements and RNAi-based defense systems [79-81] in all life forms is compatible with the origin of these phenomena at a very early stage of evolution.

From the very beginning, the parasites would have different levels of genome complexity, from the minimal replicable unit, like in viroids, to more complex entities that might encode one or more proteins facilitating their own replication. Capsid proteins, in particular, JRC were the crucial innovation that might have led from virus-like genetic elements to entities resembling true viruses (as long as viruses are defined as intracellular parasites, we are not entitled to speak of viruses evolving prior to the appearance of membrane-bounded cells; however, this is a semantic problem only). Under this model, the advantage conferred on a selfish genetic element by encapsidation is obvious: stabilization of the genome and increased likelihood of transfer between compartments and even between different networks of compartments. Given this advantage of the capsid and the extensive reassortment and recombination that is thought to have reigned in the primordial gene pool, it is not surprising that different genes that could contribute to virus replication, such as polymerases, helicases, ATPases, and nucleases, combined with the JRC gene on multiple occasions, and several such combinations were propagated by selection to give rise to distinct viral lineages. Alternative capsid proteins that form helical capsids in diverse RNA and DNA viruses [82] might have evolved already at this early stage of evolution. However, the capsid is by no means a pre-requisite of evolutionary success for selfish genetic elements. Indeed, virus-like entities, such as the prokaryotic retroelements (retrons and group II introns) and various small plasmids, have maintained the capsid-less, selfish life style from their inferred emergence within the primordial gene pool to this day.

Thus, under this scenario, virus-like entities are older than typical, modern-type cells: agents resembling modern viruses, some of them even with capsids, existed before the emergence of membrane-bounded cells with large dsDNA genomes. Moreover, these agents would, effectively, have a virus-like life style because the primordial life forms are perceived as non-cellular but compartmentalized and would support ancient virus-like agents moving between compartments much like modern cells support viruses. The origin of ancient virus-like agents, the ancestors of the major classes of modern viruses, would follow the stages of evolution of genetic systems: RNA-only selfish agents resembling group I introns or viroids might have emerged in the primitive RNA world. The RNA-protein world begot RNA viruses, the postulated stage of an RT-based, mixed RNA-DNA system spawned retroid elements, and the DNA stage yielded several lineages of DNA viruses (Fig. ​(Fig.2).2). It appears likely that the primordial gene pool harbored an extraordinary variety of virus-like entities. When archaeal and bacterial cells escaped the compartment networks, they would carry with them only a fraction of these viruses that subsequently evolved in the cellular context to produce the modern diversity of the virus world.

General considerations on the course of evolution from the RNA world to modern-type systems, together with the spread of different virus classes among modern life forms (Table ​(Table11 and Fig. ​Fig.1),1), suggest that the following classes of virus-like selfish elements emerged from the primordial gene pool and infected the first bacteria and/or archaea (Fig. ​(Fig.2):2): i) RNA-only elements, such as group I introns, ii) positive-strand RNA viruses (and, possibly, dsRNA viruses as well), iii) retroid elements similar to prokaryotic retrons and group II introns, iv) small RCR replicons, v) at least one, and probably, more groups of larger, dsDNA viruses ancestral to the order Caudovirales (tailed phages). Incidentally, the logic of this inference suggests that the dsDNA viruses emerging from the primordial gene pool already possessed DNA polymerases and, accordingly, that the DNA polymerase is a true viral hallmark gene, notwithstanding the difficulty in obtaining strong evidence of this (see above).

The manuscript by Koonin et a. describes a scenario for virus evolution that links the origin of virus to the early evolution and origin of cells. In particular, the authors suggest that viruses and phages are descendents of selfish genetic elements that were already present before the evolution of cells and genomes. The argument is based on the wide, but not universal, distribution of viral "signature" genes, and agrees with hypercycle models of early molecular evolution that showed that these early networks are prone to infection by molecular parasites. The basic hypothesis presented in this manuscript is reasonable, well developed and provides a good alternative to the scenarios that describe virus' origins as genes escaped from cellular organisms.

