175 | William Ratcliff on Multicellularity, Physics, and Evolution
What this covers
Sean Carroll interviews William Ratcliff, a biologist who studies how single cells evolve into multicellular organisms. Rather than treating multicellularity as a rare, contingent event shrouded in mystery, Ratcliff presents it as a repeatable physical transition that can be engineered and observed in real time. The conversation builds from lab experiments with snowflake yeast—a strain that evolved multicellular clusters under simple size selection in just thousands of generations—to broader questions about whether size or complexity comes first in evolution, and whether oxygen enabled or constrained the rise of multicellular life.
The discussion spans several distinct claims. Ratcliff argues that the Darwinian Algorithm operates at the group level once cells remain attached after division, making multicellularity evolutionarily easy and nearly inevitable. He also presents a counterintuitive finding about oxygen: rather than monotonically enabling larger organisms, it shows a "Nike swoosh" pattern in his experiments, suppressing size at intermediate oxygen levels and potentially constraining rather than catalyzing the Great Oxidation Event. The conversation then pivots to materials physics—how cells evolved entanglement) and cell-packing strain to increase structural toughness by orders of magnitude, and how physical fracture mechanisms create the genetic bottlenecks necessary for true group-level selection. Throughout, Ratcliff emphasizes that size precedes and drives complexity: once organisms become large enough, diffusion limits and internal-external chemistry gradients create selective pressure for circulatory systems and cell communication.
Ratcliff argues that multicellularity is a repeatable, physically-constrained evolutionary transition that can be reproduced in the lab in real time, and that increased size (driven by factors like oxygen and cell-packing physics) precedes and enables biological complexity rather than the reverse.
- Snowflake yeast evolve multicellularity rapidly under simple size selection, showing the transition is easy and Darwinian
- Oxygen's effect on multicellular size is non-monotonic (a 'Nike swoosh'), so the first Great Oxidation Event may have constrained rather than enabled complexity
- Physical mechanisms like cell-packing strain and entanglement, not just gene regulation, govern how multicellular bodies become tough and heritable
Oxygen availability sets hard limits on multicellular size through metabolic trade-offs between energy and resource access.
- Because a small amount of oxygen suppresses multicellular size, the first Great Oxidation Event (~2.1 billion years ago) likely acted as a constraint on multicellular complexity rather than a catalyst, consistent with the geological 'Boring Billion' of low apparent biospheric change, whereas only the later rise to near-modern oxygen levels ~500 million years ago enabled the flowering of multicellular diversity.
“at first, we probably would have been better off without the Great Oxidation Event”
- Over roughly 5000 generations (1000+ daily transfers) of settling selection, anaerobic snowflake yeast evolved to be macroscopic — visible to the naked eye and bigger than a fruit fly, about 20,000 times larger than their ancestor — while aerobic and mixotrophic yeast that can use oxygen stayed only a few times bigger, demonstrating both rapid evolvability and the size-suppressing effect of usable oxygen.
“we saw the evolution of snowflake yeast that are visible to the naked eye. So they're macroscopic. They're bigger than a fruit fly.”
- Oxygen confers two major metabolic benefits: respiration yields up to 18 times more energy from a resource than fermentation, and oxygen acts as a co-factor enabling organisms to metabolize certain carbon compounds (such as alcohol) that cannot be fermented but can be respired.
“oxygen provides two huge benefits, one, you can get a lot more energy out of the resource up to 18 times more energy”
- Contrary to the long-held assumption of a monotonic relationship where more oxygen yields larger organisms, evolution experiments with snowflake yeast show a 'Nike swoosh' pattern: yeast evolve large size with no oxygen (fermenters) and with abundant oxygen, but stay small under limiting oxygen, because a small amount of oxygen is a valuable resource that creates a selective incentive to stay small so a larger proportion of cells can access it.
“unlike the expectation that the more oxygen they have, the bigger they get, we find that when they have no oxygen, that is they're an anaerobe, or they have a low amount of oxygen, it doesn't matter if they're a mixotroph or an obligate aerobe or they have high amounts of oxygen, we end up getting this Nike swoosh, it's not a monotonic increase.”
Size itself drives complexity evolution by creating diffusion and mechanical pressures that select for specialized systems.
