How Did the Complex Eukaryotic Cell Arise?

Somewhere in a hydrothermal seep or a stratified anoxic sea, sometime between two and one and a half billion years ago, something happened that would never happen again — or at least, as far as we can tell, has never happened since. A simple cell absorbed another simple cell, or was invaded by one, or fused with one, and instead of digesting it, killing it, or dying itself, the two settled into an uneasy, then permanent, partnership. From that singular event, or cascade of events, came every redwood, every whale, every mushroom, and every human being who has ever wondered about it. It is the closest thing biology has to a creation story, and we still do not know how it was told.

This is strange, because we know so much around the edges of the event. We can sequence the genomes of the organisms implicated in it. We can watch its molecular residue — mitochondrial DNA — replicate in our own cells right now. We have found, in ocean-floor sediment, the closest living relatives of whatever cell first hosted the guest that became the mitochondrion. And yet the actual mechanics of the transition — who engulfed whom, in what order, under what conditions, and why it only worked once in four billion years of evolutionary experimentation — remain genuinely, actively contested. Not contested the way flat-earthers contest geology, but contested the way frontier science is supposed to be: by serious people with serious data reaching different conclusions from the same fossil-free, time-shrouded event.

The stakes of settling this are not merely genealogical. The eukaryotic cell — with its nucleus, its internal membranes, its power-generating organelles — is the chassis on which all complex, multicellular life is built. Understand how it arose, and you understand why complex life took three billion years to appear after life itself began, why it may be rare or common on other worlds, and what, structurally, makes something like a nervous system — and eventually a mind — possible at all.

svg[Figure 1. The proposed structure of eukaryogenesis: an archaeal host, a bacterial partner, and a merger whose exact steps remain disputed]

The convergence of these two research programs — comparative biochemistry of organelles, and phylogenomics of newly discovered archaea — is genuinely one of the great achievements of twenty-first century biology. It has replaced vague hand-waving about "a symbiosis happened" with a reasonably specific cast of characters: a host somewhere in the Asgard archaeal radiation, and a symbiont somewhere in or near the alphaproteobacteria. That much is about as settled as anything in deep evolutionary history gets.

But "reasonably specific cast of characters" is very different from "a working script," and this is where consensus dissolves into genuinely open argument.

Start with the order-of-events problem. The traditional picture — sometimes called the "archezoan hypothesis," associated with the work of Thomas Cavalier-Smith — held that a fully eukaryotic cell, complete with nucleus, cytoskeleton, and phagocytic ability, existed before it acquired mitochondria, and that some living eukaryotic lineages without mitochondria were primitively so, direct descendants of this pre-mitochondrial stage. This picture had intuitive appeal: it explains how the host could have engulfed the bacterium in the first place, since phagocytosis was already in its toolkit. But it has taken heavy damage. Every eukaryotic lineage once thought to be primitively without mitochondria — including Giardia and microsporidia — has turned out, on closer genomic inspection, to carry a stripped-down, functionally minimal remnant organelle (a mitosome or hydrogenosome), evidence that they had mitochondria and lost most of the organelle's function rather than never having it. This finding, driven substantially by the work of researchers such as Miklós Müller and later confirmed genomically, essentially killed the idea of a mitochondria-free eukaryotic ancestor as a living lineage, and shifted the burden of proof onto rival scenarios.

The rival family of scenarios — broadly, symbiosis-first models — argues that the host was not yet a complex, phagocytosing eukaryote, but something much simpler: an archaeon, perhaps not too different from Prometheoarchaeum, that acquired its bacterial partner through a mechanism other than phagocytosis, and that essentially all the hallmark complexity of the eukaryotic cell — nucleus, endomembrane system, elaborate cytoskeleton — evolved after, and arguably because of, that acquisition. The energetic argument here, most forcefully made by Nick Lane and William Martin, is that a simple prokaryotic cell is bioenergetically capped: its ATP-generating membrane is its outer cell membrane, and there's a limit to how much of that membrane a single free-living cell can maintain and fuel relative to the amount of genome it can afford to express. Get an internal, endosymbiotic power plant — an ancestral mitochondrion — and you multiply your available energy per gene enormously, freeing evolution to elaborate a vastly larger genome, more regulatory complexity, and, in time, the nucleus and the rest of the eukaryotic toolkit. On this view, mitochondria weren't a luxury acquired by an already-complex cell; they were the enabling condition for complexity to evolve at all.

