The theory you're describing is endosymbiotic theory, and its central claim is that mitochondria and chloroplasts descend from free-living prokaryotes that were engulfed by a larger host cell roughly 1.5 to 2 billion years ago and never left. The key boundary: it explains only two organelles — mitochondria and plastids — not the nucleus, ER, or Golgi, which arose by other routes. Lynn Margulis formalized the modern version in a 1967 paper, and the evidence — double membranes, their own circular DNA, bacterium-sized ribosomes — has only strengthened since.
What Endosymbiotic Theory Actually Claims, Cell by Cell
Endosymbiotic theory holds that a host cell — likely an archaeon or an early eukaryote — took in a bacterial cell by phagocytosis and instead of digesting it, kept it alive and put it to work. The guest eventually lost the ability to live independently and became an organelle. That's the whole idea in one sentence.
The theory makes specific, testable claims, not a vague gesture. First, mitochondria descend from an alphaproteobacterium. Second, chloroplasts descend from a cyanobacterium. Third, the transfer happened once for mitochondria and a small number of times for plastids. Fourth, the host lineage was already capable of phagocytosis or something like it.
Two organelles are covered; everything else is not. The nucleus, endoplasmic reticulum, Golgi apparatus, and peroxisomes have separate origins under different hypotheses. Students often assume "eukaryotes came from symbiosis, full stop" — the theory is narrower than that.
Lynn Margulis published "On the Origin of Mitosing Cells" in the Journal of Theoretical Biology in 1967. The paper was rejected by roughly a dozen journals before acceptance, a detail Margulis repeated in interviews for decades. Her earlier 1966 book Origin of Eukaryotic Cells laid out the case at length.
The timeline matters. Molecular clock estimates place the mitochondrial acquisition at roughly 1.5 to 2 billion years ago, and the chloroplast acquisition later, around 1 to 1.5 billion years ago, after cyanobacteria had diversified. Those windows come from phylogenomic analyses rather than fossils, because the event left no direct fossil trace.
Lynn Margulis, the 1967 Paper, and Why the Idea Was Rejected for Years
Margulis didn't invent the idea from nothing — the Russian botanist Konstantin Mereschkowski proposed a symbiosis-based origin for chloroplasts in 1905, and Ivan Wallin argued for bacterial mitochondria in the 1920s. Both were largely ignored. Margulis's contribution was a full mechanistic case backed by then-available cell biology.
The rejection pattern had a specific cause. Mid-century cell biology treated organelles as products of the cell's own membranes, and the idea that a bacterium could become a permanent part of another cell conflicted with that framework. Margulis also bundled the argument with broader claims about symbiosis as a major evolutionary force, which made the core idea easier to dismiss.
What turned the field was DNA. Starting in the 1960s and accelerating through the 1970s, researchers found that mitochondria and chloroplasts contain their own genomes, circular like bacterial chromosomes, and that these genomes are maternally inherited. Sequencing in the 1980s and 1990s then placed mitochondrial genes squarely inside the alphaproteobacteria.
The 1967 paper is the citation most textbooks use, but accept the nuance: the modern consensus version differs from Margulis's in one important way. She leaned toward an autogenous-plus-symbiotic hybrid for some features; the current view is that the host was an archaeon and the merger with an alphaproteobacterium created the eukaryotic lineage. Her broader symbiogenesis claims never achieved the same acceptance.
Five Lines of Evidence: Membranes, DNA, Ribosomes, Size, and Gene Transfer
The evidence for endosymbiotic theory clusters into five categories. Each is independently checkable, and together they make the case hard to escape.
Double membranes. Mitochondria and chloroplasts have two membranes. The inner one resembles a bacterial plasma membrane in lipid composition and protein transport machinery; the outer one resembles the host's endomembrane. The leading explanation is that the inner membrane is the original bacterial membrane and the outer membrane came from the host's phagocytic vesicle.
Their own DNA. Both organelles carry a small circular genome. Human mitochondrial DNA is about 16,569 base pairs and encodes 37 genes — 13 proteins, 22 tRNAs, and 2 rRNAs. That number is worth holding onto: it's tiny compared with the roughly 20,000 protein-coding genes in the human nuclear genome.
Bacterial-style ribosomes. Mitochondrial and chloroplast ribosomes are 70S, the bacterial size, not the 80S eukaryotic size. This is why certain antibiotics that target bacterial ribosomes — chloramphenicol, tetracycline, aminoglycosides — can interfere with mitochondrial protein synthesis. It's the mechanistic reason behind a real clinical side effect.