- Larger size is necessary for and drives the evolution of complexity: once an organism becomes big, diffusion limitations and inside/outside environments create evolutionary incentives to evolve circulatory systems, cell communication, and resource transport, so size came first and complexity followed rather than the reverse.
“size itself is a driver of increased complexity. Once you have a big organism, now you really have an incentive to evolve the circulatory system and mechanisms of cellular communication and resource transport within the organism. So science probably came first and complexity probably followed”
- Early multicellular organisms must confront mechanical stresses operating over length scales (tens to hundreds of microns) their unicellular ancestors never dealt with, so understanding multicellular evolution requires treating the body as a 'Darwinian material' that must become biophysically tough and robust, not just a question of gene regulation and differentiation; modern organisms achieve robustness through coordinated development, but nascent ones lack this coordination.
“forming a group of cells requires these simple multicellular organisms to confront stresses then act over length scales that they have no history of dealing with”
- A surprising departure from gradual size increase occurs when slightly longer cells begin packing more densely and entangle like vines wrapping around adjacent cells, with entanglement percolating throughout the cluster; this likely-phase-transition shifts groups from being a single fracture away from reproduction to requiring hundreds of broken bonds, raising material toughness from 100 times weaker than gelatin to as strong as wood and enabling clusters 1000 times larger than predecessors.
“We see this shift from cells being basically, branches on a tree that if you break a single cell, the entire branch falls away to cells being entangled like vines, such that the cells wrap around adjacent cells.”
- By transferring only 10% of the test tube each day, with the rest discarded, the protocol gives any cluster a 90% chance of dying before it can even compete to settle, which sets a maximum size beyond which dilution gets the better of growth; this enforced death is important because an organism that could grow forever would be biologically uninteresting and break the experimental system, which requires entities that both reproduce and die.
“we sub-sample 10% of it and those yeast compete to be transferred. So they actually have a 90% chance of dying without ever even having a chance to compete for faster settling”
Multicellularity evolved independently many times because different lineages gain distinct advantages through group formation.
- There is no single reason organisms become multicellular; different lineages gain different advantages, including avoiding predation by filter-feeders with small mouths, cooperating metabolically by collectively excreting enzymes to degrade environmental resources, and competing for directional resources like light by overgrowing competitors.
“every lineage has its own reasons to become multicellular, so there's no one reason to do it”
- Multicellularity has independently evolved many times across different lineages on Earth, probably around 100 times, with some lineages like green algae evolving it more than 25 times on their own.
“multicellularity actually has evolved many times in different lineages on earth”
- Snowflake yeast rarely appear in nature despite a fairly high mutation rate because being a cluster is ecologically costly in the yeast's natural habitat (rotting fruit, oak sap) where dispersal is at a premium; if a yeast's offspring stay stuck together and no fly transports them to a new sugar source, they die, making single-cell dispersal a winning strategy.
“that's ecologically costly for them in this environment. If you're making a living as a yeast and basically being moved from one sugar source to another, dispersal is a premium virtue.”
- Unlike animals or plants, fungi commonly revert from multicellular filamentous forms to single-celled yeast; the bakers' yeast Saccharomyces cerevisiae descended from a multicellular ancestor several hundred million years ago and has since jettisoned much of the genome that multicellular fungi possess, a pattern of devolution rarely seen in other multicellular lineages.
“our yeast descended from a multicellular ancestor, but that was several hundred million years ago”
Experimental evolution reveals multicellularity is rapid and easy to achieve, suggesting unicells sit near an evolutionary edge.
- In the lab, diluting yeast daily and selecting the cells that settle fastest to the bottom of a test tube incentivizes group formation because clusters settle faster than single cells; this is not claimed to be the natural driver of multicellularity but is an expedient proxy for selecting on size, which is the underlying advantage many lineages gain.
“this really incentivizes them to form groups, because groups settle through liquid media faster than single cells”
- Single-celled yeast can be turned into multicellular 'snowflake yeast' by breaking a single gene that normally allows daughter cells to separate from the mother cell after division, causing them to remain attached in fractal tree-like clusters; this ease of transition is probably true for most unicellular organisms, indicating they sit near the edge of multicellularity.