The trouble is that this elegant energetic argument requires an entry mechanism for the symbiont that doesn't presuppose the very engulfment machinery it's trying to explain the emergence of. If the host wasn't yet capable of phagocytosis, how did it take in — or become permanently stuck with — a bacterial partner? Proposed answers include: the bacterium was a predator or parasite that invaded the archaeon and simply failed to leave (a "Ros" — reactive oxygen scavenging, or predatory-invasion — hypothesis); the two cells lived in such tight physical and metabolic intimacy, exchanging hydrogen and other metabolites in an anaerobic or micro-oxic environment (the "hydrogen hypothesis" of Martin and Miklós Müller), that they gradually fused, membrane by membrane, without any dramatic engulfment event at all; or the archaeal host, with its long membrane protrusions of the kind actually observed in cultured Lokiarchaea, physically wrapped around and enclosed the bacterium from the outside in, a scenario sometimes called the "entangle-engulf" or "inside-out" model, championed in different forms by researchers including David Baum and Buzz Baum.

Each of these has genuine appeal and genuine trouble. The hydrogen hypothesis elegantly explains why the partnership would be selected for — mutual metabolic benefit in a specific chemical environment — but struggles to explain the mechanical transition from "two cells exchanging metabolites" to "one cell with a permanently internalized, membrane-bound organelle whose genome has been substantially transferred to the host nucleus." The inside-out model, based on the observed morphology of cultured Asgard archaea, gives a physically plausible mechanism for engulfment without pre-existing phagocytic machinery, but doesn't by itself explain the metabolic logic of why this particular fusion, among countless possible prokaryotic contacts, became permanent and load-bearing rather than transient. And there remains a harder, more basic problem underneath all of these: distinguishing, from genomic data collected in the present, what the actual sequence of ancient events was, as opposed to what sequence would produce the genomic patterns we see today — because multiple different historical pathways can converge, after two billion years of subsequent evolution, gene loss, and horizontal gene transfer, on very similar-looking genomic signatures. This is the essential epistemic bind of the whole field: we are trying to reconstruct a unique historical event using tools — comparative genomics, phylogenetic tree-building — that are powerful at detecting deep relatedness but are much weaker at resolving fine-grained temporal order, especially when the event in question happened once, roughly two billion years ago, and left no direct fossil or geochemical trace of the transition itself, only of its outcome.

Add to this the still-unresolved question of how many symbiotic events were actually involved. Textbooks present a tidy two-step story: archaeal host acquires mitochondrial ancestor, and later (in the photosynthetic lineages) also acquires a chloroplast ancestor. But some genomic analyses have found signatures suggestive of additional bacterial contributions to the ancestral eukaryotic genome — genes that don't trace cleanly to either the archaeal host lineage or the alphaproteobacterial symbiont, sometimes dubbed a possible "third" or additional early symbiotic partner, though this remains contested and could equally reflect messy early gene transfer, contamination in ancient datasets, or systematic errors in deep phylogenetic reconstruction rather than a genuine additional merger. The uncomfortable truth is that the deeper into this history researchers dig, the less it looks like a single clean handshake between two cells, and the more it looks like a turbulent, gene-swapping melee at the dawn of complex life, the fine details of which may simply be unrecoverable.

And then there is the uniqueness puzzle, which is arguably the strangest part of the whole picture. If mitochondrial acquisition were even a moderately probable outcome of prokaryotic ecology — the kind of thing evolution would stumble into now and again — we would expect to see it, or something like it, happening more than once across the tree of life, the way chloroplast acquisition has happened independently several times. But despite roughly three and a half billion years of prokaryotic history, uncountable trillions of predator-prey and parasite-host and syntrophic-partner encounters between archaea and bacteria, and an ongoing, richly documented, present-day zoo of bacterial endosymbionts living inside other cells (from Wolbachia in insects to sulfur bacteria in tubeworms), nothing else has ever produced a second eukaryote-like fusion, with permanent organellar integration, massive gene transfer to a host nucleus, and a cascade into full cellular complexity. This could mean the event required an extraordinarily specific, perhaps never-repeated combination of host and symbiont biology and environmental circumstance. It could mean that it has an intrinsically low probability per encounter, and we are simply looking at one lucky roll of a very long-odds die, in which case in some sense we got extremely fortunate to exist as complex creatures at all. Or it could mean we're not looking hard enough, or in the right places, or that early competitive exclusion — the first successful eukaryotic lineage rapidly out-competing and eliminating any nascent rivals before they could diversify — erased the evidence of parallel attempts that did once exist. We simply do not know which of these is true, and that uncertainty is not a minor footnote; it bears directly on how surprising, or how expected, complex life should be anywhere in the universe.

