Major Evolutionary Transitions in Evolution

A major evolutionary transition is a change in the way biological information is stored, transmitted, or organized such that entities previously capable of independent replication can now replicate only as parts of a larger whole. Before the transition, the lower-level units — genes, cells, organisms — vary, compete, and reproduce on their own, and selection acts at that level. After the transition, the lower-level units still exist, but their fitness is bound up entirely with the success of the group they belong to; the group itself becomes a unit of selection, with its own heritable variation in survival and reproduction. This is the language of Levels of selection made concrete: transitions are events that create new levels. Maynard Smith and Szathmáry catalogued a sequence of such transitions, each building on the last:

Two structural features recur across this list and give the concept its analytic bite. First, division of labor: after a transition, formerly identical units differentiate into specialists that depend on each other — somatic cells specialize while germ cells alone reproduce, worker ants forage while queens alone reproduce, organelles fix carbon or make ATP while the nucleus alone controls heredity. Second, conflict suppression: because the lower-level units retain some capacity for self-interested replication, every transition is vulnerable to "cheats" — a somatic cell that becomes cancerous, an organelle that resists uniparental transmission, a worker that lays her own eggs. A transition is evolutionarily stable only if mechanisms exist that make cheating rare enough, or costly enough, that the higher-level unit's integrity persists across generations.

Transitions differ enormously in their molecular details, but they share a common architecture that can be broken into four steps, illustrated here with the classic case of the origin of the mitochondrion and, in parallel, the origin of multicellularity.

Step 1 — Formation of a group. Two or more independent replicators come into physical or reproductive association. For the mitochondrion, this was almost certainly an alphaproteobacterium engulfed by, or living syntrophically with, an archaeal host cell roughly 1.5–2 billion years ago. For multicellularity, this was a dividing cell that failed to separate from its daughter cells, or independent cells that aggregated, creating a physically joined group.

Step 2 — Division of labor. Once grouped, differential gene expression or differential loss of genes allows the units to specialize. The proto-mitochondrion shed most of its genome (a modern mitochondrion retains only 13 protein-coding genes in humans, compared to roughly 1,000–4,000 genes in free-living alpha-proteobacteria) and outsourced the rest to the host nucleus, becoming an obligate ATP-generating specialist that can no longer replicate outside the cell. In an early multicellular lineage, cells begin to express different gene sets depending on position — some specializing in reproduction, others in feeding, movement, or protection.

Step 3 — Establishment of a germline / vertical transmission bottleneck. For the new unit to be heritable as a whole, there must be a reliable mechanism that transmits the group, not just its parts, to the next generation. Mitochondria are transmitted through the egg cytoplasm (uniparental, maternal inheritance in most animals), which prevents different mitochondrial lineages within one organism from competing indefinitely and aligns the mitochondrion's fitness interest with the whole cell's. In multicellular organisms, a germline is set aside early in development (in animals, primordial germ cells are specified in the first few cell divisions in some lineages, or later via inductive signals in others), physically insulating reproductive cells from mutations and selfish variants that might arise in soma.

Step 4 — Suppression of within-group conflict. Even with specialization and a transmission bottleneck, selection at the lower level does not vanish — it is merely opposed. Mechanisms of conflict suppression include:

• Bottlenecking of organelle populations at each generation (reducing the number of mitochondria transmitted through the female germline to a small founding pool, on the order of tens to a few hundred, which increases variance among offspring and lets purifying selection at the organismal level purge deleterious mitochondrial mutants — this is the mitochondrial bottleneck).
• Policing behavior in social insects — workers destroying eggs laid by other workers ("worker policing"), which removes the fitness benefit of individual worker reproduction and stabilizes sterility.
• Apoptosis and immune surveillance suppressing cells that escape normal growth control (cancer as a recurring "reversion" of the multicellularity transition, discussed below).
• Fair meiosis — mechanisms that equalize the transmission chance of each allele at a locus, suppressing selfish "meiotic drive" elements that would otherwise spread by biasing their own transmission.

Because most major transitions happened hundreds of millions to billions of years ago, they cannot usually be watched directly; the evidence is a combination of comparative genomics, phylogenetics, and — crucially — experimental evolution that reproduces a transition under laboratory conditions where it can be observed happening in real time.

