1 One Ancestor for All Life
In the summer of 2024, a team at the University of Bristol in Britain published a portrait in Nature Ecology & Evolution: the common ancestor of every form of cellular life on Earth.
Here is the portrait. It lived about 4.2 billion years ago. Earth had taken shape only a little over 300 million years earlier. The Moon was much closer than it is today, and the oceans had only just settled into place.
Geologists call this earliest stretch of time the Hadean, after the god of the underworld in Greek mythology. They once imagined an Earth as hellish as the name, with molten rock everywhere. Zircons from the Jack Hills in Australia later changed that picture. In crystals 4.4 billion years old, the oxygen isotopes indicate that liquid water had already been involved in their formation. The Sun was about 30 percent dimmer than it is today. A much larger Moon hung in the sky, and over the course of a day the tides rose and fell far more violently than ours do.
It was already a proper cell, roughly as complex as a bacterium today, with a genome of about 2.75 million base pairs and some 2,600 genes. An ordinary present-day bacterium has around 4,000 genes.
It did not use oxygen. There was scarcely any in the air to use. It drew electrons from hydrogen, assembled carbon dioxide into organic matter step by step, and released small molecules such as acetic acid.
It already possessed the full machinery for making proteins, along with a rotary machine for producing energy. It may even have had an immune system for dealing with viruses: remember a stretch of an intruder's sequence, recognize it when it returns, and cut it apart. Nor was it alone. Its waste probably fed other microorganisms living on the same seabed.
This is the ancestor shared by every bacterium, every archaeon, every animal, plant, and fungus alive today, including you, reading this essay. Biologists call it the last universal common ancestor. Its initials, LUCA, sound like the familiar name Luca, and many people simply call it that.
Not a single fossil underlies this portrait. Almost no rocks from 4.2 billion years ago have survived, and those that have could never come with a label saying, “This is it.”
The portrait was worked out backward from organisms alive today: compare what they share and what they lack, then retrace the steps into the past. This essay is about how that portrait was drawn, which strokes are steady, and which still tremble.
The idea began with Darwin. In 1837, in a notebook he carried with him, he sketched a little branching diagram. A single trunk divided into several branches, and those branches divided again. Above it he wrote two words: “I think.” Twenty-two years later, at the close of On the Origin of Species, he argued from analogy that all the organisms that had ever lived on Earth probably descended from some single primordial form.
Darwin supplied the idea, but no way of finding the ancestor. More than a century would pass before people learned to read molecules and turned the idea into a question that data could answer.
First, the name needs clearing up. “Last” means the most recent point at which all the lineages meet when we trace them backward. It does not mean the first life ever to appear.
Our own species offers a useful example. Trace the family trees of everyone alive on Earth today into the past, and at some point they converge on the same person. A model published in 2004 estimated that this person might have lived only a few thousand years ago.
As the fifth essay will explain, “mitochondrial Eve,” the maternal ancestor of all modern humans, lived about 200,000 years ago. But that is the result of following just one path backward, from mother to mother. Humanity is of course far older than either of these dates. Countless people lived before that common ancestor; their bloodlines either died out or had long since merged into the same line.
The same is true of life. Before the last universal common ancestor, there may already have been a long history of living things: the earliest membranes, the earliest metabolism, the earliest molecules capable of copying themselves.
Many other forms of life may also have lived alongside it. They have no descendants alive today, so they do not appear on this tree. We will return to them in Section 8.
The last universal common ancestor, then, is the name of a relationship. It is defined as “the point where all descendants alive today converge when traced backward.” Suppose every bacterium went extinct tomorrow, leaving only archaea and their descendants. That meeting point would immediately move forward in time, into the archaeal branch itself. The ancestor that lived 4.2 billion years ago would not have changed. What changed would be who was still alive.
Relationships do not remain in rocks. Even if a cell from four billion years ago had been preserved intact and placed under a microscope, there would be no way to prove that it occupied precisely that node. It might be the ancestor's cousin, the ancestor's own ancestor, a side branch from the same period, or one of its descendants. This is an object we can infer only by working backward from its living descendants.
What this series calls a construct comes down, in the end, to drawing a line: what falls inside counts; what falls outside does not. People can draw the line, as with a set of criteria. Nature can draw it too, as with a membrane.
In this essay, the line is drawn around “having descendants alive today.” All who have living descendants converge on this node. Those who have none fall outside the line, however richly they may once have flourished.
2 The Shared Machinery
The first stroke of the portrait comes from a startling fact: every cell on Earth uses the same core machinery.
In 1961, Nirenberg and Matthaei, at the US National Institutes of Health, put a piece of artificial ribonucleic acid consisting of a single letter repeated over and over into a test tube. They added protein-making machinery extracted from cells. Out came a long protein chain made entirely of phenylalanine.