It's worth laying the main research programs side by side, because none of them is a strawman, and each has real evidentiary support along with real explanatory gaps.

The symbiosis-first / energetic hypothesis (Martin, Lane, and collaborators) holds that a metabolically intimate, probably syntrophic, relationship between an anaerobic archaeal host and a bacterial partner — possibly initiated through hydrogen or other metabolite exchange in an anoxic or micro-oxic niche — was the founding event, and that the huge bioenergetic surplus provided by an internalized, membrane-respiring symbiont then permitted the subsequent evolution of genomic complexity, the nucleus, and the endomembrane system as consequences, not causes, of the merger. Its strength is a clear, quantifiable, mechanistic story for why eukaryotic complexity would need this particular kind of energetic unlock, and it fits the observation that no mitochondria-free "pre-eukaryote" lineage has ever been found. Its weakness is the entry-mechanism gap: absent phagocytosis, the actual moment of internalization remains hand-wavy, resting on syntrophic intimacy gradually tightening into fusion, a transition that is plausible in outline but has not been observed, reconstructed experimentally, or pinned to specific molecular steps.

The phagocytosis-first / archezoan-descendant hypothesis (associated with Cavalier-Smith's tradition, though modified heavily since) holds that a genuinely complex, actively engulfing eukaryotic ancestor already existed and captured its mitochondrial ancestor as prey that escaped digestion. Its strength is mechanistic plausibility — we know phagocytosis happens today, we can watch amoebas do exactly this to bacteria in real time — and it doesn't require inventing an exotic, unobserved fusion mechanism. Its weakness is that the strongest empirical prop for it, mitochondria-free "primitive" eukaryotic lineages, has been progressively dismantled by genomic evidence of relic mitosomes and hydrogenosomes in every case examined, undermining the idea that any living lineage represents a pre-mitochondrial eukaryotic grade.

The inside-out / entangle-engulf hypothesis (associated with David Baum, Buzz Baum, and given fresh empirical grounding by the actual morphology of cultured Asgard archaea like Prometheoarchaeum) proposes that the archaeal host had long, branching membrane protrusions — exactly the structures Imachi's team observed reaching out from the cultured cell — and that these protrusions gradually wrapped around and enclosed an ectosymbiotic bacterial partner from the outside, turning what began as an external, surface-bound relationship into full internal enclosure without requiring pre-existing phagocytic machinery at all. Its strength is that it's grounded in an actual observed cell morphology, not an inferred or hypothetical one, and it neatly sidesteps the phagocytosis paradox. Its weakness is that we've cultured only a handful of Asgard archaea so far, under artificial laboratory conditions very different from a two-billion-year-old seafloor, and it remains unclear whether the protrusions observed are a general Asgard trait relevant to eukaryogenesis or an idiosyncrasy of the particular strain and culture conditions used.

The syntrophic consortium / multiple-partner hypothesis takes seriously the messier genomic signals suggesting more than a simple two-organism merger, proposing that eukaryogenesis emerged from an extended microbial consortium — an archaeal host, an alphaproteobacterial partner, and possibly other bacteria contributing genes and metabolic functions along the way, in a shared anaerobic or low-oxygen niche such as a mat or biofilm — rather than a clean bilateral fusion. Its strength is that it may better fit the genomic mosaicism actually observed in eukaryotic genomes, where genes trace to a surprisingly wide range of bacterial sources. Its weakness is that "multiple messy contributions over an extended period" is a much harder hypothesis to test, falsify, or render mechanistically precise than a clean two-partner merger, and it risks becoming an explanation flexible enough to accommodate almost any data.

Given that the event itself is unrecoverable — no fossil eukaryogenesis, no ancient DNA from the actual transition — what would count as resolution? Several distinct kinds of evidence, converging, would move this from contested to (as much as ancient history ever gets) settled.

The first is culturing more, and more diverse, Asgard archaea, and studying their actual cell biology in detail: do they have functional, dynamic actin-based cytoskeletons? Can their ESCRT-like proteins actually remodel membranes the way eukaryotic ESCRT does? Do their observed membrane protrusions do anything resembling engulfment under any culturable condition, or interact physically with bacterial partners in ways that model the proposed inside-out mechanism? This is slow, unglamorous work — Imachi's team's cultured Lokiarchaeon took over a decade from initial sediment sampling to publication, because these organisms grow with generation times of two weeks or more, in strict anaerobic co-culture with partner organisms they cannot easily be separated from. But every additional cultured Asgard lineage is a chance to test specific mechanistic predictions against a living cell rather than a genome sequence alone, and the pace of discovery of new Asgard lineages, and successful cultivation attempts, has been accelerating.

