The origin of eukaryotes is a huge enigma and a major challenge for evolutionary biology 123. There is a sharp divide in the organizational complexity of the cell between eukaryotes, which have complex intracellular compartmentalization, and even the most sophisticated prokaryotes (archaea and bacteria), which do not 456. A typical eukaryotic cell is about 1,000-fold bigger by volume than a typical bacterium or archaeon, and functions under different physical principles: free diffusion has little role in eukaryotic cells, but is crucial in prokaryotes 78. The compartmentalization of eukaryotic cells is supported by an elaborate endomembrane system and by the actin-tubulin-based cytoskeleton 910. There are no direct counterparts of these organelles in archaea or bacteria. The other hallmark of the eukaryotic cell is the presence of mitochondria, which have a central role in energy transformation and perform many additional roles in eukaryotic cells, such as in signaling and cell death.

The conservation of the major features of cellular organization and the existence of a large set of genes that are conserved across eukaryotes leave no doubt that all extant eukaryotic forms evolved from a last eukaryote common ancestor (LECA; see below). All eukaryotes that have been studied in sufficient detail possess either mitochondria or organelles derived from mitochondria 111213, so it is thought that LECA already possessed mitochondria (see below). Plants and many unicellular eukaryotes also have another type of organelle, plastids.

The organizational complexity of the eukaryotic cells is complemented by extremely sophisticated, cross-talking signaling networks 14. The main signaling systems in eukaryotes are the kinase-phosphatase machinery that regulates protein function through phosphorylation and dephosphorylation 15161718; the ubiquitin network that governs protein turnover and localization through reversible protein ubiquitylation 192021; regulation of translation by microRNAs 222324; and regulation of transcription at the levels of individual genes and chromatin remodeling 24252627. Eukaryotes all share the main features of cellular architecture and the regulatory circuitry that clearly differentiate them from prokaryotes, although the ancestral forms of some signature eukaryotic systems are increasingly detected in prokaryotes, as discussed below. Phylogenomic reconstructions show that the characteristic eukaryotic complexity arose almost 'ready made', without any intermediate grades seen between the prokaryotic and eukaryotic levels of organization 9282930. Explaining this apparent leap in complexity at the origin of eukaryotes is one of the principal challenges of evolutionary biology.

The key to the origin of eukaryotes will undoubtedly be found using comparative genomics of eukaryotes, archaea and bacteria. Complete genome sequences from all three domains of cellular life are accumulating exponentially, albeit at markedly different paces. As of March 2010, the NCBI genome database contained over 1,000 bacterial genomes, about 100 archaeal genomes, and about 100 genomes of eukaryotes 31. Here, I discuss some of the main insights that have come from comparative analysis of these genomes, which may help to shed light on the origin and the early stages of evolution of eukaryotes. So far, the comparative genomics era has brought fascinating clues but no decisive breakthrough.

Although several eukaryotic kingdoms, such as animals, fungi, plants and ciliates, are well defined and seem to be monophyletic beyond reasonable doubt, deciphering the evolutionary relationships between these kingdoms and numerous other groups of unicellular eukaryotes (also called protists) turned out to be daunting. For many years, evolutionary biologists tended to favor the so called crown group phylogeny 232. The 'crown' of this evolutionary tree included animals (Metazoa) and plants (Viridiplantae), fungi and various assortments of protists, depending on the methods used for tree construction 3334. The rest of the protists, such as microsporidia, diplomonads and parabasalia, were considered 'early branching eukaryotes'; for some of them, this conclusion was reached because they appeared to lack mitochondria and were therefore thought to have evolved before the mitochondrial symbiosis. The scenario resulting from the crown group phylogeny was called the archezoan scenario: the archaezoan was defined as a hypothetical ancestral form that lacked mitochondria but possessed the other signature features of the eukaryotic cell. However, during the past decade, the early branching groups have lost their positions at the root of the eukaryotic tree, one after another 353637. The improved taxon sampling as a result of genome sequencing together with new, more robust methods for phylogenetic analysis indicate that the deep placing of these groups seen in early trees was a long-branch artifact caused by the fast evolution of the respective organisms 373839. At the same time, comparative-genomic and ultrastructural studies destroyed the biological underpinning of the near-root positions of the (former) early branching groups of protists by showing that none of them ancestrally lack mitochondria, as they all have genes of apparent mitochondrial origin and mitochondria-related organelles, such as hydrogenosomes and mitosomes 111213 40.