The first “word” of the genetic code had been deciphered: three identical letters stood for one amino acid. Within a few years, all 64 codons had been read. These 64 three-letter combinations specified 20 amino acids and the stop signals that say, “This far, and no farther.” Many amino acids had several synonymous spellings. In 1968, Nirenberg shared a Nobel Prize with two other researchers who had worked on the genetic code.
What came next was more surprising still. Bacteria, yeast, maize, mice, and humans used almost exactly the same codebook. Whatever amino acid a three-letter combination specified in Escherichia coli, it specified the same one in a human being.
Only mitochondria and a handful of organisms had made small revisions. In human mitochondria, for example, a combination that means “stop” elsewhere is read as tryptophan.
In 1968, Crick called this codebook a “frozen accident.” Chance may have played a large part in establishing it. But once every protein was written according to its rules, changing any one word would make thousands upon thousands of proteins go wrong at once. And so the code froze.
The codebook is not the only thing they share. The machine that makes proteins is called the ribosome, and in almost every kind of cell it follows the same design: two subunits, one large and one small, with the messenger held between them, linking amino acids together one at a time.
Around the year 2000, researchers finally resolved the ribosome's atomic structure. They found that the step that joins one amino acid to the next is carried out by ribonucleic acid inside the ribosome; the surrounding proteins are merely helpers. Inside the ribosome, ribonucleic acid does the real work. This discovery convinced many people that, before life used DNA and proteins, it had passed through a stage that depended mainly on ribonucleic acid. In 2009, the three scientists who had resolved the ribosome's structure shared the Nobel Prize in Chemistry.
The 20 amino acids that make up proteins are the same set everywhere. Apart from glycine, which has no left- or right-handed form, life uses the left-handed versions; the right-handed ones are almost never used. Why the left hand? There is still no settled answer.
In a meteorite that fell near the Australian town of Murchison in 1969, researchers found dozens of amino acids. A few had a slight excess of their left-handed versions. Some have wondered whether a process in space gave life's preference a nudge before those compounds ever reached Earth.
The energy molecule with which cells pay their bills is always the same: adenosine triphosphate. Over a single day, an adult human makes and uses up an amount of it roughly equal to their own body weight. Bacteria use the same molecule. Genes are written in DNA, copied into RNA, and then translated into proteins.
It is unlikely that this whole collection was independently invented several times over. The simplest explanation is that it all came from the same source.
In 2010, Theobald, at Brandeis University in the United States, turned this “simplest explanation” into a formal statistical test. He selected twenty-odd proteins found in all life and compared two models: all life descended from one common ancestor, or life arose in several independent origins. Which better explained the sequences we see today?
The result: the odds in favor of the common-ancestor model were ten to the power of 2,860 to one. The total number of atoms in the universe is only about ten to the power of 80.
Shared machinery makes comparison possible. Every organism contains a set of molecules doing the same jobs. The molecule responsible for a given job is assembled from different sequences in different organisms, and those sequences can be compared letter by letter.
The fewer differences two organisms have in the same molecule, the more recently they usually parted ways. Compare many organisms in pairs, and a map of relationships emerges: which split came first, and which came later. This is the tree of life at the molecular level.
You can draw a tree from appearances too, but that only takes you so far. Dogs resemble wolves, so they are close relatives. Dogs do not resemble fish, so the relationship is distant. Once you cross between very different groups, appearances cease to be reliable. Two organisms can look alike because they lead similar lives, without sharing a particularly recent ancestor.
We will encounter this problem again in the ninth essay, on the Ediacaran. Resemblance may come from a shared ancestor, or from separate lineages independently arriving at a similar form. Only the first kind can be used to build a tree. Molecules make it much easier to tell the two apart.
Follow the tree backward, and all its branches meet in one place. That meeting point is the last universal common ancestor.
3 The Shape of the Tree
Before reconstructing an ancestor, we need to know what the tree looks like. The shape of this tree was itself the subject of a forty-year dispute.
The tree of life had been drawn several times before Woese. In 1866, Haeckel, in Germany, drew one with three main branches: plants, animals, and a third branch holding all sorts of single-celled organisms.
In 1969, the American biologist Whittaker divided life into five kingdoms: animals, plants, fungi, protists, and bacteria. These trees were drawn from visible form and ways of feeding. When it came to deciding which single-celled organisms were close relatives and which were distant, there was scarcely anywhere to begin.
In 1977, Woese and Fox, at the University of Illinois in the United States, did something that seemed decidedly odd at the time. They looked neither at form nor at metabolism. Instead, they chose a single molecule with which to compare all single-celled life: a stretch of ribonucleic acid in the ribosome.
There were three reasons to choose it. Every cell has it, so it can be used to compare all organisms. It changes slowly, preserving signals from the deep past. Yet it is not wholly unchanging: enough differences remain to make comparisons possible.
The stretch of ribonucleic acid Woese chose became a standard tool for identifying microorganisms. Today, when an unfamiliar bacterium is found in soil, seawater, or the human gut, the first step is almost always to sequence that stretch and place the organism on the tree Woese drew.