Size and division. Mitochondria are roughly 0.5 to 1 micrometer in diameter, in the same range as many bacteria. They divide by binary fission-like processes using proteins related to bacterial FtsZ in plants and algae, and by dynamin-related machinery in animals.
Gene transfer to the nucleus. Most of the original symbiont's genes now sit in the host nucleus. The mitochondrial genome shrank from an estimated 1,000–5,000 ancestral genes to the 37 in humans. Those genes moved to the nucleus and their protein products are now imported back into the organelle via targeting sequences.
| Evidence line | Mitochondrion | Chloroplast | Bacterial match |
|---|---|---|---|
| Membranes | Double | Double | Inner membrane bacterial-type |
| Genome | Circular, ~16,569 bp in humans | Circular, ~120,000–160,000 bp in plants | Circular bacterial chromosome |
| Ribosome | 70S | 70S | 70S bacterial ribosome |
| Size | ~0.5–1 µm diameter | ~2–10 µm long | Bacterial range |
| Division | Fission, dynamin/FtsZ-related | Fission, FtsZ-based | Binary fission |
| Gene count now | 37 in humans | ~120 in Arabidopsis | Shrunk from thousands |
The Host Was an Archaeon: The Two-Domain Tree and the Asgard Connection
The host lineage question was genuinely unresolved until the 2000s and 2010s. The leading answer now: the host was an archaeon, specifically one related to the Asgard archaea.
Asgard archaea were identified from environmental DNA and named for Norse gods — Lokiarchaeota, Thorarchaeota, Heimdallarchaeota, Odinarchaeota. In 2015 a Lund University team led by Thijs Ettema reported Lokiarchaeota from deep-sea sediments near a hydrothermal vent field between Greenland and Norway. Their genomes contained eukaryotic signature proteins — actin-related cytoskeletal components, small GTPases, ESCRT machinery — none of which had been seen in prokaryotes before.
That reshaped the tree. The two-domain tree places eukaryotes as a branch within the archaea rather than as a separate domain, with the alphaproteobacterial merger as the event that split the eukaryotic lineage off. Some researchers argue for a three-domain tree where the merger happened earlier and differently; the debate is live in the literature.
A 2020 cultivation milestone mattered here. A Japanese team reported culturing a Lokiarchaeon, Prometheoarchaeum syntrophicum strain MK-D1, from deep-sea sediment, after a 12-year effort. It grew only in coculture with other microbes — it needed partners — and its cells had long, branching protrusions. That's not proof of phagocytosis, but it's suggestive.
The Mitochondria-First Versus Mitochondria-Late Debate
When did the mitochondrion arrive relative to other eukaryotic features? Two positions dominate, and the difference changes the story.
Mitochondria-early holds that the alphaproteobacterial merger happened near the root of the eukaryotic lineage, before the nucleus and other organelles were fully formed. The host was a relatively simple archaeon. Genes transferred from the symbiont drove much of eukaryotic complexity.
Mitochondria-late holds that the host was already a complex cell with a nucleus and internal membranes, and the mitochondrion was acquired afterward. This was more popular in the 1980s and 1990s when the "archezoan" hypothesis claimed some amitochondriate eukaryotes were primitively so.
The archezoan hypothesis mostly collapsed. Groups once thought to lack mitochondria — Giardia, Trichomonas, Entamoeba — were found to carry mitochondrial remnants: mitosomes or hydrogenosomes. Giardia lamblia has mitosomes; Trichomonas vaginalis has hydrogenosomes. These derived from mitochondria and lost most functions. So the "late" camp lost its key evidence, and the current consensus leans mitochondria-early.
The practical implication is that the merger wasn't a late addition to an already-complex cell. It's increasingly seen as the trigger for eukaryotic complexity, not a consequence of it.
Secondary and Tertiary Endosymbiosis: The 3-Membrane and 4-Membrane Exceptions
Here's where the simple version of the theory starts to fail, and where it gets interesting. Chloroplasts didn't arise just once. They arose once as a primary endosymbiosis — a cyanobacterium engulfed by a eukaryotic host — and then spread by secondary and tertiary events.
Primary plastids have two membranes and are found in the Archaeplastida: green plants, red algae, and glaucophytes. That's the single cyanobacterial event.
