Origin of Eukaryotes: Did Energy Make Complex Cells Possible?

The origin of eukaryotes is one of the strangest single events in the history of life. Every animal, plant, fungus and amoeba descends from one ancestral complex cell, while bacteria and archaea, after roughly four billion years of trying, never produced another lineage like it. Why? One influential answer, put forward by Nick Lane and William Martin, is about power: only a cell carrying mitochondria could afford the energy that a large, gene-rich genome demands (Lane and Martin 2010). Critics have pushed back hard, and the dispute is still open.

This article sets out what is known, what each side claims, and then reworks the central arithmetic with a small sensitivity analysis to see which assumptions carry the conclusion. The question is not academic trivia. It shapes how biologists think about the odds of complex life elsewhere, and it gets asked in ordinary places too: a Tildes science thread titled “the unique merger that made you and ewe and yew” turned up in the searches done for this piece, debating exactly this “only once” puzzle.

What exactly is unresolved

Two separate questions tend to get blurred together. The first is historical: did the eukaryotic lineage arise only once? The second is causal: if so, what stopped prokaryotes from doing it again, or from becoming complex in other ways?

The historical claim has broad support. All known eukaryotes share a long list of traits (nucleus, endomembrane system, cytoskeleton, spliceosomal introns, sexual cycles) that trace back to a last eukaryotic common ancestor, usually called LECA. The causal claim is the contested part. A single origin could mean the transition was enormously hard. It could also mean that the first eukaryotes outcompeted or erased any later attempts, or that the event needed a rare coincidence of partners rather than a rare amount of energy. The energy hypothesis is a claim about the first reading, and it makes testable predictions about how genomes, cell size and metabolic rate scale.

What is established

Several points are on firm ground. Phylogenomic work places eukaryotes inside the Asgard archaea; the most extensive analysis to date puts them as a sister lineage to the Hodarchaeales within a clade the authors call Heimdallarchaeia (Eme et al. 2023). So the host cell was an archaeon, and the mitochondrion descends from an alphaproteobacterial partner. Laboratory culture has also produced an Asgard archaeon, Prometheoarchaeum syntrophicum, that grows only slowly and in partnership with other microbes (Imachi et al. 2020), a reminder that the host was probably small and metabolically entangled with neighbors.

Metabolic scaling also differs across the divide. Across body sizes, metabolic rate rises faster than linearly with mass in prokaryotes, roughly linearly in protists, and more slowly than linearly in animals (DeLong et al. 2010). That pattern is measured, not inferred, though the data sets behind each group are smaller and noisier than the clean exponents suggest.

Finally, gene expression dominates a growing cell’s energy budget. Protein synthesis takes the largest share; Ed Yong’s 2010 report gives a figure of about 75 percent for gene expression, which matches the textbook picture. Beyond these points, consensus thins out quickly.

The energy-per-gene hypothesis, stated fairly

Lane and Martin start from an observation that is hard to dispute: prokaryotes hit a ceiling of complexity. Their abstract says that eukaryotes arose once and that “otherwise prokaryotes show no tendency to evolve greater complexity” (Lane and Martin 2010). The proposed reason is that bacteria and archaea generate ATP across their outer membrane. If a prokaryote grows larger, its volume, and the genes and proteins needed to fill it, grow faster than its surface. Energy supply per gene therefore falls as size rises.

Mitochondria, on this view, broke the constraint in two ways. They moved respiration onto internal membranes, multiplying bioenergetic surface without enlarging the cell’s outer boundary. And they shed most of their genes to the host nucleus, keeping only a small local genome that controls respiration. The host could then afford a much larger nuclear genome, with each gene backed by far more power. The paper’s headline number is blunt:

“permitting a remarkable 200,000-fold expansion in the number of genes expressed”

Lane and Martin, 2010, Nature

The supporting numbers came from a table of metabolic rates, cell masses and genome sizes. As reported in a 2010 Panda’s Thumb summary of the paper, a typical bacterium in their data runs at about 0.2 W per gram and weighs about 2.6 picograms, while a typical protist runs at about 0.06 W per gram and weighs around 40,000 picograms. Giant bacteria provide a check on the logic. Epulopiscium, a gut symbiont of surgeonfish, carries a very large number of genome copies (Yong’s report cites up to 600,000), as if each patch of membrane needed its own local genes. Lane and Martin read this extreme polyploidy as evidence that big prokaryotes cannot run on a single genome.

The strongest form of the argument is not that eukaryotes burn more energy per gram. They do not. It is that energy available per gene, the budget for evolving and expressing new genes, sits far higher in a eukaryote, and that no prokaryote can reach that regime by growing larger.