There are therefore no grounds to consider any group of eukaryotes primitive, a presymbiotic archezoan. Rather, taking into account the small genomes and high rate of evolution characteristic of most of the protist groups thought to be early branching, and their parasitic lifestyle, it is becoming increasingly clear that most or perhaps all of them evolved from more complex ancestral forms by reductive evolution 3739. Reductive evolution refers to the evolutionary modality typical of parasites: they tend to lose genes, organelles and functions when the respective functionalities are taken over by the host. So the archezoan (crown group) phylogeny seems to have been disproved, and deep phylogeny and the theories of the origin of eukaryotes effectively had to start from scratch.

This time phylogenomic approaches were mainly used, that is, phylogenetic analysis of genome-wide sets of conserved genes; this was made possible by the much larger number of genomes that had been sequenced 4142. The key accomplishment at this new stage was the proposal of 'supergroups' of eukaryotes that are suggested to combine highly diverse groups of organisms in a monophyletic group 36434445. Most of the phylogenomic analyses published so far converge on five supergroups (or six if the Amoebozoa and Opisthokonts do not form a single supergroup, the Unikonts; Figure 1). Although proving monophyly is non-trivial for these groups 464748, the general structure of the tree, with a few supergroups forming a star-like phylogeny (Figure 1), is reproduced consistently, and the latest results 49505152 seem to support the monophyly of the five supergroups.

The relationship between the supergroups is a formidable problem as the internal branches are extremely short, suggesting that the radiation of the supergroups occurred rapidly (on the evolutionary scale), perhaps resembling an evolutionary 'big bang' 535455. Two recent, independent phylogenetic studies 5152 each analyzed over 130 conserved proteins from several dozen eukaryotic species and, after exploring the effects of removing fast-evolving taxa, arrived at a three-megagroup structure of the eukaryotic tree. The megagroups consist of Unikonts, Excavates, and the assemblage of Plantae, Chromalveolata and Rhizaria 5152.

Furthermore, there have been several attempts to infer the position of the root of the eukaryotic tree (Figure 1). The first alternative to the crown group tree was proposed by Cavalier-Smith and coworkers 565758, who used rare genomic changes (RGCs) 59, such as the fusion of two enzyme genes 5657 and the domain structure of myosins 58, to place the root between the Unikonts and the rest of eukaryotes (I (red arrow) in Figure 1). This separation seems biologically plausible because Unikont cells have a single cilium, whereas all other eukaryotic cells have two. However, this conclusion could be suspect because the use of only a few RGCs makes it difficult to rule out homoplasy (parallel emergence of the same RGC, such as gene fusion or fission, in different lineages). Rogozin and coworkers 60 used a different RGC approach based on rare replacements of highly conserved amino acid residues requiring two nucleotide substitutions and inferred the most likely position of the root to be between Plantae and the rest of eukaryotes (II (green arrow) in Figure 1). Again, this seems biologically plausible because the cyanobacterial endosymbiosis that gave rise to plastids occurred on the Plantae lineage.

The controversy about the root position and the lack of consensus regarding the monophyly of at least some of the supergroups, let alone the megagroups, indicate that, despite the emerging clues, the deep phylogeny of eukaryotes currently should be considered unresolved. In a sense, given the likely 'big bang' of early eukaryote radiation, the branching order of the supergroups, in itself, might be viewed as relatively unimportant 61. However, the biological events that triggered these early radiations are of major interest, so earnest attempts to resolve the deepest branches of the eukaryotic tree will undoubtedly continue with larger and further improved datasets and methods.