Endosymbiotic origin of mitochondria — the phylogenetic case. Lynn Margulis revived and championed the endosymbiotic theory in 1967, proposing that mitochondria and chloroplasts began as free-living bacteria engulfed by a host cell. The decisive evidence came from molecular phylogenetics once mitochondrial and chloroplast genomes were sequenced. Mitochondrial ribosomal RNA genes cluster, in gene trees, within the alphaproteobacteria — specifically nested near the order Rickettsiales — rather than with the nuclear genome's own phylogenetic affinities. Mitochondria retain their own circular genome (in humans, 16,569 base pairs encoding 13 proteins, 22 tRNAs, and 2 rRNAs), their own double membrane (the inner membrane homologous to a bacterial plasma membrane, the outer membrane derived from the host's engulfing vesicle), and their own bacterial-type 70S ribosomes rather than the eukaryotic cytoplasm's 80S ribosomes. Chloroplasts show the same pattern with cyanobacterial ancestry. This concordant, multi-line evidence — genome phylogeny, membrane topology, ribosome type, and even antibiotic sensitivity (mitochondrial translation is blocked by bacterial-targeting antibiotics like chloramphenicol but not by eukaryotic-targeting cycloheximide) — converged on a single conclusion that no single line of evidence could have established alone.

Experimental evolution of multicellularity — the Ratcliff experiment. In 2012, William Ratcliff, Michael Travisano, and colleagues took single-celled Saccharomyces cerevisiae yeast and imposed an artificial selection regime that favored settling quickly through liquid medium: after growth, they took only the fastest-settling fraction of cells and used it to found the next culture, repeating this for tens of generations. Within about 60 rounds of selection, yeast lineages evolved multicellular "snowflake" clusters — branched aggregates of cells that failed to separate after budding and instead stayed physically attached, settling faster than single cells because of their larger size. Critically, these clusters showed genuine differentiation: some cells within a cluster underwent apoptosis (programmed cell death) that created a physical fracture point, allowing propagule-sized clusters to break off and disperse — a rudimentary germ-soma division of labor, evolved from scratch in under a year of laboratory time. This experiment demonstrated that a transition to multicellularity does not require novel genes; it can arise from simple changes in cell-separation behavior under selection that favors group-level traits (settling speed) over cell-level traits, and division of labor can follow quickly once groups exist.

Worker sterility and kin selection — Hamilton and the social insects. W. D. Hamilton's 1964 papers on inclusive fitness explained why sterile worker castes — the paradigmatic case of individuals giving up their own reproduction for the reproductive success of relatives — could evolve by natural selection at all. Hamilton noted that in Hymenoptera (ants, bees, wasps), haplodiploid sex determination means sisters share, on average, three-quarters of their genes (they inherit an identical paternal haploid genome set and share half of their mother's genes on average), higher than the one-half a female would share with her own daughter. This asymmetry means a worker can propagate more of her genes by helping her mother produce more sisters than by reproducing herself, provided she is helping raise full sisters. This prediction — that eusociality should be disproportionately common in haplodiploid lineages — matches the observation that Hymenoptera account for the large majority of independent origins of eusociality (estimated at roughly 8–11 independent origins in that order alone), although eusociality also evolved in diplodiploid lineages (naked mole rats, termites), showing haplodiploidy helps but is not strictly required. Later work (Bourke, Ratnieks) documented worker policing directly: in honeybee colonies, workers detect and destroy eggs laid by other workers (because those eggs are less related to the policing worker than the queen's eggs are), and in species with effective policing, worker reproduction is measurably suppressed — a mechanistic demonstration of conflict suppression stabilizing a transition.

Example 1 — Predicting the direction of a transition from relatedness asymmetry. Consider a hypothetical haplodiploid wasp species in which queens mate with a single male (monandry) versus a related species where queens mate with four males equally (polyandry, effective mate number = 4). Under monandry, a worker's relatedness to a full sister is 0.75, and to a brother is 0.25 (haplodiploid asymmetry: sisters share the father's whole haploid genome plus half of the mother's genes on average, giving r = 0.5×1 + 0.5×0.5 = 0.75; brothers get none of the father's genes, so r = 0.5×0.5 = 0.25). A worker's relatedness to her own potential son (if she laid an unfertilized egg) would be 0.5. Using Hamilton's rule, rearing a full sister (r = 0.75) is favored over personal reproduction of a son (r = 0.5) whenever the costs and benefits are roughly comparable, strongly favoring altruistic helping. Now compute the polyandrous case: with four equally represented fathers, the average relatedness to a random female nestmate becomes r = 0.5×0.25 + 0.5×0.5 = 0.375 (only a 0.25 chance a random sister shares the same father, contributing 1 to paternal relatedness, averaged across the four possible fathers, plus the constant 0.5 maternal share). This average relatedness (0.375) is now below a worker's relatedness to her own son (0.5). Hamilton's rule predicts that high polyandry should erode the genetic benefit of helping and select for reduced worker altruism or increased worker reproduction — a real, tested prediction (Hughes et al. 2008, Science, found that a single mating by the founding queen — low paternity skew, i.e., high relatedness — was the ancestral state across origins of eusociality in bees, wasps, and ants, consistent with monandry being required to get a transition to eusociality started, even though many derived lineages later evolved polyandry after workers had already become obligately sterile).

