The Origin of Eukaryotes and Endosymbiosis

Endosymbiosis theory (formalized as the "serial endosymbiotic theory," SET) proposes that key eukaryotic organelles originated as free-living prokaryotes that were engulfed by another cell and, rather than being digested or destroyed, survived as internal symbionts. Over evolutionary time these symbionts became permanently integrated, losing their autonomy and most of their genomes while retaining a specialized metabolic role: the mitochondrion for oxidative ATP production, the plastid for photosynthesis. This is fundamentally different from the way most organelles evolved (e.g., the endoplasmic reticulum, Golgi, or nucleus, which appear to have arisen through internal membrane elaboration of the host's own compartments) — mitochondria and plastids are the products of symbiogenesis, a merger of two separate genomes into one lineage. Two nested acquisitions must be kept distinct. First, at the base of the eukaryotic tree, an archaeal host cell acquired an alphaproteobacterial endosymbiont that became the mitochondrion; this event (or a closely spaced series of events) is now thought to coincide with, or even to have driven, the origin of the eukaryotic cell itself — essentially all known eukaryotes either have mitochondria or descend from ancestors that did (a few anaerobic lineages have reduced, mitochondrion-derived organelles like hydrogenosomes and mitosomes, rather than lacking them primitively). Second, much later, one eukaryotic lineage independently engulfed a cyanobacterium that became the plastid, giving rise to the Archaeplastida (red algae, green algae, land plants, and glaucophytes). Plastids were then passed sideways to other eukaryotic lineages through secondary and tertiary endosymbiosis — a eukaryotic alga being engulfed by another eukaryotic host — explaining the bewildering diversity of algal groups with plastids surrounded by three or four membranes instead of two. The key terms to keep straight: an endosymbiont is an organism living inside another; an organelle derived from endosymbiosis retains vestiges of that ancestry (its own genome, double membrane, prokaryote-like ribosomes); endosymbiotic gene transfer (EGT) is the process by which genes move from the symbiont's genome to the host nucleus over time, so that the organelle genome shrinks while the host genome takes over control of most organellar proteins, which are then re-imported into the organelle after translation in the cytoplasm.

Step 1 — The host cell. Molecular phylogenetics of ribosomal RNA and, more recently, of hundreds of concatenated proteins places the eukaryotic host's ancestry within the archaea, specifically nested inside or sister to a recently discovered archaeal superphylum called the Asgard archaea (named for members like Lokiarchaeota, Thorarchaeota, Odinarchaeota, and Heimdallarchaeota, first identified from marine sediment metagenomes near hydrothermal vents). Asgard archaea encode a surprising number of "eukaryotic signature proteins" — actin homologs, ESCRT membrane-remodeling machinery, small GTPases — proteins once thought unique to eukaryotes, suggesting the host already had a cytoskeleton-like system and membrane-trafficking capacity before acquiring mitochondria. This matters mechanistically: a host with a flexible, actin-supported membrane could plausibly engulf another cell by phagocytosis-like means, or alternatively could have formed close physical/metabolic associations that later resulted in engulfment.

Step 2 — The metabolic handshake. The leading current model (a version of the "hydrogen hypothesis" proposed by William Martin and Miklós Müller, refined since) suggests the host was an anaerobic, hydrogen-dependent archaeon, and the symbiont was a facultatively anaerobic alphaproteobacterium capable of producing hydrogen and CO₂ as fermentation byproducts while also being capable of aerobic respiration when oxygen was present. A syntrophic (mutually feeding) relationship — the archaeon consuming H₂ produced by the bacterium — could have drawn the two cells into permanent physical association, eventually leading to internalization of the bacterium. This explains why mitochondria are so nearly universal in eukaryotes: the merger and the origin of the "eukaryotic" cellular architecture may be one and the same event, not a preexisting proto-eukaryote later acquiring an organelle as an accessory.

Step 3 — Internalization and genome reduction. Once inside, the endosymbiont's genome experienced relaxed selection on functions now redundant with the host (cell wall synthesis, independent DNA repair, biosynthetic pathways duplicated in the host) and was subject to Muller's-ratchet-like genetic drift because endosymbiont population sizes within a host are small. Genes were lost outright, or transferred to the host nuclear genome (endosymbiotic gene transfer, EGT). A modern human mitochondrial genome retains only:

The vast majority of genes needed to build and run a mitochondrion (over 1,000 in humans) now reside in the nucleus; their protein products are synthesized on cytoplasmic ribosomes and imported back into the organelle through dedicated TOM/TIM translocase complexes (Translocase of the Outer/Inner Mitochondrial membrane) using N-terminal targeting sequences. Plastids show a parallel, even more extreme reduction: the plastid genome of land plants typically encodes only about 100–120 genes, down from an ancestral cyanobacterial genome of several thousand, with import managed by analogous TOC/TIC translocases.