Comparative analysis of representative genomes from different eukaryotic supergroups enables the reconstruction of the gene complement of LECA using maximum parsimony (MP) or more sophisticated maximum likelihood (ML) methods 626364. Essentially, genes that are represented in diverse extant representatives of different supergroups, even though lost in some lineages, can be mapped back to LECA. The results of all these reconstructions consistently point to a complex LECA, in terms of both the sheer number of ancestral genes and, perhaps even more importantly, the ancestral presence of the signature functional systems of the eukaryotic cell (see below). A MP reconstruction based on phyletic patterns in clusters of orthologous genes of eukaryotes mapped 4,137 genes to LECA (Figure 1) 636566. Remarkably, an even simpler estimation, based on the recent analysis of the genome of Naegleria gruberi, the first sequenced genome of a free-living excavate 67, revealed about a nearly identical number of genes, 4,134, that are shared by Naegleria and at least one other supergroup of eukaryotes, suggesting that these genes are part of the LECA heritage (Figure 1). Such estimates are highly conservative as they do not account for lineage-specific loss of ancestral genes, a major aspect in the evolution of eukaryotes. Indeed, even animals and plants, the eukaryotic kingdoms that seem to be the least prone to gene loss, have still lost about 20% of the putative ancestral genes identified in the unicellular Naegleria (Figure 1). Given that the current estimate for the gene complement of LECA must be conservative, the genome of LECA is likely to have been as complex as those of typical extant free-living unicellular eukaryotes 68.

This conclusion is supported by reconstructions from comparative genomics of the ancestral composition of the key functional systems of the LECA, such as the nuclear pore 2869, the spliceosome 29, the RNA interference machinery 70, the proteasome and the ubiquitin signaling system 71, and the endomembrane apparatus 10. The outcomes of these reconstructions are all straightforward and consistent, even when different topologies of the phylogenetic tree of eukaryotes were used as the scaffold for the reconstruction: LECA already possessed all these structures in its fully functional state, possibly as complex as the counterparts in modern eukaryotes.

Reconstruction of other aspects of the genomic composition and architecture of LECA similarly points to a highly complex ancestral genome. Comparative-genomic analysis of intron positions in orthologous genes within and between supergroups suggests high intron densities in the ancestors of the supergroups and in LECA, at least as dense as in modern free-living unicellular eukaryotes 72737475. A systematic analysis of widespread gene duplications in eukaryotes indicates that hundreds of duplications predate LECA, especially duplications of genes involved in protein turnover 636566. Taken together, these results clearly indicate that LECA was a typical, fully developed eukaryotic cell. The subsequent evolution of eukaryotes has seemingly shown no consistent trend toward increased complexity, except for lineage-specific embellishments, such as those seen in animals and plants. There was obviously an important stage of evolution on the 'stem' of eukaryotes, after they first evolved but before LECA, which included extensive duplication of numerous essential genes, so that the set of ancestral genes approximately doubled 636566.

Eukaryotes are hybrid organisms in terms of both their cellular organization and their gene complement. All eukaryotes seem to possess mitochondria or related organelles derived from α-proteobacteria, whereas Plantae and many groups of Chromalveolata additionally have cyanobacteria-derived plastids 7677. The gene complement of eukaryotes is an uneven mix of genes of apparent archaeal origin, genes of probable bacterial origin, and genes that so far seem eukaryote-specific, without convincing evidence of ancestry in either of the two prokaryote domains (Figure 2). Paradoxical as this might appear, although trees based on rRNA genes and concatenated alignments of information-processing proteins, such as polymerases or splicing proteins, both put archaea and eukaryotes together, genome-wide analyses consistently and independently show that there are three or more times more genes with bacterial homologs than with archaeal homologs 62637879 (Figure 2). The archaeal subset is strongly enriched in information processing functions (translation, transcription, replication, splicing), whereas the bacterial subset consists largely of metabolic enzymes 6278 (see below for more details).