The manuscript treats the early evolution of viruses as a speculative topic. Given that evolution is traced back to the origin of cells, this might appear justified; however, I find the article more speculative than necessary. The authors link virus origins to one particular model of cellular origins and early evolution, the authors chose the scenarios by Martin, Muller, and Russell (see my references 1 & 2). This leads to unnecessarily detailed speculation. These models made an important contribution in detailing possible pathways to cellular life and towards the eukaryotic cell, but many alternative syntrophic relationships at the root of the eukaryotes were suggested, for examples see my refs. 3–6. Furthermore, some details of the scenario followed in this manuscript have been debated in the past (e.g., molecular phylogenies do not indicate a close relationship between eukaryotes and methanogenes); an RNA based genome might not necessarily be less complex, because early RNA polymerases were inferred to be error correcting (see my ref. 7); and the assumptions that the authors make for the most recent common ancestor of bacteria, archaea and eukaryotes contradict much of what was learned about early evolution during the previous decades: Molecular evidence points towards energy coupling membranes being already present in the MRCA of all known life (see discussion in my ref. 8). Apparently, the MRCA of all known cellular organisms was not devoid of membranes, but already had a complex targeting machinery for membrane proteins (my ref. 9), terminal oxidases (my ref. 10 and Simonetta Gribaldo, pers. communication) and ATP synthases driven by transmembrane electricochemical ion gradients (my ref. 11). All of these systems apparently predated the MRCA of all known cellular organism. The idea of a primitive, pre-cellular common ancestor dates back to the first molecular trees of life, when Fox and Woese concluded that this organism might have been a progenote, i.e. an organism without a tight coupling between geno- and phenotype (my refs. 12 & 13). While a pre-cellular organism with a distributed, communal genome likely was a stage in early cellular evolution, molecular evidence suggests that at the time of the organismal MRCA cellular structures were much more advances than envisioned in the scenario described by the authors. Horizontal gene transfer complicates a simple back extrapolation, but the molecular phylogenies of ATP synthases, elongation factors, ribosomes and signal recognition proteins are in surprising agreement, suggesting that with few recognizable exceptions, the genes in question were transferred only between closely related organisms. One way to arrive at a more primitive MRCA of the three domains is to place the root of the tree of life on the eukaryotic branch as suggested by Forterre and collaborators (my ref. 14). However, several shared derived characters of ATPases (my ref. 15) and elongation factors (my ref. 16) suggest that the root is located outside the clade comprised by the archaea and the eukaryotic nucleocytoplasm.

A frequent argument in favor of a pre-cellular MRCA, also invoked in the present manuscript, is based on the different lipid and cell wall composition of archaea, bacteria and eukaryotes. I think this is a red herring: All three domains synthesize isoprenoids (and while some of the enzymes are different between the two domains, the pathway as a whole and some of the enzymes appear to be homologous) (my refs. 17 & 18); furthermore, all cells use polyprenols like dolichol to transport sugars through membranes (either as activated cell wall precursors or for glycosilation reactions inside the ER or in the periplasmic space), indicating that the ability to synthesize long chain branched aliphatic alcohols was around early; and S-layer proteins are considered by some as the likely ancestral cell wall material (my ref. 19). If the ester linked fatty acid based membrane lipids were a later bacterial invention, it would not be surprising to find this pathway also in eukaryotes, because all known eukaryotes apparently evolved from ancestors that once possessed mitochondria (my ref. 20).

While I do not agree with some of the details of the described scenario for cellular evolution, these details are not crucial to the central thesis of the manuscript (there is no arrow in figure ​figure22 that connects the MRCA of all life and the bacterial and archaeal MRCAs to the virus world). The proposed hypothesis on the origin of viruses depends only on the presence of a progenote stage in early evolution, regardless whether this stage was part of the "stem" leading to the organismal MRCA, or whether this stage was reached independently by the lineages leading to the archaeal and bacterial domains.

Author response:We agree with Gogarten that our concept of the ancient virus world does not critically depend on the nature of LUCA (MRCA of all modern cells); what is actually required is an advanced, diverse pre-cellular pool of genetic elements. Indeed, this is a useful point to make and we do so explicitly in the revised manuscript. It might be useful to note that the pre-cellular stage of evolution in our model does not seem to be the progenote (Woese, Fox, 1977, J. Mol. Evol. 10: 1–6) in the original sense because that latter was supposed to possess a primitive, imprecise translation system which would not work for the level of pre-cellular complexity envisaged here.