Secondary plastids occur when a eukaryote engulfed a eukaryotic alga that already had a primary plastid. The result has three or four membranes, and sometimes a vestigial nucleus called a nucleomorph. Euglenids, chlorarachniophytes, dinoflagellates, diatoms, and brown algae all trace to secondary events — and at least one tertiary event where an alga with a secondary plastid was engulfed again.
The membrane count is diagnostic. Two membranes means primary; three or four means secondary or tertiary. The cryptophyte and chlorarachniophyte nucleomorphs are the surviving remnants of a red or green algal nucleus — one of the most direct pieces of evidence for serial endosymbiosis anywhere in biology.
Watch for the edge case that trips people up: Paulinella chromatophora is a cercozoan amoeba with a photosynthetic organelle that looks like a plastid but arose from a separate cyanobacterial endosymbiosis roughly 60 to 100 million years ago, independent of the canonical primary event. It's a putatively ongoing endosymbiosis, one of the few cases where researchers can watch the process in mid-flight.
| Event | Membranes | Lineages | Age estimate |
|---|---|---|---|
| Primary plastid | 2 | Archaeplastida (plants, red algae, glaucophytes) | ~1–1.5 billion years |
| Secondary plastid | 3–4 | Euglenids, diatoms, brown algae, dinoflagellates, cryptophytes | ~500 million to 1 billion years |
| Tertiary plastid | 4 | Some dinoflagellates | Later still |
| Paulinella chromatophore | 2 | One amoeba genus | ~60–100 million years |
Why Endosymbiotic Theory Isn't the Same as Symbiogenesis or the Hydrogen Hypothesis
Three terms get confused constantly, and the confusion matters when you're reading primary literature.
Endosymbiotic theory is the specific claim that mitochondria and chloroplasts descend from prokaryotes that were engulfed and retained. It's a statement about two organelles and their origins.
Symbiogenesis is Margulis's broader framework: the idea that new cell types and species can arise through merging of distinct organisms. Endosymbiotic theory is a subset of symbiogenesis, but symbiogenesis also includes claims about flagella or nuclei that were never confirmed.
The hydrogen hypothesis, proposed by William Martin and Miklós Müller in 1998 in Nature, is one specific model for the merger. It argues the host was a methanogenic archaeon that received hydrogen and carbon dioxide from a facultatively anaerobic bacterium, and the symbiosis began as a metabolic exchange before any phagocytosis. It's a model for how the merger started, not a separate theory.
The misuse: calling any case of two organisms living close together endosymbiosis in the technical sense. Corals and photosynthetic algae live in symbiosis, but the algae aren't organelles — they're not inherited strictly through the host lineage and haven't lost independent genomes. That's not what endosymbiotic theory claims.
Gene Transfer in Action: How 99% of the Symbiont's Genome Moved House
The most underrated part of the story is the aftermath. The mitochondrion didn't arrive fully formed as a co-equal partner; it arrived as a bacterium and was systematically disarmed.
The mechanism is endosymbiotic gene transfer. Over time, chunks of the symbiont genome were copied into the host nucleus, and the symbiont's copies were lost. The transferred genes acquired nuclear promoters, introns, and targeting sequences for import back into the organelle. A protein made in the cytoplasm is imported through the TOM and TIM complexes on the mitochondrial membranes.
The numbers are striking. Human mitochondrial DNA carries 37 genes. The nucleus contains roughly 1,000 to 1,500 mitochondrial proteins. So the ratio runs above 30:1 in favor of nuclear-encoded mitochondrial genes. In Arabidopsis thaliana, the chloroplast genome holds about 120 genes while the nucleus holds an estimated 2,000–2,500 chloroplast proteins.
Why keep anything? Organelle genomes retain mainly genes for hydrophobic proteins that resist import — the core membrane subunits of the electron transport chain in mitochondria, Rubisco's large subunit in some plastids. The "hydrophobicity hypothesis" argues those proteins can't be threaded through the import machinery, so the genes stay in place. The hypothesis has critics, but the pattern is consistent.
Endosymbiotic Theory in 2026: What's Settled, Contested, and Still Unknown
The core claim is settled. Mitochondria descend from alphaproteobacteria and chloroplasts from cyanobacteria; that is treated as established in every major cell biology text. No serious challenge to the core exists.