Step 4 — Functional integration and specialization. With most of its genes relocated, the organelle became permanently dependent on the host, and the host became permanently dependent on the organelle's unique biochemistry (oxidative phosphorylation; the Calvin cycle). The organelle divides by binary fission in synchrony with the host cell cycle, its numbers partitioned to daughter cells, and cristae or thylakoid membranes elaborate the original bacterial plasma membrane to maximize surface area for chemiosmotic ATP synthesis.

Step 5 — Secondary and tertiary endosymbiosis (plastids only). After the primary cyanobacterial capture produced red and green algae, several unrelated eukaryotic lineages independently engulfed a whole eukaryotic alga (not a free bacterium) and retained its plastid, producing plastids bounded by three or four membranes (the original cyanobacterial double membrane, plus one or two more from the phagocytic and algal plasma membranes). This explains plastids in euglenids (secondary green-algal origin), diatoms and brown algae (secondary red-algal origin, via the "chromalveolate" hypothesis, now understood as more complex with likely multiple independent or serial events), and dinoflagellates (which in some lineages have even undergone tertiary replacement of their plastid by yet another algal endosymbiont).

Endosymbiotic theory did not begin as a fringe idea and end as fact overnight — it took roughly a century of accumulating structural, biochemical, and finally genomic evidence.

Early proposals (1880s–1920s). Botanist Andreas Schimper noted in 1883 that chloroplasts divide independently of the cell nucleus and resemble free-living cyanobacteria (then called "blue-green algae"); he speculated in a footnote that they might have originated as symbiotic algae. Russian biologist Konstantin Mereschkowsky formalized this into a full symbiogenesis theory in 1905, proposing that plastids were domesticated cyanobacteria. Ivan Wallin extended the idea to mitochondria in the 1920s, even claiming (unsuccessfully, and prematurely) to have cultured mitochondria outside cells as free bacteria. These proposals were largely dismissed for decades — mainstream cell biology favored the view that organelles arose by gradual internal specialization of host membranes, and there was no way, before molecular biology, to test the two hypotheses rigorously.

Structural evidence (1950s–60s). Electron microscopy revealed that mitochondria and chloroplasts have a double membrane (unlike other eukaryotic organelles, which are single-membrane-bound, having budded from the endomembrane system) and that they divide by a bacterial-like fission process independent of the host cell's mitotic machinery, rather than being built de novo each generation.

The discovery of organellar DNA (1963). Margit and Sylvan Nass, using electron microscopy of osmium-fixed chick embryo cells treated with DNase, identified fibrous DNA-containing structures inside mitochondria — the first direct demonstration that mitochondria carry their own genetic material, separate from the nuclear genome. Around the same time, chloroplast DNA was independently identified by Hans Ris and Walter Plaut in Chlamydomonas. This was the single most important structural clue: an organelle with its own genome is not something the nucleus builds from scratch each generation, it is something that replicates its own hereditary material.

Lynn Margulis's synthesis (1967). Lynn Sagan (later Lynn Margulis), then a young researcher, submitted "On the Origin of Mitosing Cells" — rejected by roughly fifteen journals before being published in the Journal of Theoretical Biology. She synthesized structural, biochemical, and paleontological evidence into a coherent, falsifiable Serial Endosymbiotic Theory, proposing mitochondria arose from aerobic bacteria, plastids from cyanobacteria, and (more controversially, and now largely unsupported) that eukaryotic flagella arose from symbiotic spirochetes. The paper was initially met with considerable skepticism, partly because it directly contradicted the dominant "autogenous" model of organelle origin.

The decisive molecular evidence: ribosomal RNA phylogenies (1970s–80s). The test that converted endosymbiosis from a plausible story into an established fact came from comparing nucleotide sequences. Carl Woese, developing the tools of molecular phylogenetics using small-subunit ribosomal RNA (SSU rRNA) sequences, showed in the 1970s that mitochondrial and chloroplast rRNA genes are far more similar to bacterial rRNA sequences than to the nuclear/cytoplasmic rRNA of their own host cell. Specifically:

• Mitochondrial rRNA sequences cluster within the alphaproteobacteria, close to modern genera like Rickettsia (an intracellular parasite) — this is a phylogenetic tree, not a resemblance by eye, and it is the single strongest piece of evidence for endosymbiosis.
• Chloroplast rRNA sequences cluster within the cyanobacteria.
• Nuclear/cytoplasmic ribosomal RNA of the same eukaryotic cell falls into an entirely separate branch of the tree of life (the eukaryotic domain), confirming that the organelle genome and the nuclear genome have genuinely independent evolutionary histories fused within one cell.

