MuLTEE
History of the experiment
One · 2010 to 2014
Building the model system
Snowflake yeast were made in Michael Travisano's lab at the University of Minnesota, where Will Ratcliff was a postdoc. Out of a conversation about experiments that could address the major evolutionary transitions, the two designed a simple experiment around a new selection regime: let yeast cells settle through the liquid culture, transfer the bottom fraction into the next day's fresh tubes, and see whether yeast could evolve multicellular clusters. The design was remarkably effective. Ten replicate populations of unicellular Saccharomyces cerevisiae Y55 were grown in liquid and settled daily, and all ten evolved clonal snowflake clusters within just 60 transfers, about 300 generations. Clusters form because mother and daughter cells fail to separate after budding, so every group is clonal by construction. Under settling selection the multicellular genotypes carried a 34% fitness advantage. Without it they paid about 10%.
Two results from this period were important for the work that followed. The first was genetic. RNA-seq identified 1,035 differentially expressed genes, and seven of the ten most strongly downregulated were targets of the transcription factor ACE2. Knockout and complementation confirmed that loss of ACE2 is both necessary and sufficient for the snowflake phenotype. Because a single gene deletion produces the multicellular phenotype, later experiments could begin from a constructed ace2Δ ancestor rather than waiting for the mutation to arise.
The second was methodological. Because larger clusters sink faster, settling selection screens enormous numbers of clusters for size every day without needing a predator or any other ecological manipulation, and the strength of selection can be adjusted simply by shortening the settling window. That combination of scale and control is what made snowflake yeast a practical system for evolving and studying increased multicellular size.
Plate 01
- Imaging
- Differential interference contrast
- Subject
- Snowflake yeast cluster
A snowflake yeast cluster imaged in transmitted light, with no fluorescent label. Cells remain attached to the mothers that budded them, so the branched form is a consequence of how the group grows. The archive records no treatment or timepoint for this image.
Ratcliff moved to Georgia Tech in 2014 and continued this work there. Over the next few years the lab studied how high rates of cell death promote the formation of multicellular aggregates, how aggregative and clonal multicellularity compete against one another, and the biophysics behind the first increases in cluster size. This work was done through short-term evolution experiments and modelling. There was no long-term evolution experiment, since the multicellular phenotype had shown no striking change in the earlier work.
Two · 2018
Testing whether oxygen limits multicellular size
In these early experiments, the model system appeared to have an upper limit on size. Size increase flattened after a few hundred generations, and the longest run in the lab, a mixotrophic population carried to roughly 400 days, remained microscopic. The interpretation available at the time was that snowflake yeast could evolve into multicellular clusters but could not evolve much beyond that, and that the system was therefore not well suited to studying open-ended multicellular evolution.
Ozan Bozdag joined the Ratcliff lab as a postdoc and proposed a new project exploring the relationship between multicellular size, oxygen concentration, and carbon metabolism, building on his doctoral work growing yeast under different metabolic regimes. In an organism that respires, oxygen reaches interior cells by diffusion, and diffusion limits how large a mass of respiring cells can become. If oxygen delivery was what constrained cluster size, then the limit was a consequence of the culture conditions rather than a property of snowflake yeast, and changing those conditions should remove it.
This can be tested directly, because metabolism in yeast can be manipulated. Respiration-deficient petite mutants cannot use oxygen at all, so increased cluster size carries no cost in oxygen delivery. Sparging air through the medium raises dissolved oxygen for populations that do respire.
Bozdag's postdoc project comprised twenty populations across four metabolic and oxygen regimes: anaerobic petites, strictly aerobic yeast at intermediate oxygen (about 24–25% of present atmospheric level), strictly aerobic yeast at high oxygen (about 72–75% PAL), and mixotrophic yeast at intermediate oxygen. All twenty were maintained under identical daily growth and settling selection for 145 transfers, roughly 812 generations.
Oxygen strongly affected size evolution, and the effect was not monotonic. Populations at intermediate oxygen increased in mean size by only 8.9%, while anaerobic populations increased by 93% and high-oxygen populations by 97%. The size limit that had made the system appear closed was therefore set by the metabolic regime rather than by the yeast, and it could be removed.