“you can just break the machinery that normally allows daughter cells to separate from a mother cell after they're done dividing, and then you just get these tree-like multicellular groups forming”
- The potential for traits to evolve is often effectively unlimited over geological time scales, with traits evolving very quickly given suitable selection; observed evolutionary stasis is usually not a constraint but simply the absence of significant selection on the organism's state.
“Things can evolve very, very, very quickly given suitable selection. And often when you see stasis, it's simply because you don't have any selection, any significant selection on the organism's state.”
- To compress evolutionary time, the lab uses synthetic biology to insert features that took millions of years to evolve — engineering yeast to express sperm whale myoglobin (making them pink), peanut worm myohemerythrin, and an ancestrally reconstructed myoglobin/hemoglobin ancestor — to test whether increasing oxygen diffusion through tissue synergistically removes size constraints and drives larger size, which it appears to do.
“we can use synthetic biology tools to bring in features that evolved over millions of years and see what they do in our system”
- Freezing evolving yeast populations at minus 80 degrees roughly every 50 generations creates a 'living fossil record' that lets researchers revive ancestors, compete them against descendants at any time point, and sequence all the mutations, enabling unprecedented mechanistic understanding of how the populations changed.
“we freeze our populations in a minus 80 deep freezer about every 50 generations, and so we can actually go back and pull out our living fossil record”
- Unlike most hypothesis-driven biology experiments that test a low-dimensional, additive effect requiring prior understanding of the system, experimental evolution is valuable precisely because you do not have to know in advance what you will get, allowing genuinely surprising discoveries that the experimenter could never have predicted.
“A lot of experiments in biology are hypothesis-driven, where you already have a pretty good idea of the way the system works before you even propose the experiment”
Eukaryotic cells possessed pre-existing regulatory tools and life-cycle complexity enabling far more sophisticated multicellular organization.
- Eukaryotic cells likely became more capable of complex multicellularity because they have a much larger repertoire of gene regulation mechanisms, larger size, more complex cellular morphology, and ancestral life cycles with multiple temporal cell states (amoeboid, flagellated) that can be repurposed from temporal expression to spatial expression, providing modules for building multicellular organisms.
“eukaryotic cells have a much larger number of types of gene regulation available to them, and it's easier to get cellular differentiation in eucaryotes”
- Eukaryotic cells arose from a fusion between an archaeal host and an alphaproteobacterium that became domesticated as the mitochondrion; evidence includes Asgard archaea (lokiarchaeota) that nest as a sister group to eukaryotes on phylogenetic trees and possess eukaryote-specific cell biology, plus the fact that the eukaryotic lipid membrane is bacterial rather than archaeal in chemistry.
“They're actually the result of a fusion between archaea and alphaproteobacteria, that alphaproteobacterium was at one point a Symbion and became domesticated as our mitochondrion”
- Once life achieves the cellular state, it has accomplished roughly 95% of the hard work of becoming complex life, because the transition to cellularity represents a sea change in how life evolves, with information transmitted vertically.
“once you have a cell, you've already sort of... You've done 95% of it”
Fracture-driven genetic bottlenecks in snowflake yeast enable group-level Darwinian selection to operate on emergent traits.
- In snowflake yeast, growth causes cell-packing strain that fractures branches, and because the cell at the base of a broken branch is the genetic progenitor of everything downstream, this physical fracture mechanism introduces genetic bottlenecks that segregate genetic variation between groups, creating a statistical correlation between group-level emergent traits and underlying cellular genotypes that allows group-level Darwinian selection to operate.
“fracture in this case is driven by cells dividing into the interior of the group, causing packing, strain accumulating from cell-cell packing, and when that strain accumulating from packing exceeds the strength of a cell-cell connection, you break a branch”
- Darwinian evolution is a substrate-independent algorithm requiring only that entities replicate, vary, that variation be at least somewhat heritable, and that persistence depend on those varying heritable traits; this applies to single cells, ants, dogs, chemistry, and computer programs alike, so for multicellular groups to evolve complexity they must acquire these Darwinian-entity properties at the group level.
“the cool thing about a Darwinian entity is that it's an algorithm, this applies to anything, living or non-living. You only need a few things to be true in order to have a population of things evolve via Darwinian evolution.”