This was, in effect, a controlled comparison: if organelles had arisen by internal specialization of host membranes, their DNA should show no special affinity to any particular free-living bacterial lineage — it should simply be host DNA. Instead, organelle DNA phylogenetically nests inside specific bacterial groups, exactly as predicted if they are the descendants of once free-living bacteria that have been diverging for roughly 1.5–2 billion years, but still show clear common ancestry with their bacterial sisters.

Biochemical confirmation. Organelles were shown to have 70S ribosomes (the smaller bacterial/archaeal type, sensitive to antibiotics like streptomycin and chloramphenicol that block bacterial translation but not the 80S cytoplasmic ribosomes of the eukaryotic host); a bacterial-type circular (in most lineages), naked, histone-free chromosome; a bacterial-type membrane lipid, cardiolipin, concentrated in the mitochondrial inner membrane and otherwise found in bacterial plasma membranes; and translation initiated with N-formylmethionine, as in bacteria, rather than plain methionine as in the eukaryotic cytoplasm.

Genomic confirmation of the Asgard archaeal host (2015–2020). Metagenomic sequencing of marine sediments near the Loki's Castle hydrothermal vent field (Norwegian-Greenland Sea) by Thijs Ettema's group recovered a nearly complete genome they named Lokiarchaeum, showing it encoded actin homologs, ESCRT proteins, and other "eukaryotic signature genes" previously thought to be eukaryote-specific innovations. Subsequent culturing efforts (Imachi and colleagues, 2020) succeeded in growing a related Asgard archaeon (Candidatus Prometheoarchaeum syntrophicum) in the laboratory under strict anaerobic, syntrophic co-culture conditions — the organism grew extremely slowly (doubling time of roughly 14–25 days) and developed long branching membrane protrusions, offering a physical model for how such a cell might engulf a bacterial partner. This provided experimental, not just phylogenomic, support for the archaeal-host, syntrophic model of eukaryogenesis. Together, these independent lines of evidence — electron microscopy, organellar DNA discovery, ribosomal RNA phylogeny, biochemical ribosome/lipid comparisons, and Asgard archaea genomics — converge on a single conclusion robust to any one method's limitations: eukaryotic organelles are domesticated bacteria, and the host lineage itself is archaeal.

Example 1 — Reading a rRNA gene tree. Suppose you are given four SSU rRNA sequences to compare: (1) nuclear rRNA from a diatom, (2) mitochondrial rRNA from the same diatom, (3) rRNA from Rickettsia prowazekii (an alphaproteobacterium), and (4) rRNA from Synechocystis (a cyanobacterium). A phylogenetic tree built from sequence alignment groups sequence (2) tightly with sequence (3), both nested well inside the alphaproteobacterial clade, with a bootstrap support of, say, 97%. Sequence (1) instead falls in a completely separate eukaryotic clade with other diatom and algal nuclear sequences, closer to other eukaryotic host lineages than to any bacterium. How do you interpret this? The mitochondrial genome's phylogenetic position, deeply nested within one specific bacterial group rather than branching off before the origin of that group, means the mitochondrion did not evolve its rRNA machinery independently — it inherited it, with modification, from a genuine alphaproteobacterial ancestor. The fact that the host's nuclear rRNA belongs to an entirely different domain-level lineage confirms that two independently evolving genomes are physically co-located in one cell — the signature of endosymbiosis, not internal membrane elaboration. If you additionally sequenced the diatom's plastid rRNA and found it nested not with green algae or cyanobacteria directly but inside a red algal clade, that pattern (a plastid whose closest phylogenetic relative is another eukaryote's plastid, rather than a free-living cyanobacterium) is the signature of secondary endosymbiosis — the diatom's ancestor engulfed a red alga, not a cyanobacterium directly.