There were no automated sequencing machines then. Woese labeled the ribonucleic acid with radioactive isotopes, cut it into fragments with enzymes, spread the fragments across filter paper, and pressed photographic film against it. Then he identified the spots on each sheet of film, one by one. For more than a decade, he studied thousands upon thousands of those sheets on the laboratory light box.
The conclusion they reached was deeply unsettling at the time. Until then, life had been divided into two broad categories: organisms with a cell nucleus, and organisms without one. Bacteria belonged to the second. “No nucleus” had become a catchall basket.
Woese found that the basket actually held two profoundly divergent lineages. One consisted of ordinary bacteria. The other included methane-releasing microorganisms living in swamps, hot springs, and cows' stomachs. The several hundred liters of methane a cow burps out each day are produced mainly by these microbes in its stomach. The two lineages differed from each other as much as each differed from organisms with nuclei.
He called the second group archaebacteria; the name was later changed to archaea. In November 1977, the New York Times put the research on its front page, under a headline announcing that scientists had discovered a form of life older than complex organisms such as plants and animals.
Many colleagues did not believe it, and Woese spent years out in the cold. In 1990, he joined Kandler in Germany and other researchers in formally establishing bacteria, archaea, and the nucleus-bearing eukaryotes as the three highest-ranking groups of life. They called them domains, a rank above the old kingdoms. In 2003, Woese received the Crafoord Prize, established for disciplines outside the Nobel Prizes' remit. He died in 2012.
In the three-domain picture, bacteria stand on one side, while archaea and eukaryotes are equal sister groups. In 1984, Lake, at the University of California, Los Angeles, in the United States, proposed a different drawing: eukaryotes were not the sister of archaea but a branch growing from within them.
He was comparing the shapes of ribosomes. Under the electron microscope, eukaryotic ribosomes and those of a certain group of heat-tolerant archaea had the same small protrusions. He called those archaea “eocytes”: dawn cells.
The idea did not become the mainstream view. There were too few features to compare, and too narrow a sample of archaea. The deepest branches were also prone to a misleading effect: two rapidly evolving branches could be pulled together on the diagram even when they were unrelated. This is called long-branch attraction.
The turning point came from the floor of the Arctic Ocean. On the mid-ocean ridge between Greenland and Norway lies a field of hydrothermal vents discovered in 2008. Its black chimneys resembled a castle, and its discoverers named it Loki's Castle, after the mercurial god of Norse mythology. In 2015, a team at Uppsala University in Sweden assembled the genome of an entirely unknown kind of archaeon from mud on the nearby seabed. They called it Lokiarchaeota.
Its route into science was unusual. No one had seen the organism. The researchers extracted all the genetic material from a lump of mud, sequenced the lot, then used computation to piece together the fragments that belonged to the same organism.
The assembled genome contained relatives of proteins previously thought to exist only in eukaryotes: skeletal proteins that give cells their shape, and proteins that package and ship cargo in membrane vesicles. Two years later, the group was expanded into a large family. The researchers stayed with Norse mythology and named it Asgard archaea, after the home of the gods.
Branches within the family were named for individual Norse gods: Loki, Thor, Odin, and Heimdall. On the evolutionary tree, eukaryotes fell inside this family. A large-scale analysis in 2023 placed their closest relatives in a small group within the Heimdall branch.
Then someone managed to grow one. In 2020, Hiroyuki Imachi and colleagues at the Japan Agency for Marine-Earth Science and Technology reported that, after about twelve years, they had finally cultured a strain of Asgard archaea from seabed mud more than 2,500 meters deep in Japan's Nankai Trough.
For twelve years they had recreated the conditions of deep-sea sediments in the laboratory, gradually enriching this extraordinarily slow-growing organism. A single cell division takes about twenty days. They named it after Prometheus, the fire-stealer of Greek mythology.
The cultured cells were tiny, about half a micrometer across, and spherical, with long, slender, branching tentacles extending from them. Inside there was no nucleus, no internal membrane, and no eukaryotic organelle of any kind.
They broke down amino acids, releasing hydrogen and formic acid. Partner microorganisms nearby had to keep consuming these products for the reaction to be energetically worthwhile. They could not do without their partners.
A little archaeon with long tentacles, unable to live without its partners, gave a living form to the idea that “the eukaryotic cell arose from a long symbiosis between archaea and bacteria.” The fifth essay will tell the story of that symbiosis.
The prevailing view today is that eukaryotes grew from within archaea, and that the tree of life has two branches at its highest level: bacteria and archaea. Exactly which Asgard branch is closest to eukaryotes remains disputed.
Finding the root of the tree is a separate problem. For an ordinary map of relationships, you can use a group known to lie outside the set you are studying as a reference, looking inward from its side. This is called an outgroup.