Since the nature of LUCA is not central here, it is not the place to present the argument for a pre-cellular LUCA that has been discussed previously [37]and, more briefly, in this paper. Just in a nutshell: we do not believe that the use of non-homologous pathways for lipid biosynthesis by archaea and bacteria is a "red herring"; on the contrary, it is a major conundrum in need of a solution. Yes, all bacteria do synthesize isoprenoids, and some of them do so with the use of the archaeal enzymatic machinery, probably, acquired via HGT. Whether or not the classical bacterial pathway of isoprenoid biosynthesis derives from a common ancestor with the archaeal pathway is a more complex matter. What is important, however, is that bacteria never use isoprenoids to build their membranes. So the notion of a cellular LUCA would require displacement of the ancestral, archaeal-type, isoprenoid-based membrane by the newly emerged, bacterial-type fatty-acid-based membrane, without elimination of the isoprenoid biosynthesis pathway that would then serve other functions (as they do in modern bacteria). Not an impossible scenario in itself but a mechanistically challenging one, and with the underlying selective forces utterly mysterious.

Gogarten responds in a second review: I fail to see a mechanistic challenge. To a large extent lipids based on fatty acids, long chain alcohols, and even non-biogenic lipids, for example extracted from the Murchison meteorite (my ref. 21), appear mechanistically equivalent.

Author response:However, the lipid argument is not the only one for a non-cellular LUCA. The lack of homology between the core components of the DNA replication machineries in archaea and bacteria, which implies a fragmented RNA genome in LUCA, is equally important. This effectively rules out accurate genome segregation and does not bode well for a cellular LUCA at all. We certainly do not claim to "know" what LUCA was like but we do perceive the non-cellular model discussed in [37]to be the current solution of choice.

Gogarten responds in a second review: Even an RNA based genome might have been less fragmented than assumed (see above), furthermore the lack of perceived sequence homology between the bacterial and archaeal/eucaryal DNA replication machinery could be due to divergence, not lack of shared ancestry. Functionally the processes and sub-processes in the replication fork are very similar in all three domains of life, which seems to be difficult to explain by convergent evolution.

Alternative explanations for the features that were used to argue for a non-cellular MRCA exist; in contrast, the findings that indicate a cellular MRCA of the three domains (e.g., the machineries used in chemiosmotic coupling, and for the targeting of membrane proteins apparently predate the MRCA of the three domains, see above) at present have not been reconciled with a non-cellular entity. Therefore, at present a pre-cellular MRCA of the three domains (LUCA) appears at odds with the available data. I do not perceive this scenario as the solution of choice.

Gogarten first review continues: Other suggestion:

Add additional citations: To me the idea that virus and phage evolution began early in the evolution of life appears very reasonable, and I would be surprised if others had not formulated similar ideas in the past.

Author response: We believe that the notion of the virus world as explicated here is new. The idea of a primordial origin of virus-like entities, of course, is old, even if unpopular lately (at least prior to the work of the Bamford group on the JRC structure in diverse viruses and the discovery of the mimivirus – all this is cited here). We cite the classic textbook of Luria and Darnell [40]which offers an insightful discussion of the early ideas in this area. In the revision, we added the citation of Felix D'Herelle's 1922 book which is where the idea that viruses might precede cells in evolution, probably, was proposed for the first time [46].

Gogarten responds in a second review: The addition of the D'Herelle citation is an excellent choice, the following might be interesting as well, it seems more similar to the ideas developed in the manuscript: According to Sapp (my ref. 22) the idea of early co-existence of viruses and cells was expressed by Peter Raven in a letter to R. E. Buchanan on November 3, 1970 "Raven suggested that viruses, probably as old as life itself, might be regarded as by-products of bacterial reproduction, in which segments of DNA or RNA protected with protein coats spread from cell to cell, directing the host cell's metabolism to reproduce more of the viral DNA or RNA."