Example 2 — Estimating gene relocation from genome size. Free-living alphaproteobacteria today (e.g., Rhodospirillum rubrum or close relatives used as proxies for the mitochondrial ancestor) typically have genomes encoding roughly 3,000–4,000 protein-coding genes. Modern human mitochondrial DNA encodes only 13 proteins. Assume, as a simplifying estimate, that the ancestral endosymbiont needed approximately 1,200 genes to sustain its own core respiratory and replicative functions once housed inside a host cell (a smaller number than a free-living relative, since some functions — cell wall synthesis, motility, many biosynthetic pathways — became redundant or were lost outright rather than relocated). Of those ~1,200 genes, only 13 protein-coding genes remain in the organelle genome. That means:

Roughly 99% of the ancestral gene set relevant to mitochondrial function has either been lost or transferred to the nucleus over ~1.5–2 billion years. This is consistent with observed rates: comparative genomics across eukaryotes shows mitochondrial genomes ranging from as few as 3 protein-coding genes (in some parasitic protists) up to about 97 genes (in the protist Reclinomonas americana, thought to retain a relatively ancestral mitochondrial gene set) — a real empirical range that brackets the > 1,000-gene nuclear-encoded mitochondrial proteome of humans, most of which arrived in the nucleus via endosymbiotic gene transfer, was retargeted with a mitochondrial import sequence, and is now translated in the cytosol and re-imported through TOM/TIM channels.

Lynn Margulis (born Lynn Petra Alexander in Chicago, 1938) became the central figure in establishing endosymbiotic theory as serious science, though she built on ideas from Schimper, Mereschkowsky, and Wallin. Trained at the University of Chicago and later earning her PhD at Berkeley, Margulis worked largely outside the mainstream of her era's cell biology, which favored gradual, autogenous models of organelle evolution. Her 1967 paper, submitted under the name Lynn Sagan, synthesized cytological, biochemical, and comparative evidence into a rigorous, falsifiable framework — and was rejected by roughly fifteen journals before acceptance. She continued developing and defending the theory through her 1970 book Origin of Eukaryotic Cells, at a time when molecular sequencing tools did not yet exist to settle the question decisively; that vindication came a decade later through Carl Woese's ribosomal RNA phylogenies. Margulis remained a combative and unconventional figure throughout her career, championing symbiosis as a major creative force in evolution (sometimes controversially overextending the idea, as with her insistence — not supported by subsequent evidence — that eukaryotic flagella derived from symbiotic spirochetes) while also becoming an influential co-developer, with James Lovelock, of Gaia theory. She held a position at the University of Massachusetts Amherst for much of her later career and was elected to the National Academy of Sciences in 1983.

Medicine. Mitochondrial dysfunction underlies a large class of human diseases — from Leber's hereditary optic neuropathy to MELAS syndrome (mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes) — caused by mutations in mtDNA or in nuclear genes governing mitochondrial import and function. Because mtDNA is maternally inherited (sperm mitochondria are typically destroyed after fertilization), mitochondrial disease transmission follows distinctive, non-Mendelian pedigrees, and "mitochondrial replacement therapy" (so-called "three-parent" IVF techniques) directly exploits the fact that mitochondria are a semi-autonomous, separately inherited genome to prevent transmission of pathogenic mtDNA mutations. Understanding mitochondrial genetics as the legacy of a bacterial ancestor also explains why some antibiotics (targeting bacterial 70S ribosomes) can have off-target toxicity on human mitochondrial translation.

Forensics and population genetics. Because mtDNA is maternally inherited, non-recombining (largely), and present in high copy number per cell, it is widely used to trace maternal lineages, in forensic identification, and in reconstructing human population history and migrations (e.g., mitochondrial "Eve" studies of deep human ancestry).

Agriculture and biotechnology. Chloroplast genetic engineering ("chloroplast transformation") exploits the plastid's bacterial-like genome architecture to introduce transgenes with high expression levels and reduced risk of pollen-mediated gene flow (since plastids are typically maternally inherited in most crops), a strategy used in efforts to engineer disease resistance and improved photosynthetic efficiency in crop plants.

Evolutionary and origin-of-life research. The Asgard archaea discoveries reframed active research on the tree of life, prompting revision of eukaryotic classification (some now argue eukaryotes should be classified as a lineage within archaea rather than a separate domain). Ongoing culturing efforts with organisms like Candidatus Prometheoarchaeum are used to test physical models of how membrane protrusions could engulf a bacterial partner, informing broader theories of how Major Evolutionary Transitions in Evolution like symbiogenesis occur mechanistically, not just genetically.

Conservation and ecology. Endosymbiotic and symbiotic relationships remain central to modern ecosystems — reef-building corals depend on photosynthetic dinoflagellate symbionts (Symbiodinium), and coral bleaching under thermal stress is fundamentally a breakdown of a eukaryote–eukaryote symbiosis mechanistically related to the ancient plastid-acquisition events discussed here, underscoring that symbiogenesis is not a one-time historical curiosity but an ongoing evolutionary process.

Ask the tutor: How exactly does the hydrogen hypothesis explain why a hydrogen-dependent archaeon and a hydrogen-producing bacterium would end up merging into one cell rather than just living side by side?