The tree of all cellular life has no outgroup. There is no fourth kind of life to serve as a reference. In 1989, Takashi Miyata's team at Kyushu University in Japan and Gogarten and colleagues in the United States independently hit upon the same ingenious solution. Some genes had duplicated before the last universal common ancestor, leaving two copies that subsequently evolved separately. Each copy could serve as an outgroup for the other. Using two kinds of ancient duplicated genes, the two teams arrived at the same orientation: the root lay between the bacterial branch and the branch containing archaea and eukaryotes.
The shape of the tree directly determines the reconstruction. In a three-domain tree, a gene found in bacteria, archaea, and eukaryotes appears to have been retained independently three times, making it easy to assign to the ancestor.
In a tree with two main branches, eukaryotes are a later offshoot within archaea. “Eukaryotes have it too” no longer supplies an independent third piece of evidence. The deep comparison that actually matters for reconstructing the ancestor is between the bacterial side and the archaeal side.
4 What the Ancestor Had
In 2016, Martin's team at the University of Düsseldorf in Germany drew up the first serious inventory of the ancestor's genes.
They took about 6.1 million protein-coding genes from nearly 2,000 bacterial and archaeal genomes. First they sorted them into groups with similar functions and similar sequences. Genes in the same group were considered descendants of a single ancestral gene. Altogether, they assembled about 286,500 such groups.
Then came the filtering. To count as “already present in the ancestor,” a group of genes had to occur in several major groups of both bacteria and archaea. Its own internal tree of relationships also had to be clean: the bacterial sequences clustered on one branch, the archaeal sequences on another, with a clear divide between them.
The rule was meant to keep out genes that had arrived sideways. We will come to those in Section 6. After the filtering, 355 groups remained. Of every thousand groups, only one was clean enough to pass.
The ancestor inferred from these 355 groups looked like this: it did not use oxygen, obtained electrons from hydrogen, and could turn carbon dioxide into organic matter. Many of its enzymes had iron and sulfur at their active centers. It liked high temperatures and could also fix nitrogen from the air.
From this, Martin's team concluded that it had lived on the seafloor somewhere rich in hydrogen, carbon dioxide, and iron, much like an alkaline hydrothermal vent today. Lost City, in the middle of the Atlantic, which the third essay will discuss, offers a living model of such a place, with white chimneys up to sixty meters tall. A deep-sea expedition stumbled upon it in 2000, on a submarine mountain near the Mid-Atlantic Ridge. It releases warm, alkaline water, quite unlike the scalding, acidic hydrothermal fluids of black smokers.
In 2003, Martin and the British geochemist Russell had jointly set out a systematic case for life's origin in these alkaline hydrothermal vents. Reactions between rock and seawater release hydrogen, while the difference in acidity between the inside and outside of a vent acts like a natural battery. The pathway by which the ancestor assembled carbon dioxide into organic matter is still used by a group of bacteria called acetogens. It may be the oldest carbon-fixation pathway on Earth.
Eight years after the 2016 study, the Bristol team tackled the same task by an entirely different route. They first built a large tree from the genomes of 350 archaea and 350 bacteria. Then, one by one, they compared the smaller trees of several thousand gene families against it.
Instead of throwing out suspect genes, they explicitly allowed horizontal transfer, duplication, and loss in their model, then calculated the probability that the ancestor possessed a given gene. Under the strictest criteria, they recovered 399 functional units, in the same range as the Martin team's 355.
Then they went a step further. They added up the probabilities of presence for every gene: high, middling, and low. Using the proportion of all genes accounted for by reliable core genes in modern genomes, they estimated the size of the ancestor's complete genome: about 2.75 million base pairs and roughly 2,657 protein-coding genes. This is where the portrait in Section 1 came from.
There is a nearly tenfold difference between 355 and 2,657. Both numbers are right; they answer different questions. Think of taking stock in an old storeroom.
One approach records only boxes whose labels are intact and whose provenance can be checked. The register ends up with 355 entries. The other examines every box, assigns each a probability that it “was here to begin with,” then adds those probabilities together and arrives at 2,657.
The first misses more; the second depends more heavily on estimates. Almost every entry in the register is reliable, but many boxes that really belonged in the storeroom have been left out. The estimation method leaves none out, but every number rests on an estimate.
Earlier studies, working by different rules, had produced figures of 80, 572, and around 1,500. They are like a row of rulers marked with different scales, each accurate on its own terms.
More than 2,600 genes can do a great deal. In the ancestor's genes, the Bristol team found the complete machinery for making proteins, the central metabolic network, the energy machinery, and components of an immune system that bacteria today use to remember viruses and cut them apart.
The discovery of that immune system is a story in itself. In the 1990s, while studying archaea in coastal salt ponds, Mojica at the University of Alicante in Spain noticed strange repeated sequences in their genomes. Between the repeats lay fragments of unknown origin.
It took him more than a decade to work out that the fragments came from viruses that had invaded in the past. A cell copies a small piece of the virus into its own genome and keeps it on file. When the same virus enters again, the cell consults the file and cuts it apart. In 2012, the American researcher Doudna and the French researcher Charpentier turned these scissors into a gene-editing tool; in 2020, they received the Nobel Prize in Chemistry. If the Bristol inference is right, the ancestor of those scissors was already in use more than four billion years ago.