At a coarse level, these observations are best compatible with genome fusion scenarios 7980 whereby the eukaryotic genome emerged through a fusion between two ancestral genomes, an archaeal or archaea-related one, and a bacterial, most likely α-proteobacterial, one, given the well-established ancestry of the mitochondrial endosymbiont 81. However, attempts to pinpoint the specific archaeal and bacterial 'parents' of eukaryotes reveal complicated evolutionary relationships. Although many of the bacterial-like genes in eukaryotes have α-proteobacterial homologues, these are far from dominant amongst the bacterial-like genes which show apparent evolutionary affinities with a variety of bacterial groups (Figure 2). An important cause of this complicated breakdown of the bacterial-like component of the eukaryotic gene complement is the large size of the α-proteobacterial pangenome, that is, of the combined genes found in all α-proteobacteria, and the associated diversity of the gene sets in individual members of this group 82. Thus, without knowing the exact identity within the α-proteobacteria of the bacterial endosymbiont that gave rise to the eukaryotic mitochondria, it is hard to delineate its genetic contribution. Apart from this uncertainty about the gene complement of the endosymbiont, it is impossible to rule out multiple sources of the bacterial-like genes in eukaryotes 83, which may have origins other than the genome of the bacterial endosymbiont. In particular, whatever the actual nature of the archaeal-like ancestor, it probably lived at moderate temperatures and non-extreme conditions and was consequently in contact with a diverse bacterial community. Modern archaea with such lifestyles have numerous genes of diverse bacterial origins, indicating extensive horizontal acquisition of genes from bacteria 8485. Thus, the archaeal-like host of the endosymbiont could have already had many bacterial genes, partly explaining the observed pattern.

The case of the archaeal(-like) parent is far more difficult than that of the bacterial ancestor(s) as there are no data on the ancestral lineage that would parallel the unambiguous origin of mitochondria from α-proteobacteria. Phylogenomic studies using different methods point to different archaeal lineages - Crenarchaeota 8687, Euryarchaeota 88, or an unidentified deep branch 8990 - as the candidates for the eukaryote ancestor (Figure 3). Unequivocal resolution of such deep evolutionary relationships is extremely difficult. Moreover, at least one of these analyses 89 explicitly suggests the possibility that the archaeal heritage of eukaryotes is genuinely mixed, with the largest contribution coming from a deep lineage, followed by the contributions from Crenarchaeota (Thaumoarchaeota) and the Euryarchaeota (Figure 3). In the next section I examine the possibility of multiple archaeal and bacterial ancestors of the eukaryotes with respect to distinct functional systems of eukaryotic cells.

Some of the most compelling indications on the course of evolution and the nature of ancestral forms come from signature genes that are uniquely shared by two or more major lineages and from detailed evolutionary analysis of well characterized functional systems, in particular the signature systems of the eukaryotic cell. Comparative genome sequence analysis has revealed that some of the key molecular machines of the eukaryotes, and not only those directly involved in information processing, can be confidently derived from archaeal ancestors (Table 1 and Figure 4). Strikingly, this archaeal heritage seems to be patchy with respect to the specific origins, with apparent evolutionary affinities to different groups of archaea (Table 1 and Figure 4). For instance, comparative analysis of the translation system components tends to suggest an affinity between eukaryotes and Crenarchaeota 91. Similarly, the core transcription machinery of eukaryotes shares some important proteins with Crenarchaeota, Thaumarchaeota and Korarchaeota, to the exclusion of Euryarchaeota 929394. By contrast, the histones, the primary components of nucleosomes, are missing in most of the Crenarchaeota but invariably conserved in Euryarchaeota (and also present in Korarchaeum and some Thaumarchaeota) 95.

Eukaryotic cell division components are also conserved in several but not all of the major archaeal lineages. For example, homologs of the ESCRT-III complex, which performs key roles in vesicle biogenesis and cytokinesis in eukaryotes, are responsible for cell division in the Crenarchaeota but are missing in most of the Euryarchaeota, which possess a bacterial-like division mechanism using the GTPase FtsZ, a distant homolog of tubulin 9697. However, a few members of the Euryarchaeota have both systems, with FtsZ probably responsible for division and ESCRT-III for vesicle biogenesis 98.

Eukaryote B-family DNA polymerases, a group of four paralogs that are collectively responsible for genome replication, show a complex pattern of ancestry (Figure 4): one branch of the eukaryotic polymerases seems to have evolved from archaeal PolBI, which is conserved in all archaea, whereas the other branch appears to derive from the Crenarchaea-specific PolBII 99100. Surprisingly, the eukaryotic polymerases additionally contain a Zn-finger domain homologous to that of PolD, which is restricted to Euryarchaeota 100; furthermore, the small subunits of eukaryotic Polα and Polδ are inactivated derivatives of the exonuclease subunit of PolD 101.