The second is better resolution of ancient phylogenetic trees themselves — a technical, statistical frontier as much as a biological one. Reconstructing relationships across a two-billion-year-old divergence, using genes that have been edited, transferred, and reshuffled ever since, pushes phylogenetic methods to their limits, and different modeling assumptions about rates of evolution and compositional biases in ancient sequences can shift which archaeal lineage looks like the eukaryotes' closest relative, and how deep in the archaeal tree the eukaryotic root actually sits. Improved statistical models — already an active area of method development among researchers working on deep phylogenetics — could sharpen or overturn specific claims about which Asgard lineage is closest to us, which would in turn constrain scenarios about the ancestral host's biology.

The third is direct experimental synthetic biology: attempts to reconstruct minimal syntrophic or endosymbiotic relationships in the lab between simple prokaryotes, and watch whether, and under what conditions, loose metabolic partnerships can be pushed toward tighter physical and genetic integration. This is far from recreating eukaryogenesis itself, but controlled laboratory evolution experiments — tracking, say, whether an engineered bacterial endosymbiont within a host cell begins transferring genes to the host genome under selective pressure over hundreds or thousands of generations — could at least test whether the proposed mechanisms (gene transfer, membrane fusion, metabolic integration) are plausible steps of the kind claimed, rather than untestable just-so stories. Some steps in this direction already exist: there are laboratory lineages in which bacterial endosymbionts have been experimentally introduced into eukaryotic host cells and maintained across many generations, offering partial, imperfect analogues.

The fourth, and perhaps most tantalizing, is the search for independent natural experiments — not a second full eukaryogenesis, which seems not to have happened, but simpler modern symbioses that are partway along a similar road, which could illuminate which steps are hard and which are easy. Certain insect-bacteria endosymbioses, and some remarkable cases of one bacterium living permanently inside another bacterium (a phenomenon documented in some mealybug-associated bacterial nested symbioses), show genuinely deep genomic integration, extreme genome reduction in the symbiont, and dependency comparable in some structural respects to organelle evolution, even though these are bacterium-in-bacterium or bacterium-in-eukaryote rather than the archaeon-bacterium combination relevant to mitochondrial origin. Studying exactly how far these partnerships have progressed toward organelle-like integration, and why they haven't (yet, or ever) gone all the way, could sharpen our sense of what a completed transition actually requires — echoing questions that also arise in the search for a shadow biosphere or a second genesis, where the underlying puzzle is likewise: how do you recognize a genuinely novel biological transition, or a genuinely independent origin, using only the residue it left behind?

None of these approaches will hand us a video of the actual event. But converging evidence from cultured Asgard biology, improved deep phylogenetics, synthetic reconstruction experiments, and comparative study of ongoing symbioses could, within the coming few decades, narrow the live hypotheses from four or five down to one or two with real confidence — which, for an event this old and this singular, is about as much resolution as we can reasonably hope for.

It would be easy to file this under intellectual curiosity — an extremely deep, extremely old whodunit with no practical stakes. That would be a mistake.

First, this transition is the hinge on which the entire subsequent history of complex life turns. Every multicellular organism — every plant, animal, and fungus — is eukaryotic; prokaryotes, for all their staggering metabolic ingenuity and near-total planetary dominance by biomass and diversity, have never evolved anything resembling a brain, a limb, or a leaf. Understanding why the eukaryotic cell, and only the eukaryotic cell, became the platform for large-scale multicellular complexity bears directly on debates connected to the puzzle of how genotypes map to complex phenotypes: the eukaryotic cell's internal compartmentalization, its capacity for elaborate gene regulation, and its energetic surplus from mitochondrial respiration are widely thought to be preconditions for the kind of regulatory complexity multicellular development requires. If that dependency is real and specific — rather than one of several equally viable paths evolution could have taken — it reshapes how we think about the space of possible biologies, on this planet and elsewhere.