A cell capable of holding a grudge tells us that viruses were already circling nearby. In the model, the acetic acid and hydrogen it released could feed other microorganisms. Together, they formed a small ecosystem. There may already have been a food web on the seafloor 4.2 billion years ago.
There is another way to investigate whether it liked the heat: “resurrect” ancient proteins. Researchers first infer the sequence of an ancestral protein from billions of years ago using present-day sequences. They then synthesize it in the laboratory and test how hot it can get and still work.
In 2008, American researchers used this method to reconstruct an ancient factor involved in protein production in an ancestor of bacteria. It remained stable at temperatures in the 60s Celsius. Another ancient enzyme later reconstructed by Japanese researchers tolerated still more heat. A protein from billions of years ago, folding again, working again, in a test tube today.
The two studies do not agree entirely on a few features. The Martin team's ancestor clearly favored high temperatures; the Bristol team considered the evidence less one-sided. Some researchers also argue that the ability to fix nitrogen arrived later, transferred from elsewhere. The outline of the portrait is steady. The details still tremble.
5 What the Ancestor Lacked
Having counted what it had, turn the inventory around and count what it lacked. The missing pieces are precisely the most puzzling ones.
Start with the machinery that copies DNA. In 1999, Koonin and colleagues at the US National Institutes of Health systematically compared the components bacteria and archaea use to replicate DNA. They found a contrast that still holds today.
Some of the supporting processes associated with DNA, such as making raw materials and repairing damage, are homologous on the two sides: they descend from the same ancestor. But among the core components that actually copy a chromosome, several are different in the two groups. For the enzyme that builds a new strand letter by letter, bacteria use one class and archaea use one or two others. Their structures and sequences indicate that they are not two descendants of the same ancestral enzyme. The relationship is like that between a bat's wings and an insect's. The enzymes that start the copying and those that pull the two strands apart likewise belong to different families on the two sides.
Both great groups of organisms store information in double-stranded DNA. Both must open the double strand, make a beginning, and extend a new strand along a template. They do the same job, but several of the core parts doing it come in two different sets.
The observation has stood for more than twenty years. There are, however, four explanations for it, and none has yet been ruled out.
The first connects to a larger idea. In 1986, Gilbert, at Harvard University in the United States, gave that idea a name: the “RNA world.” Before DNA and proteins divided the labor between them, ribonucleic acid did both jobs, storing information and acting as a catalyst.
As Section 2 explained, it is ribonucleic acid that actually joins amino acids together inside the ribosome: a fossil of that world. Under this explanation, the ancestor still stored its information mainly in RNA, and modern DNA replication took its final form separately after the two branches diverged. The difficulty is all the supporting DNA-related machinery the two sides share. Why is that machinery so ancient?
A second explanation holds that the ancestor already used DNA, but that the two groups later replaced their replication cores separately. The problem is how to carry out such a wholesale replacement without stopping chromosome replication. It is rather like changing a car's engine while driving it.
A third says that the ancestor possessed two, or even several, replication systems at once, and that bacteria and archaea subsequently lost different parts. This removes the need to suppose that replication was invented twice, at the cost of an ancestor more complex than we had imagined.
The final explanation brings in viruses. Early viruses carried replication machinery of their own and introduced different machines into different cellular lineages. Viruses remain a great storehouse of replication machinery today; many viral polymerases have no close relatives in cells. The difficulty is that viral lineages this ancient have left neither fossils nor a continuous record against which to check the idea.
Now consider the membrane. Hold out your two hands. They have the same shape, yet you cannot turn your left hand over and superimpose it on your right: the thumbs are always on opposite sides. Molecules can be like that too. The same atoms, joined in the same way, can come in two versions that mirror each other but cannot be superimposed.
This is the relationship between bacterial and archaeal membrane lipids. The backbone of bacterial lipids has one configuration; archaea use its mirror image. The chemical bonds joining the components differ, as do the long, oily chains attached to them.
All three layers of difference occur together. Biologists call this the lipid divide. Archaeal membranes are exceptionally tough. Some archaeal lipids connect at their ends and span the entire membrane like rows of staples. Archaea like these live in the nearly boiling acidic hot springs of Yellowstone National Park in the United States, going about their lives at temperatures and acidities that would long since have made other cells fall apart.
There are several ideas about the ancestor's membrane as well. Perhaps a modern membrane already existed, and the two groups later replaced the systems that made it. Perhaps the ancestor had a mixed membrane containing both kinds of lipids, with each branch later settling on one.
Perhaps it had only a simpler membrane, before either mirror-image configuration had become fixed. The most radical proposal is that it had no membrane of its own at all. It lived in tiny pores in the rock of seafloor hydrothermal vents, and the mineral walls of those pores served as its cell walls.
Mixed membranes were once thought impossible: lipids of opposite handedness, it was argued, would be unstable together. In 2011, researchers showed with artificially mixed membranes that such a combination could be stable.