Another major theme emerging from these studies is the bacterial contribution and the formation of archaeao-bacterial chimeras (Table 1 and Figure 4). A clear-cut case of a chimeric eukaryotic system is the RNA interference machinery, in which one of the key proteins, the endonuclease Dicer, consists of two bacterial RNAse III domains and a helicase domain of apparent euryarchaeal origin, and the other essential protein, Argonaute, also shows a euryarchaeal affinity (Figure 4) 70102. The nuclear pore complex, a quintessential eukaryotic molecular machine, does not show any indications of archaeal ancestry but rather consists of proteins of apparent bacterial origin combined with proteins consisting of simple repeats whose provenance is difficult to ascertain 28.

These observations suggest that the archaeal ancestor of eukaryotes combined a variety of features found separately in diverse extant archaea. This inference is consistent with the results of phylogenomic analysis and evolutionary reconstruction discussed above. Thus, the currently existing archaeal lineages probably evolved by differential streamlining, or reductive evolution of the complex ancestral forms, whereas eukaryotes largely retained the ancestral complexity. The diverse origins of eukaryotic functional systems has major implications for how eukaryotes originated, as explained below.

The results of comparative genomics and ultrastructural studies do not yet definitively show where the eukaryotic cell came from, but they do offer important insights. Box 1 lists the key observations that must be included in any evolutionary scenario for the evolution of eukaryotes (called eukaryogenesis) and summarizes the two alternative scenarios, which are depicted in Figure 5. The main issue revolves around the role of endosymbiosis 23103104: was it the cause of the entire chain of events that led to the emergence of LECA (the stem phase of evolution), as proposed by the symbiogenesis scenario, or was it a step in the evolution of the already formed eukaryotic cell, as proposed by the archaezoan scenario? In other words, was the host of the α-proteobacterial symbiont (the future mitochondrion) a prokaryote (as in the symbiogenesis scenario) or an amitochondrial eukaryote, an archaezoan?

Given that eukaryogenesis may have been a unique event and that intermediate stages in the process cannot be seen, these questions are enormously difficult, and final answers might not be attainable. But the symbiogenesis scenario seems to be more plausible than the archaezoan scenario 105, for three main reasons. First, under the archaezoan scenario, there is no plausible selective force behind the evolution of the nucleus, and in particular the elaborate nuclear pore complex. The nucleus disrupts the transcription-translation coupling that is typical of bacteria and archaea 106107108 and necessitates the evolution of the time- and energy-consuming mechanism of nucleocytosolic transport of mRNA. By contrast, the symbiogenesis hypothesis offers a plausible selective factor: defense against the invasion of the host genome by Group II self-splicing introns, which are abundant in α-proteobacteria and could have been unleashed as a result of exposure of the archaeal host genome to the bacterial endosymbiont DNA; these would disrupt gene expression unless transcription and translation were decoupled and compartmentalized 106. At least some additional innovations of eukaryogenesis, such as the evolution of the nonsense-mediated decay of transcripts containing premature stop codons and expansion of the ubiquitin system, can be envisaged as part of the same chain of adaptations to the intron bombardment as the origin of the nucleus 109 (Figure 5).

Second, functional studies in prokaryotes, particularly archaea, show that not only the molecular components of the several signature eukaryotic systems but also their actual structures and functions have evolved in archaea and thus predate eukaryogenesis. These include the archaeal proteasome 110, exosome 111 and Sm-protein complex, the progenitor of the spliceosome 112, the ESCRT-III membrane remodeling system 113114, actin-like proteins 105 and a prototype of the ubiquitin system of protein modification 115. Each of these molecular machines found in different groups of archaea has been shown or predicted to be mechanistically similar to the eukaryotic counterpart, but they all function within the prokaryotic cell. The endomembrane system and the nucleus are dramatic exceptions, and so are the mitochondria themselves. It is tempting to connect these dots by proposing that eukaryogenesis was triggered by endosymbiosis, and that the endomembrane systems including the nucleus evolved as defense against invasion of Group II introns and perhaps foreign DNA in general 106109. It does not seem accidental that many key components of these endomembrane systems seem to be of bacterial origin whereas others are repetitive proteins that might have evolved de novo 28. Under the symbiogenesis scenario, diverse pre-existing systems of the archaeal host were co-opted and expanded within the emerging eukaryotic cellular organization 66.