Second, mitochondria are not just historical curiosities; they are active, central players in modern human physiology, and our understanding of their evolutionary origin feeds directly into research on why they fail with age. Mitochondrial dysfunction is one of the recognized hallmarks connected to the biology of aging, implicated in neurodegeneration, metabolic disease, and the general decline of cellular energy production over a human lifespan. Some of this dysfunction traces to exactly the evolutionary legacy of the endosymbiotic event: the mitochondrion's stripped-down genome, its residual vulnerability to oxidative damage from the very respiratory chemistry that made the symbiosis valuable in the first place, and the intricate, still imperfectly understood cross-talk required between mitochondrial and nuclear genomes to keep two once-independent genomes working in permanent, precise coordination. Understanding the deep history of that coordination — how it was forged, what compromises were baked in at the origin — informs present research into why it degrades.

Third, and perhaps most profoundly, this question sits at the center of how we should think about the probability of complex life elsewhere in the universe. If eukaryogenesis required an extraordinarily improbable, near-singular convergence of a specific archaeal lineage, a specific bacterial partner, and a specific environmental context — the kind of event that might reasonably be expected to occur once per observable universe rather than once per habitable planet — then simple microbial life could be common across the cosmos while complex, macroscopic, differentiated life remains vanishingly rare. If, on the other hand, the transition reflects a more general and repeatable class of syntrophic-to-symbiotic escalation that simply happens to have gone to completion only once on this particular planet by chance, then complex life's rarity here may say less about its intrinsic difficulty and more about contingent bad luck, or about early competitive exclusion erasing rival attempts before they could flourish. We cannot yet distinguish these possibilities, and the distinction matters enormously for how we should calibrate our expectations about life, and mind, elsewhere.

The intellectual lineage of this question runs from outsiders to institutions and back to outsiders again.

The idea itself is older than most people realize. The Russian botanist Konstantin Mereschkowski proposed in 1905 that chloroplasts were the descendants of symbiotic cyanobacteria, and the American Ivan Wallin extended similar reasoning to mitochondria in the 1920s — both were largely dismissed or ignored for decades, their ideas considered fanciful by a biological mainstream more comfortable with gradualist, non-symbiotic accounts of organelle origin.

history

• 1905/1920s — Mereschkowski and Wallin independently propose symbiotic origins for chloroplasts and mitochondria; largely dismissed.
• 1967 — Lynn Margulis (as Lynn Sagan) publishes "On the Origin of Mitosing Cells," reviving endosymbiotic theory with a systematic case.
• 2015 — Spang, Ettema, and colleagues report Lokiarchaeota from Arctic seafloor sediment, the first strong genomic link to a eukaryote-like archaeal host.
• 2019/2020 — Imachi, Nishihara, and colleagues culture Candidatus Prometheoarchaeum syntrophicum, directly observing an Asgard archaeon's cell biology for the first time.

Stand back from the argument over mechanisms, and what remains is something almost vertiginous: every multicellular creature that has ever lived, every fern and every philosopher, is the downstream consequence of a single ancient partnership between two microbes whose precise handshake we still cannot fully describe. We know the participants — an archaeon from somewhere near the Asgard radiation, a bacterium from somewhere near the alphaproteobacteria. We know, roughly, the geological and chemical stage on which it happened — an anoxic or low-oxygen Proterozoic world. We do not know the choreography.

This is not a failure of biology; it is a mark of how genuinely difficult the question is, and an honest field's tolerance for saying so. Deep evolutionary history, unlike much of experimental biology, cannot always be re-run, isolated, or directly observed — it must be inferred, with appropriate humility, from the traces a single unrepeated event left scattered across billions of years of subsequent evolution. That inference has gotten dramatically better in the past decade. It is not yet complete, and it may never fully be.

What remains genuinely open:

• Did the symbiosis or the cellular complexity come first — was the host a simple archaeon transformed by acquiring an energetic partner, or an already-sophisticated proto-eukaryote that captured prey it could not digest?
• What was the actual physical mechanism of entry — engulfment, invasion, gradual syntrophic fusion, or an inside-out enclosure by archaeal membrane protrusions — and can any of these be reconstructed step by step from cultured organisms or synthetic experiments?
• Was there truly a single bacterial partner, or a more tangled consortium of contributing microbes whose genomic fingerprints have been blurred by two billion years of subsequent evolution?
• Why has this transition, so evolutionarily consequential, apparently happened only once — bad luck, an intrinsically rare mechanism, or early competitive exclusion erasing rival attempts we will never see?
• What does the answer imply for how common or rare complex, differentiated, potentially conscious life should be elsewhere in the universe?

Ask the tutor: If eukaryogenesis really did happen only once in four billion years of evolutionary experimentation on Earth, what would that imply about the odds of complex life arising anywhere else — and how would we even begin to test such a claim?