- A multicellular group is not automatically a multicellular organism; per the philosophical literature, an organism requires interacting components exhibiting mutual adaptation and functionality, so a mere clump of cells (like a biofilm or naive snowflake) must evolve these integrated, functional features over generations to become a true organism.
“in order for something to be an organism, you need to sort of have... There's actually a philosophical literature on what counts as an organism, and it kinda boils down to the organism contains interacting components in which there's essentially mutual adaptation and essentially functionality”
Physical entanglement and cell elongation explain rapid material-property evolution independent of genetic differentiation mechanisms.
- Across all evolving populations, yeast first get bigger by evolving more elongated cells, which increases free space around each cell and reduces the jamming that occurs as cells divide into the cluster interior; like building a structure from hot dogs rather than ping-pong balls, the fluffier packing delays the strain-driven jamming and fracture, allowing larger sizes before the cluster breaks apart.
“initially, what all of our yeasts that are evolving to become macroscopic do is they make more and more elongate cells”
- Entanglement is an understudied but potentially widespread mechanism underlying the materials properties of biological structures, exemplified by Prototaxites (a Devonian fungus forming 25-foot-tall columns of beautifully entangled mycelium) and possibly rhino horn (matted, non-parallel hair), suggesting much of biological toughness arises from entanglement rather than only from glues or cell-cell bonds.
“if you look at the largest fungi that have ever existed, there's something called prototaxities... that formed these 3-foot... 3-foot wide, 25-foot tall columns of fungal Mycelium”
- Although snowflake yeast cells are genetically identical, they are not physiologically identical: over the experiment, older interior cells evolved to divide from the middle making right-angled branches that drive entanglement, while younger surface cells bud from the tip, a difference apparently coordinated by evolved changes in chaperone protein expression, which may represent the beginning of true age-dependent cellular differentiation.
“our yeast cells are all genetically the same, but that does not mean that they're physiologically the same. In fact, we see big differences, evolved differences... in the behavior of young cells that are sort of weighted towards the surface and older cells that are on the interior”
Early bacterial multicellularity emerged through metabolic cooperation and functional specialization among attached cells.
- Cyanobacteria began forming chains of cells around 2.1 billion years ago in which different cells specialized for distinct functions — carbon fixation via photosynthesis versus nitrogen fixation converting atmospheric nitrogen into ammonium for protein building — representing an early form of multicellular division of labor among bacteria.
“cyanobacteria began forming chains of cells in which you'd have cells that are specialized for carbon fixation, for photosynthesis or nitrogen fixation”
- Mitochondria retain their own small genome (~20,000 base pairs versus a few million for free-living bacteria) because most of their original genes have been exported into the host's chromosomes, with only genes that are hard to import through the membrane or that benefit from local production remaining behind.
“it's a 20000 DNA basis when a free-living bacteria is typically on the order of a few million”
- Because evolution proceeds by small steps, it generally cannot reach a more advantageous state if the only route there passes through a less-fit valley; this is why the anaerobic yeast, having lost nearly all their mitochondrial genome, will likely never re-evolve oxygen use, and why horizontal gene transfer and synthetic biology are powerful, allowing valley-crossing or moving machinery between disparate lineages.
“if there's a place you could imagine getting to that would be better for you, but the only route to get there is through a valley that is worse for you, evolution will, roughly speaking, never get you there.”
- Because cells take up space like sand grains, the distribution of cell-free volumes within multicellular clusters follows a universal k-gamma distribution arising from maximum entropy considerations with no free-parameter fitting; both the unevolved snowflake yeast and the 20-million-year-old Volvox fit this distribution well, implying a 'multicellular ground state' that coordinated development can only partly deviate from.
“it turns out that you get universal packing statistics for things where their particles take up space, and in fact, this arises from maximum entropy considerations and the distribution of cell-free volumes... is distributed with a K-Gamma distribution”
- Richard Lenski's long-term evolution experiment in E. coli, running for 30 years across 12 populations and approaching 75,000-80,000 generations, has taught fundamental lessons about evolution, most notably that adaptation never stops but follows a power law that does not asymptote, plus insights into novel gene function, diversified microbial strategies, and the roles of chance and necessity.
“We've learned that, basically, adaptation never stops. It's a power law. It doesn't asymptote.”