In 2018, researchers at the University of Groningen in the Netherlands engineered E. coli to make large amounts of archaeal-type membrane lipids alongside its own bacterial lipids. The modified cells continued to grow and divide. One cell can live inside a membrane containing both mirror-image forms at once.
We should also be clear about what the ancestor certainly had. The core parts of the protein-making ribosome are highly homologous across all life, far more uniform than the machinery of replication.
The genetic code is almost the same everywhere. The small molecules that transport amino acids have an extremely conserved basic structure. And although the bacterial and archaeal versions of the rotary machine that makes the cell's energy currency have diverged deeply, both its working principle and ancient duplications of some of its components can be traced to before the two groups split.
The third essay will discuss this machine: a difference in ion concentration across a membrane drives its rotation, and each turn produces several energy molecules. Using an ion gradient across a membrane to generate power is a very ancient principle. Which ion was used at the beginning is still uncertain.
So the ancestor is a peculiar mixture. It had a complete set of translation and energy machinery characteristic of a highly developed cell, yet we cannot reconstruct a modern replication system homologous on both sides, nor establish what kind of membrane it had. Some take this to mean that it had not yet become fully cellular and was more like a loose community.
Bristol's figure of more than 2,600 genes points in the opposite direction, toward an ancestor with the organization of a prokaryotic cell. The missing pieces are the most intriguing blanks in the portrait.
6 Genes That Move Sideways
Every reconstruction rests on an assumption: genes are passed mainly from parents to offspring. In bacteria, that assumption is not secure.
In 1928, Griffith, a medical officer in Britain's Ministry of Health, performed a curious experiment while studying the pneumococcus. One strain had a capsule on its surface and could infect and kill mice. Another lacked a capsule and was harmless.
He killed the virulent strain with heat and injected it into mice together with the living, harmless bacteria. The mice died. From their bodies he isolated living, encapsulated, virulent bacteria. Something in the dead bacteria had transformed the living ones.
In 1944, Avery and colleagues at the Rockefeller Institute in the United States proved that the substance was DNA. Two years later, the Americans Lederberg and Tatum discovered that E. coli could also “conjugate” directly: one cell handed a stretch of DNA to another. Lederberg received a Nobel Prize for this at the age of thirty-three.
Genes have several ways to travel sideways. A bacterium can pick up a piece of DNA directly from the surrounding water and incorporate it into its own genome. It can receive a piece from another bacterium through a thin, temporary connecting tube.
A virus can also move it to a new home. On leaving one bacterium, the virus carries off a bit of its host's DNA and injects it into the next. None of these three routes requires the other organism to be a relative. This is called horizontal gene transfer.
It is close to everyday life. In 1959, Japanese researchers discovered that the resistance of dysentery bacteria to several antibiotics could pass as a complete package from one bacterium to another, and even from harmless E. coli to dysentery bacteria. The resistance genes were carried on a small, independently replicating ring of DNA called a plasmid. One important reason bacterial drug resistance spreads so quickly in hospitals is that whole stretches of resistance genes can pass between bacteria. They need not be relatives, or even members of the same species.
How common is it? In 2008, German researchers analyzed 181 bacterial and archaeal genomes and estimated that, on average, at least 80 percent of the genes in each genome had taken part in horizontal transfer at some point in their history. That is “at least once” accumulated over billions of years. It is like saying that 80 percent of people have moved house at least once in their lives, not that 80 percent are currently living in a house they have just moved into.
In 1998, Woese proposed something more radical. In life's earliest stages, he argued, the parts of cells had not yet become tightly locked together. They depended only loosely on one another, and a foreign part was easy to fit. Horizontal exchange was the normal state of affairs.
As cellular design became more integrated and the dependencies between parts grew tighter, an incoming version had to fit an ever greater number of partners at once. Horizontal replacement became harder, and stable vertical lineages gradually stood out. In 2002, he formally named this transition the Darwinian threshold.
Under this view, the tree was not there from the beginning. A tree-like shape may itself be a property that arose later in evolution and gradually grew clearer. This is a theoretical model. No one has measured the year in which the transition occurred.
In 1999, Doolittle, at Dalhousie University in Canada, wrote in Science that if the genomes of bacteria, archaea, and early eukaryotes all contained large numbers of genes mixed together from different sources, there might be no natural basis for forcing entire genomes into a strictly hierarchical classification.
He also said that classification based on molecular sequences remained indispensable. His doubt concerned something else: whether a single tree of life existed that was correct for every gene.
That same year, Lake and colleagues at the University of California, Los Angeles, in the United States, proposed an elegant hypothesis. Successful horizontal transfer is harder for genes in large, tightly coordinated systems than for genes that work more independently.
A ribosome has dozens of parts fitted closely together. An incoming part must match all those others at once before it can be accepted. An enzyme responsible for one step in metabolism usually has only to perform its own step properly.