Example 2 — Reading the mitochondrial genome as evidence of reductive evolution. A student is given two genome sizes: a free-living alphaproteobacterium, Rickettsia prowazekii (obligate intracellular but not organelle-derived, ~1.1 million base pairs, ~834 protein-coding genes), and the human mitochondrial genome (16,569 base pairs, 13 protein-coding genes). The student is asked to explain the ~85-fold reduction in genome size using the transition framework rather than simply citing "genome reduction." Reasoning: once the proto-mitochondrion became permanently housed inside a host cell with reliable nutrient supply and no need for independent motility, cell-wall synthesis, or defense genes, most of its metabolic and structural gene repertoire became redundant, especially once functional copies were transferred to the host nuclear genome (a process called endosymbiotic gene transfer, documented directly by finding mitochondrion-derived gene sequences embedded in the nuclear genomes of plants and animals, some still bearing mitochondrial-type codon usage). Because the host nucleus now controls replication, transcription, and most metabolic genes needed by the organelle, the mitochondrion's remaining 13 genes are almost entirely core, hydrophobic subunits of the oxidative phosphorylation complexes — proteins whose extreme hydrophobicity makes them difficult to import across two membranes from the cytoplasm, which is one leading hypothesis (the hydrophobicity/CoRR hypothesis) for why these particular 13 genes, and not others, were retained rather than transferred. This is a textbook instance of Step 2 (division of labor) and Step 3 (transmission bottleneck) working together: the organelle became an obligate specialist precisely because its remaining genome is now inseparable from, and totally dependent on, the host's nuclear genome.

John Maynard Smith trained first as an aeronautical engineer before turning to genetics under J. B. S. Haldane at University College London, a background that shaped his lifelong habit of thinking about biology through explicit mathematical models rather than narrative alone. He became one of the founders of evolutionary game theory, introducing the concept of the evolutionarily stable strategy (with George Price) in 1973 to explain why animal contests rarely escalate to lethal violence. Late in his career, in collaboration with the Hungarian biologist Eörs Szathmáry, he turned to a much larger question: not why any single trait evolves, but why the entire architecture of life is organized into nested levels — genes in chromosomes, chromosomes in cells, cells in organisms. Their 1995 book synthesized molecular biology, ecology, and game theory into a single explanatory scheme, and it reframed the history of life as a sequence of solvable problems in cooperation rather than a list of unconnected "origin of X" stories.

The transitions framework is not only historical bookkeeping; it actively guides research and applied biology. In cancer biology, tumors are increasingly understood as a reversion of the multicellularity transition — cell lineages that escape the suppression mechanisms (apoptosis, immune surveillance, growth-signal dependence) that normally keep somatic cells cooperating, and revert to cell-level selection for proliferation, sometimes described using the same multilevel-selection mathematics developed for transitions in general. In synthetic biology and biotechnology, researchers deliberately engineer division of labor between microbial strains in consortia — for example splitting a costly biosynthetic pathway across two or more bacterial strains so each carries only part of the metabolic burden — mirroring the logic of Step 2 above and testing, in real time, how conflict (one strain evolving to exploit the other's product) must be managed to keep engineered consortia stable. In agriculture, understanding cytoplasmic male sterility — a case of mitochondrial genes favoring female-only reproduction against the nuclear genome's interest in producing both sexes — has been directly exploited to produce hybrid seed in crops like maize and sunflower, turning an instance of unresolved transition-level conflict into a practical breeding tool. In conservation biology, the study of eusocial insect colonies as "superorganisms" informs how population viability should be modeled for species where the true demographic unit is the colony, not the individual — a queen bee's death is a very different event, demographically, than a single worker's death. And the framework's emphasis on relatedness, conflict, and transmission bottlenecks continues to guide mitochondrial replacement therapy research (so-called "three-parent" IVF techniques for preventing mitochondrial disease), where clinicians must reason carefully about heteroplasmy, bottleneck size, and the risk of reintroducing pathogenic mitochondrial variants — the same population-genetic logic that explains why the mitochondrial bottleneck evolved in the first place.

Ask the tutor: How does the mitochondrial bottleneck's population size affect the efficiency of purifying selection against harmful mtDNA mutations?