Several arguments can be and have been put forward against the symbiogenesis scenario. First, prokaryotic endosymbionts in prokaryotic hosts are not widespread, prompting the view that phagocytosis, which is apparently unique to eukaryotic cells, was critical for the acquisition of the mitochondrion 3. This argument is not compelling because: (1) eukaryogenesis is extremely rare, probably unique, in the history of life; (2) endosymbiotic bacteria within other bacteria are rare but known 116117118, and intracellular bacterial predation has been suggested as a potential route to endosymbiosis 119; and (3) recent observations on membrane remodeling systems and actin-like proteins in archaea suggest the possibility of still unexplored mechanisms for engulfment of other prokaryotes, perhaps resembling primitive phagocytosis 105.

Second, a potentially strong argument against the symbiogenesis scenario could be the existence of a substantial number of eukaryote signature proteins (ESPs), so far found only in eukaryotes 120. The provenance of ESPs is an intriguing question. However, on many occasions, careful sequence and structure searches have revealed archaeal and/or bacterial homologs of proteins originally considered ESPs, or else the existence of such homologs became obvious with the appearance of new genomes 66. The discovery of prokaryotic homologs of tubulin, actin and ubiquitin are well known examples 7197, and more recent cases include the GINS proteins, which are involved in DNA replication 121, the ESCRT-III systems and the subunits of the TRAPP3 complex, which have a key role in eukaryotic vesicle trafficking 122. Under the symbiogenesis scenario, the former and remaining ESPs result primarily from acceleration of evolution of genes whose functions have substantially changed during eukaryogenesis.

A third, potentially serious difficulty with the symbiogenesis scenario is that neither archaeal-like nor bacterial-like genes can be traced to a single prokaryotic lineage (although the origin of the mitochondria from α-proteobacteria is well established). However, the pangenomes of prokaryotes are large whereas the gene composition of individual organisms is highly flexible 123124, so reconstruction of the actual partners of the endosymbiosis that led to eukaryogenesis might not be feasible from a limited set of extant genomes. Moreover, many if not most archaea and bacteria might have evolved by streamlining, so eukaryogenesis could have been triggered by symbiosis between two prokaryotes with complex genomes.

In short, it is currently impossible to strictly rule out the possibility that the key eukaryotic innovations evolved independently from and prior to the mitochondrial endosymbiosis. In other words, the host of the endosymbiont might have been an archaezoan. However, the archaezoan scenario does not provide a plausible staging of events during the evolution of the complex internal organization of the eukaryotic cell, does not offer a raison d'être for the nucleus, and does not account for the presence of signature functional systems of eukaryotes in different archaeal lineages. In contrast, the symbiogenesis scenario can tie all these diverse lines of evidence into a coherent, even if still woefully incomplete, narrative.

Comparative genomics has so far neither solved the enigma of eukaryogenesis nor offered a definitive picture of the primary radiation of the major eukaryote lineages. However, although falling short of decisive answers, phylogenomic analysis has yielded many insights into the origin and earliest stages of evolution of eukaryotes. Recent findings indicate that several key cellular systems of eukaryotes exist in archaea. The scattering of these systems among different archaeal lineages, along with the phylogenies of conserved proteins, suggests that the archaeal ancestor of eukaryotes belonged to a deep, possibly extinct archaeal branch with a highly complex genome and diverse cellular functionalities. In contrast, the endomembrane systems of eukaryotes, and in particular the nucleus with its elaborate nuclear pore complex, are not found in archaea, and seem to be derived, at least in part, from bacterial ancestral components. These findings seem to be best compatible with a symbiogenesis scenario for the origin of eukaryotes under which eukaryogenesis was triggered by the endosymbiosis of an α-proteobacterium with an ancestral archaeon, with the nucleus evolving as a defense against intron invasion.

Phylogenomic analysis has clarified the evolutionary links between major groups of eukaryotes and led to the delineation of five or six supergroups. The relationships between the supergroups and the root position in the tree of eukaryotes remain extremely difficult to decipher, probably owing to a compressed cladogenesis or 'big bang' phase of evolution that followed eukaryogenesis. The expanding sampling of genomes from diverse branches of life is far from being a trivial pursuit, but has rather delivered unexpected biological insights.