Gogarten's first review continues: Section on "The primordial gene pool: the crucible of the major virus lineages", last paragraph: Why would the transfer need to be rampant? The connection to a non-cellular model for early evolution could be better developed. Pre-cells, or cells with small, possibly partial genomes (my ref. 23) should do just fine for the indicated stages, as long as there is a moderate level of transfer allowing for recombination and for molecular parasites to evolve.

2. Martin W, Muller M: The hydrogen hypothesis for the first eukaryote. Nature 1998, 392(6671):37–41.

3. Searcy DG: Origins of mitochondria and chloroplasts from sulfur based symbiosis. In: The Origin and Evolution of the Cell. Edited by Hartman H, Matsuno, K.: World Scientific; 1992: 47–78.

4. Margulis L: Symbiosis in Cell Evolution: Microbial Communities in the Archean and Proterozoic Eons, 2nd edn: W H Freeman & Co; 1995.

5. Searcy DG: Metabolic integration during the evolutionary origin of mitochondria. Cell Res 2003, 13(4):229–238.

6. Lopez-Garcia P, Moreira D: Metabolic symbiosis at the origin of eukaryotes. Trends Biochem Sci 1999, 24(3):88–93.

7. Poole AM, Logan DT: Modern mRNA proofreading and repair: clues that the last universal common ancestor possessed an RNA genome? Mol Biol Evol 2005, 22(6):1444–1455.

8. Gogarten JP, Taiz L: Evolution of proton pumping ATPases: Rooting the tree of life. Photosynthesis Research 1992, 33:137–146.

9. Gribaldo S, Cammarano P: The root of the universal tree of life inferred from anciently duplicated genes encoding components of the protein-targeting machinery. Journal Of Molecular Evolution 1998, 47(5):508–516.

10. Castresana J, Lubben M, Saraste M, Higgins DG: Evolution of cytochrome oxidase, an enzyme older than atmospheric oxygen. Embo J 1994, 13(11):2516–2525.

11. Gogarten JP, Kibak H, Dittrich P, Taiz L, Bowman EJ, Bowman BJ, Manolson MF, Poole RJ, Date T, Oshima T et al: Evolution of the vacuolar H+-ATPase: implications for the origin of eukaryotes. Proc Natl Acad Sci USA 1989, 86(17):6661–6665.

12. Woese CR, Fox GE: Phylogenetic structure of the prokaryotic domain: the primary kingdoms. Proc Natl Acad Sci USA 1977, 74(11):5088–5090.

13. Woese CR, Fox GE: The concept of cellular evolution. J Mol Evol 1977, 10(1):1–6.

14. Forterre P, Philippe H: Where is the root of the universal tree of life? BioEssays 1999, 21(10):871–879.

15. Zhaxybayeva O, Lapierre P, Gogarten JP: Ancient gene duplications and the root(s) of the tree of life. Protoplasma 2005, 227(1):53–64.

16. Skophammer RG, Herbold CW, Rivera MC, Servin JA, Lake JA: Evidence that the Root of the Tree of Life Is Not within the Archaea. Mol Biol Evol 2006, 23(9):1648–1651.

17. Boucher Y, Kamekura M, Doolittle WF: Origins and evolution of isoprenoid lipid biosynthesis in archaea. Mol Microbiol 2004, 52(2):515–527.

18. Boucher Y, Doolittle WF: The role of lateral gene transfer in the evolution of isoprenoid biosynthesis pathways. Mol Microbiol 2000, 37(4):703–716.

19. Claus H, Akca E, Debaerdemaeker T, Evrard C, Declercq JP, Harris JR, Schlott B, Konig H: Molecular organization of selected prokaryotic S-layer proteins. Can J Microbiol 2005, 51(9):731–743.

20. Keeling PJ, Burger G, Durnford DG, Lang BF, Lee RW, Pearlman RE, Roger AJ, Gray MW: The tree of eukaryotes. Trends Ecol Evol 2005, 20(12):670–676.

21. Deamer DW: Role of amphiphilic compounds in the evolution of membrane structure on the early earth. Orig Life Evol Biosph 1986, 17(1):3–25.

22. Sapp J: The prokaryote-eukaryote dichotomy: meanings and mythology. Microbiol Mol Biol Rev 2005, 69(2):292–305.