The case against: continuity, gene costs and a data dispute

Michael Lynch and Georgi Marinov have mounted the most sustained challenge. In Lynch and Marinov 2015 they estimated the cost of a gene at the DNA, RNA and protein levels relative to a cell’s total lifetime ATP use. That relative cost falls as cell volume grows, in bacteria and eukaryotes alike. In bacteria, with their huge effective population sizes, the cost is large enough for selection to notice. In most eukaryotes it is swamped by genetic drift. On this account, genome expansion in eukaryotes is a population-genetic outcome of weak selection, not something mitochondria had to pay for.

Their follow-up examined membranes directly (Lynch and Marinov 2017). Their reading of the data is that bioenergetic capacity scales continuously across the prokaryote-eukaryote boundary, with no step change where mitochondria appear, and that eukaryotes are not more efficient per unit of cell volume. Their conclusion was direct:

“there is no reason to think membrane bioenergetics played a direct, causal role”

Lynch and Marinov, 2017, eLife

Martin’s group answered by auditing the data. Gerlitz et al. 2018 checked 80 cell-volume and ribosome-count values from the 2017 paper and reported that 48 of them were absent from the cited sources, differed from them, or could not be reproduced. Lynch and Marinov replied that the values came from other tables and that their reanalysis found no systematic errors; that response appears in a reply by Lynch and Marinov posted on a Stanford server, which also says the derivation of the 200,000-fold factor is unclear. The exchange settled little, except that both sides agree the underlying measurements are thin.

An independent test came later. Chiyomaru and Takemoto 2020 assembled a larger data set of genome sizes, metabolic rates, cell sizes and ploidy, then controlled for cell size and phylogeny. The energetic gap shrank. Differences in energy per gene and per genome “were less significant than those previously reported”, and the authors concluded:

“the prokaryote–eukaryote divide is hard to explain from the energetic perspective”

Chiyomaru and Takemoto, 2020, Royal Society Open Science

They did not declare the hypothesis dead; they called for more careful examination. Still, an independent reanalysis that weakens the core contrast counts as evidence against the strong version.

Other answers to “why only once”

The energy view has rivals that do not depend on watts per gene at all. Each has real support and real gaps.

• Rare partnership. The bottleneck was the improbable, stable union of an archaeal host and a bacterial endosymbiont. Energy matters less than the odds of that pairing persisting. Evidence for: known cases of bacteria living inside other prokaryotes are very rare. Against: rarity of the merger explains a single event, but not why prokaryotes never became complex by other routes.
• Population genetics. Following Lynch, complexity drifts in when selection is weak, which requires small populations, which in turn requires big cells. Evidence for: gene-cost scaling and the patchiness of genome sizes among eukaryotes. Against: it does not say what first made the cells big enough to start the process.
• Incumbency. Once eukaryotes existed, they occupied the niches any later attempt would need, much as one might argue about a single origin of life. Evidence for: plausible ecology. Against: hard to test directly, since erased competitors leave no record.
• Gradual archaeal complexity. Asgard genomes carry many eukaryote-like genes, and their lineage shows gene duplication patterns resembling eukaryotes. Complexity may have started accumulating before the mitochondrion arrived, which would weaken the claim that mitochondria came first and enabled everything else.

These are not mutually exclusive. Most researchers likely hold some mix; the disagreement is over which factor was limiting. Readers wanting background on selection versus drift will find it in this explainer on whether evolution is random.

Redoing the arithmetic: a sensitivity check built for this article

The core of the energy argument is a chain of multiplications. That makes it a good candidate for a simple question: if each input is uncertain, which ones decide the answer? A small Python script was written for this article to recompute the ratios from the published point inputs and then shake each one.

The inputs are the values as reported in secondary summaries of Lane and Martin’s table (0.2 and 0.06 W per gram; 2.6 and 40,000 picograms), plus round gene counts of 4,400 for a bacterium and 20,000 for a small eukaryote, taken from the same summaries. The full text of the 2010 paper was not accessible while this was written, so these inputs are a reconstruction, not a reproduction. The script computes three ratios: eukaryote versus bacterium power per cell; power per gene; and power per gene against a hypothetical “giant bacterium” inflated to eukaryotic mass, whose power grows with mass to an exponent b and whose genome copies grow in proportion to mass. With b = 2/3, power tracks surface area, which is the membrane-limited case at the heart of the hypothesis.