Later research found that the tendency was real, but that the barrier was not absolute. In one heat-tolerant bacterium, researchers replaced the most classic stretch of ribosomal RNA used to build evolutionary trees with a version from a different phylum. The cell still survived.
Extensive horizontal transfer and a vertical backbone can coexist. With a few dozen to a hundred or more conserved core genes, researchers can still reliably recover the main framework of bacteria and archaea. The farther out toward the periphery, the denser the network becomes.
In 2018, researchers compared the genomes of 105 bacteria belonging to the same family. They found that the individual trees of 33 photosynthesis-related genes disagreed extensively with the tree of the species. At least seven transfers of the entire set were needed to explain its distribution today.
If a bacterial lineage can photosynthesize today, it need not have inherited that ability from its ancestor. The lineage of the species follows one line, and the lineage of the photosynthesis genes another. The two do not coincide.
7 When
When did the last universal common ancestor live? To answer that, we need a molecular clock.
In the early 1960s, Zuckerkandl and Pauling at the California Institute of Technology in the United States compared hemoglobin from different animals. They found that the longer ago two animals had diverged, the more their hemoglobin sequences differed, and that the differences accumulated at a roughly constant rate.
They proposed that the changes accumulating in molecules could be used as a clock. Imagine two texts copied many times over. The more errors they contain, the earlier their separate copying histories must have begun from a shared original.
The molecular clock caused trouble almost as soon as it appeared. In 1967, Sarich and Wilson at the University of California, Berkeley, in the United States, used immunological methods to compare blood-serum proteins in humans and chimpanzees. They calculated that the two had separated only about five million years ago.
The prevailing view among paleoanthropologists put the split at more than fifteen million years. The argument lasted for over a decade. In the end, the molecular-clock side largely won.
The analogy immediately brings two difficulties with it. One is that copying speeds differ. Some lineages change quickly, others slowly; the rate for the same molecule can vary severalfold between branches.
The modern approach allows rates to vary across the tree and estimates them statistically. The other difficulty is that the clock gives only relative timing. To turn that into a number of years, at least a few nodes must have their dates pinned down independently. Those pins can come only from fossils and rocks.
Here is how the Bristol team pinned down its 2024 estimate. The tree of all cellular life has no outgroup, so they used five groups of ancient genes that had already duplicated before the ancestor lived.
The same divergence leaves a node in each of the two copies, and those two nodes must have the same date. Tying them together gets around the absence of an outgroup. The team used thirteen calibration points. The oldest upper limit came from the Moon's formation about 4.51 billion years ago; the deepest biological calibration came from a geochemical constraint about 2.95 billion years old.
The result was about 4.2 billion years ago, with a credible interval of roughly 4.09 to 4.33 billion years ago. Changing the rate model moved the lower bound by about ninety million years. Taking the different analyses together, the range extended from about 3.94 to 4.52 billion years ago.
Earlier studies also pointed far back in time. In 2018, a study using 102 species and 29 conserved genes placed the ancestor before 3.9 billion years ago. In 2023, another Bristol study, this one of the energy machinery, placed it roughly between 4.52 and 4.32 billion years ago.
The hardest part is that all the pins lie a long way from the target. A fossil of a single-celled organism from four billion years ago can usually tell us only, “There was life here.” It can rarely say which bacterial branch that life belonged to, much less announce, “This is the last universal common ancestor.”
Most biological calibration points that can actually be used are more than a billion years younger than the target. Rates must then be extrapolated from those points back toward the root. Dates near the root depend heavily on how that extrapolation is done.
The rocks have clues of their own. The Jack Hills of Western Australia contain zircon grains 4.1 billion years old. In 2015, researchers found a tiny speck of graphite inside one of them. Its carbon was isotopically light, resembling the carbon left behind by living things.
This does not prove that life already existed then. But it does not conflict with the portrait of “a complex cell already present 4.2 billion years ago.” The previous essay discussed clear evidence of life around 3.5 billion years ago. The common ancestor must be older than that evidence.
Work out what this implies. Earth formed about 4.5 billion years ago. The Moon was born in a great collision about 4.51 billion years ago: a Mars-sized body struck the young Earth, and the scattered debris gathered into the Moon.
As the first essay explained, people once thought there had been an intense meteorite bombardment around 3.9 billion years ago. It now appears that this event may never have happened. By the Bristol dates, even if it did, life had already emerged before it and endured it. If the last universal common ancestor lived about 4.2 billion years ago, only some 300 million years separated the formation of Earth from the appearance of a cell with 2,600 genes.
And before that cell, there still had to be an earlier history of life, building membranes, the code, and ribosomes one by one. Life arrived on this planet astonishingly early.
8 Those That Left No Descendants
The definition of the last universal common ancestor requires only that every form of cellular life alive today converge on it when traced backward. It does not require it to have been the only life of its time.