Contested: the precise host, the timing of the mitochondrial event relative to other features, and whether phagocytosis was involved at all. Some researchers argue the merger could have happened without full phagocytosis, via protrusions or membrane fusion. The 2020 Prometheoarchaeum cultivation work supports protrusion-based contact, not engulfment.
Also contested: how many times plastids were acquired, and where exactly the Asgard archaea sit. New draft genomes from environmental samples appear regularly. Treat any specific tree topology as provisional.
Fundamentally unknown: the exact identity of the alphaproteobacterial ancestor. Every published candidate — based on mitochondrial gene phylogenies — ends up as a lineage with no close cultured relatives today, which may reflect a highly derived symbiont rather than a real gap in sampling. The ancestor is presumably extinct as a free-living organism.
The theory's continuing value is that it makes predictions. If it's right, undiscovered lineages should show intermediate stages; organelle genomes should keep shrinking in lineages where selection is relaxed; and putting mitochondrial genes into a bacterial genome should be hard in specific ways tied to import machinery. The Paulinella system and ongoing Asgard sampling are exactly where those predictions get tested.
FAQ
What single piece of evidence most strongly supports endosymbiotic theory?
Mitochondrial and chloroplast genomes are circular and maternally inherited, and their genes cluster with bacterial sequences in phylogenetic trees. Specifically, human mitochondrial DNA's 37 genes on a 16,569-base-pair circle sit inside the alphaproteobacterial radiation — not near eukaryotic nuclear genes. That tree placement would be nearly impossible to explain if the organelle evolved from host membranes.
Do all eukaryotes have mitochondria?
No, but the exceptions are derived, not primitive. Several anaerobic eukaryotes — Giardia lamblia, Trichomonas vaginalis, Entamoeba histolytica — lack typical mitochondria but carry relics: mitosomes or hydrogenosomes. Trichomonas hydrogenosomes produce ATP anaerobically and likely descended from mitochondria. So no known eukaryote lineage is truly ancestrally amitochondriate on the current evidence.
Is endosymbiotic theory the same as what Margulis argued in 1967?
Mostly, with one important shift. Margulis's 1967 paper argued for a bacterial origin of mitochondria and chloroplasts, which is now consensus. But she also pushed symbiogenesis as a general driver of major evolutionary transitions, including claims about undulipodia and the nucleus that never gained support. The organelle origins stuck; the broader framework did not.
How long ago did the mitochondrial endosymbiosis occur?
Molecular clock analyses place it roughly 1.5 to 2 billion years ago, though the uncertainty range spans several hundred million years and depends heavily on calibration points. There is no fossil record of the event itself. The date is inferred from genome divergence, and different models give different answers — the range, not a point value, is the honest figure.
Why do chloroplasts have three or four membranes in diatoms but two in plants?
Membrane count reflects how many engulfment events occurred. Plants sit in the Archaeplastida, which descend from a single primary cyanobacterial engulfment yielding two membranes. Diatoms and other chromalveolates descend from a eukaryote that engulfed a red alga, adding the alga's plasma membrane and its plastid's two membranes on top of the host's — three or four total.
Can a bacterium become an organelle today?
It appears to be happening, at least once. Paulinella chromatophora carries a photosynthetic organelle called a chromatophore that arose from an independent cyanobacterial endosymbiosis roughly 60 to 100 million years ago — separate from all canonical plastids. Its genome has already shrunk and begun transferring genes to the host nucleus, which is exactly the pattern the theory predicts for an organelle mid-formation.
Does endosymbiotic theory apply to the nucleus?
No. The nucleus's origin is a separate question, and no bacterial antecedent is accepted for it. Proposed models involve invagination of membranes in an archaeal ancestor or a viral-eukaryogenesis hypothesis, both contentious. Some of Margulis's later writing grouped the nucleus into symbiogenesis claims, but that part never carried the evidence the mitochondrial and chloroplast claims did.
What antibiotic targets mitochondria, and why?
Aminoglycosides, chloramphenicol, tetracyclines, and macrolides can interfere with mitochondrial 70S ribosomes because those ribosomes resemble bacterial ones. That's the mechanistic basis for side effects like aminoglycoside-induced ototoxicity and chloramphenicol-associated bone marrow suppression in susceptible patients. This article is general educational information, not medical advice — anyone weighing antibiotic risks or benefits, including patients with mitochondrial disease, should consult a qualified clinician who knows their history.