import numpy as np
rng = np.random.default_rng(19)
# Point inputs (as reported from Lane & Martin 2010, Table 1, via secondary summaries)
base = dict(qb=0.2, mb=2.6e-12, gb=4400, qe=0.06, me=4.0e-8, ge=20000, b=2/3)
def ratios(qb, mb, gb, qe, me, ge, b):
    Pb, Pe = qb*mb, qe*me                    # watts per cell
    per_gene = (Pe/ge) / (Pb/gb)             # eukaryote vs bacterium, per gene
    s = me/mb                                # mass scale-up factor
    Pbig = Pb * s**b                         # bacterium blown up to eukaryote mass
    copies = s                               # genome copies scale with volume
    giant = (Pe/ge) / (Pbig/(gb*copies))     # vs giant polyploid bacterium
    return Pe/Pb, per_gene, giant
r = ratios(**base)
print("per cell %.0f, per gene %.0f, vs giant bacterium %.0f" % r)
for b in (2/3, 1.0, 1.96):
    print("b=%.2f giant ratio %.3g" % (b, ratios(**{**base, 'b': b})[2]))
# Sensitivity: each input over a plausible range (low, high)
rng_ = dict(qb=(0.02, 2.0), qe=(0.006, 0.6), mb=(2.6e-13, 2.6e-11),
            me=(4e-9, 4e-7), gb=(1500, 10000), ge=(6000, 40000), b=(2/3, 1.0))
print("tornado on giant ratio (low, high):")
for k, (lo, hi) in rng_.items():
    vals = [ratios(**{**base, k: v})[2] for v in (lo, hi)]
    print(f"  {k}: {vals[0]:.3g} .. {vals[1]:.3g}")
# Monte Carlo, log-uniform draws; share of log-variance per input
N = 200000
draws = {k: (rng.uniform(lo, hi, N) if k == 'b' else
             np.exp(rng.uniform(np.log(lo), np.log(hi), N))) for k, (lo, hi) in rng_.items()}
out = np.log10(ratios(**draws)[2])
print("giant ratio median %.3g, 5-95%%: %.3g .. %.3g" % tuple(10**np.percentile(out, [50, 5, 95])))
for k in rng_:
    x = np.log(draws[k]) if k != 'b' else draws[k]
    print(f"  R^2 share {k}: {np.corrcoef(x, out)[0,1]**2:.2f}")
print("P(giant ratio < 100): %.3f" % np.mean(out < 2))

Ranges for the sensitivity runs are deliberately wide: a factor of ten either way for each metabolic rate and cell mass, 1,500 to 10,000 genes for bacteria, 6,000 to 40,000 for eukaryotes, and b between 2/3 and 1. The factor of ten is not arbitrary; Lynch and Marinov’s reply notes that Martin himself once described measurement uncertainty of one or two orders of magnitude. All random draws use a fixed seed (19), and the run takes a few seconds.

What the numbers do and do not show

Three things stand out. First, the size of the advantage depends heavily on what is being compared. Per cell, the reconstructed gap is about 4,600-fold. Per gene in the same two cells it drops to about 1,000-fold. Only when the bacterium is hypothetically blown up to eukaryotic size, with surface-limited power and volume-proportional genome copies, does it climb into the tens of thousands. None of these routes reproduced the abstract’s 200,000-fold figure; the exact steps behind that number could not be checked here, which is the same complaint Lynch and Marinov raised.

Second, the conclusion is far more sensitive to cell masses and metabolic rates than to gene counts. Arguments over whether a small eukaryote has 8,000 or 30,000 genes barely move the result. Arguments over typical cell size and measured respiration rates move it by orders of magnitude, and those are exactly the numbers Gerlitz and colleagues found hard to trace.

Third, the exponent b is the hinge of the giant-bacterium thought experiment. At 2/3, the advantage is large. At 1, meaning a big prokaryote could somehow scale its respiration with volume, the extra advantage from size vanishes entirely. And plugging in the superlinear exponent DeLong and colleagues measured among prokaryotes produces a nonsensical result in which the giant bacterium wins; that exponent describes small prokaryotes and cannot be extrapolated 15,000-fold in mass. The point is not that any of these is correct. It is that the debate reduces to a physiological question nobody has measured: how respiration scales in prokaryotes far larger than those we can culture.

Limits matter here. This analysis uses secondary-source inputs, independent uniform ranges and a single toy model of scaling. It says nothing about phylogeny, population size, or whether energy per gene is the right currency in the first place. It tests the internal sensitivity of one argument; it does not test the argument against nature.

The author’s reading, and what would settle it

What follows is opinion. On the evidence above, the strong version of the energy hypothesis (that mitochondria were a strict precondition for any genome expansion) looks overstated. The independent reanalysis by Chiyomaru and Takemoto weakened the contrast, the population-genetic account explains genome bloat without it, and the headline factor is fragile to inputs that are poorly measured. A weaker version survives comfortably: internal bioenergetic membranes very likely relaxed a real constraint on cell size, and larger cells then opened the door that Lynch’s drift argument describes. Energy and population genetics may be two halves of one story rather than rivals.