Before the last universal common ancestor lies a longer story. In an 1871 letter to his friend Hooker, Darwin imagined life's beginnings: in some warm little pond, with ammonia, phosphates, light, heat, and electricity, a protein-like compound might form, ready to undergo more complex changes. Most of the protagonists in that history left no descendants.
There may have been lineages that branched off earlier, sister groups that lived alongside the ancestor and later went extinct, or communities of microorganisms exchanging genes so freely that their boundaries were hard to draw. Did they leave anything behind?
In 2005, Cleland, a philosopher at the University of Colorado in the United States, went so far as to suggest that a “shadow biosphere” might still be alive on Earth: microorganisms that share no ancestor with us and use a different chemistry, overlooked because all our detection methods were designed around the life we already know. No evidence for the idea has yet been found. But it puts a finger on something important: when we go looking for life, we cast a net made in our own image.
At present, no geochemical signal can reliably be identified as coming from “an independent lineage outside the descendants of the last universal common ancestor.” Carbon isotopes, sulfur isotopes, and ancient microscopic structures may establish biological activity, but they cannot tell us the identity of its lineage.
If a lineage left no descendants, passed no recognizable genes sideways to lineages that survived, and left no physical record bearing distinctive features of its lineage, then living genomes cannot, in principle, recover it. The information about it has been irreversibly lost.
If it did pass genes sideways to a surviving lineage, it might theoretically have left isolated, deeply branching genes in today's genomes that cannot be joined to the main trunk. But distinguishing “came from an extinct early lineage” from “evolved exceptionally fast” or “came from a living group we have not yet sampled” is extraordinarily difficult.
The entanglement of viruses and cells is still written into our own bodies. About eight percent of the human genome consists of traces inserted by ancient retroviruses, passed from generation to generation over tens of millions of years. Some have even been borrowed for useful work: a protein that helps cells fuse in the placenta comes from an ancient virus's envelope gene.
Viruses are often called upon to fill this gap. Some researchers have long argued that certain core viral replication genes are extremely ancient. Viruses may be more than fragments that later escaped from cells; they may even have helped bring different replication systems into early cells. This is the last of the explanations in Section 5.
Viruses do contain some widely distributed proteins with no straightforward counterparts in cells. But genes pass between viruses and hosts too frequently, and viruses have no universal marker equivalent to the ribosome. We cannot build a viral tree that can reliably be joined to the roots of bacteria and archaea.
Giant viruses once made people think a fourth kind of life had been found. In 1992, a microorganism parasitizing amoebae was isolated from water in a hospital cooling tower in Bradford, Britain. It was big enough to be seen under an ordinary microscope and was initially taken for a bacterium.
In 2003, French researchers established that it was actually a virus, with more genes than some bacteria. They named it Mimivirus, meaning “a virus that mimics bacteria.” Ten years later, researchers in Marseille, France, found still larger pandoraviruses off the coast of Chile and in a pond in Australia. Their genomes were larger than those of many bacteria, and the vast majority of their genes had no known counterparts in other organisms. In 2010, researchers comparing several giant-virus proteins found that some relationship trees placed them on a deep branch outside the three great groups of life. They proposed a “fourth domain.”
A reanalysis a year later attributed those deep branches to long-branch attraction and the choice of model. A broader analysis in 2014 concluded that giant viruses could have grown from small DNA viruses by gradually collecting genes from their hosts. No extinct fourth domain was needed. To this day, there is no generally accepted fourth cellular domain, nor any widely accepted candidate for a “shadow lineage.”
What a construct cannot contain, yet cannot make disappear, this series calls the remainder. The construct in this essay is a node with a clear position: trace every form of cellular life alive today backward, and they converge there. It has two remainders.
The first remainder is that the building material itself is in motion. For a tree to hold, genes must pass mainly from parent to offspring. Yet great numbers of bacterial genes travel sideways.
The core genes still preserve a recognizable vertical framework. Farther out, the pattern looks more and more like a net. If Woese was right, the tree-like shape itself only gradually crystallized later on. The stretch in which this node lies may be precisely the period when the tree had not yet fully taken shape.
The second remainder is the life that left no descendants. It lies beyond the reach of the evidence. The last universal common ancestor itself has no specimen because it is a relationship, and relationships do not remain in rocks. The forms of life that lived alongside it and left no descendants did not even leave a relationship behind. They are in neither the rocks nor any living genome.
This predicament, in which two kinds of absence cannot be told apart, will recur again and again. The Ediacaran organisms of the ninth essay and the fossil record of the eleventh pose the same question: when something is missing from the record, was it never there, or was it there but left no trace?
And so the node is drawn this way: what living organisms share locates it; what they all lack traces its outline. The shared part is continually stirred by horizontal transfer. Within the absent part, two things remain mixed together and inseparable: what never existed, and what existed but left no trace at all.
A cell from 4.2 billion years ago. Some 2,600 genes. Feeding on hydrogen, releasing acetic acid, perhaps already remembering viruses. Each year the portrait's outline grows clearer, while all around it remains an invisible blank. The criteria are still sifting. The genes are still moving sideways.