Several findings would move the needle:

• Direct respiration measurements in giant prokaryotes. If Epulopiscium or Thiomargarita respiration scales close to volume rather than surface, the membrane-limit premise weakens badly. A clear surface-like exponent would strengthen it.
• A sourced, open data table. Both camps agree the measurements are thin. A curated, versioned compilation of cell volume, metabolic rate, ploidy and gene number, with every value traced to a primary measurement, would let anyone rerun a script like the one above.
• Gene content of complex Asgard archaea. If newly cultured Asgard lineages show large, expanding gene repertoires and elaborate membranes without any endosymbiont, the claim that mitochondria came first loses ground.
• A complex prokaryote with internal energy membranes. Some bacteria have internal membrane systems. If any lineage with extensive internal respiratory membranes also shows unusual genome expansion, that is a natural experiment in favor of the hypothesis. If none does, that also counts.

For background on how energy-generating organelles carry their own genome, see this piece on mitochondrial DNA; for the even earlier puzzle of how cells started at all, the leading theories of abiogenesis are covered separately.

Common questions

Did complex cells really evolve only once?

The eukaryotic cell, as a lineage, appears to have a single origin, since all known eukaryotes share a common ancestor with a nucleus and mitochondria. Multicellularity is different; it arose independently many times within eukaryotes. The “once” claim is about the cell type, not about complex bodies.

Do eukaryotes use more energy than bacteria?

Per cell, yes, by a wide margin, because eukaryotic cells are much larger. Per gram of tissue, no; the reported figures suggest growing bacteria burn energy faster per unit mass. The argument is about energy available per gene, which is a different and more contested quantity.

Why would mitochondria help a cell carry more genes?

The proposal is that mitochondria put many small, locally controlled power stations inside the cell, so energy supply no longer depends on the outer membrane’s area. That frees the host to grow and to support a larger nuclear genome. Critics argue that larger genomes in eukaryotes arise mainly because selection against extra DNA is weak in small populations.

Is this question likely to be settled soon?

Partly. Better physiological data on giant bacteria and on cultured Asgard archaea could decide the membrane-limit premise within years. Whether energy or population size was the limiting factor four billion years ago may stay a matter of weighing evidence rather than a clean proof.

Where does this fit with other evolutionary puzzles?

It belongs with questions about rare, possibly singular transitions, alongside the origin of life itself and of complex organs. A related case of a single structure with a long evolutionary backstory is discussed in how the eye evolved.

Further reading

• The Vital Question: Energy, Evolution, and the Origins of Complex Life by Nick Lane. Lane's book-length case for the energy hypothesis of eukaryote origins, written for general readers. See it on Amazon (affiliate link)

As an Amazon Associate I earn from qualifying purchases. A commission may be earned if you buy through this link, at no extra cost to you.

References

1. Lane, N. and Martin, W. (2010). The energetics of genome complexity. Nature 467, 929–934. https://doi.org/10.1038/nature09486
2. Lynch, M. and Marinov, G. K. (2015). The bioenergetic costs of a gene. Proceedings of the National Academy of Sciences 112(51), 15690–15695. https://doi.org/10.1073/pnas.1514974112
3. Lynch, M. and Marinov, G. K. (2017). Membranes, energetics, and evolution across the prokaryote-eukaryote divide. eLife 6, e20437. https://doi.org/10.7554/eLife.20437
4. Gerlitz, M., Knopp, M., Kapust, N., Xavier, J. C., Martin, W. F., et al. (2018). Elusive data underlying debate at the prokaryote-eukaryote divide. Biology Direct 13, 21. https://doi.org/10.1186/s13062-018-0221-x
5. Lynch, M. and Marinov, G. K. (2018). Response to Gerlitz et al. (unpublished reply, PDF hosted at Stanford). mitra.stanford.edu
6. Chiyomaru, K. and Takemoto, K. (2020). Revisiting the hypothesis of an energetic barrier to genome complexity between eukaryotes and prokaryotes. Royal Society Open Science 7(2), 191859. https://doi.org/10.1098/rsos.191859
7. DeLong, J. P., Okie, J. G., Moses, M. E., Sibly, R. M. and Brown, J. H. (2010). Shifts in metabolic scaling, production, and efficiency across major evolutionary transitions of life. Proceedings of the National Academy of Sciences 107(29). https://doi.org/10.1073/pnas.1007783107
8. Eme, L., Tamarit, D., Caceres, E. F., et al. and Ettema, T. J. G. (2023). Inference and reconstruction of the heimdallarchaeial ancestry of eukaryotes. Nature 618, 992–999. https://doi.org/10.1038/s41586-023-06186-2
9. Imachi, H., Nobu, M. K., Nakahara, N., et al. and Takai, K. (2020). Isolation of an archaeon at the prokaryote–eukaryote interface. Nature 577, 519–525. https://doi.org/10.1038/s41586-019-1916-6
10. Yong, E. (2010). The origin of complex life: it was all about energy. National Geographic (Not Exactly Rocket Science blog). nationalgeographic.com
11. Panda’s Thumb (2010). How to afford a big sloppy genome (blog summary of Lane and Martin 2010). pandasthumb.org