Bozdag, Libby, Pineau, Reinhard & Ratcliff 2021 · Nature Communications 12:2838
Three · 2018 onward
The decision to keep transferring
At 145 transfers the experiment had answered the question it was designed to ask, and the result was being prepared for publication in Nature Communications. Bozdag did not stop the experiment. He continued transferring the populations daily, past 300 transfers and then past 400. That decision is what turned a postdoc project into a long-term evolution experiment.
There was no signal in the data at the time to justify continuing. Between roughly day 150 and day 400 the anaerobic populations did not increase in size. The motivation for continuing came instead from Richard Lenski's long-term evolution experiment with E. coli, which has been running since 1988 and has produced many of its most important results decades after it began, from populations maintained through long periods in which little appeared to change. A plateau in a selection experiment does not necessarily mean that adaptation has stopped.
Past day 300 the anaerobic populations began increasing in size again, and continued to do so. Within a few hundred further transfers the clusters were large enough to see without a microscope. Bozdag showed the macroscopic lines to Ratcliff, and the two agreed that the experiment should be maintained and developed as a long-term evolution experiment.
Fifteen of the original twenty populations are the ones still running: five anaerobic petite (PA1–5), five mixotrophic (PM1–5) and five obligately aerobic (PO1–5), all founded in 2018 from a single isogenic ace2Δ Y55 ancestor, so the starting genotype is known exactly. These populations were not re-founded after the oxygen result was published. They are the original populations, transferred continuously, which is why the experiment has an unbroken time series running from its first transfer to the present.
By 600 transfers, about 3,000 generations, all five anaerobic populations had become macroscopic. Mean cluster radius increased from 16 µm to 434 µm, roughly a 20,000-fold increase in volume, and biophysical toughness increased about 10,000-fold. These results were published in Nature in 2023, with Bozdag as first author and Ratcliff as senior author. The experiment was named the MuLTEE during the preparation of that paper.
Bozdag, Zamani-Dahaj, Day et al. 2023 · Nature 617:747–754
The experiment has since become the central project of the Ratcliff Lab and the basis of a decadal research program. Bozdag is principal investigator on the NSF LTREB award that funds it and Ratcliff is co-principal investigator. Bozdag continues to run the daily transfers, with Dung T. Lac managing the frozen archive and strain distribution. The populations are past 9,000 generations and the archive holds more than 3,000 samples.
A note on this history
The MuLTEE was not planned as a long-term experiment. It began in 2018 as a test of whether oxygen availability limits multicellular size, and it became a long-term experiment because the populations were kept running after that test was complete. It was named in 2023, five years after its first transfer. We describe it this way because the sequence of decisions is part of the record of how the experiment came about.
Dated record
Timeline
Day 400. Cell walls in cyan, nuclei in orange. The yellow magnification label and the 20 µm scale bar are the microscope's own annotation and are left in place.
-
2012
Multicellularity evolves de novo
Ratcliff, then a postdoc with Michael Travisano at the University of Minnesota, grew ten replicate populations of unicellular S. cerevisiae Y55 under daily settling selection, and all ten evolved the clonal snowflake phenotype within 60 transfers, about 300 generations. Multicellular genotypes held a 34% advantage under settling selection (P = 0.004). By transfer 60, apoptosis rate and settling rate were tightly correlated (r² = 0.91). Ratcliff, Denison, Borrello & Travisano, PNAS 109:1595–1600.
-
2013
A second organism, and the first tempo-and-mode study
Ratcliff and colleagues evolved multicellularity de novo in Chlamydomonas reinhardtii, a green alga with no multicellular ancestor, in 219 days, with a facultative life cycle and an evolved unicellular bottleneck (Nature Communications). In the same year Ratcliff, Pentz & Travisano tracked 227 days of post-origin snowflake adaptation and found three ordered modes: more cells per cluster (days 7–28), a 2.21-fold increase in per-cell volume (days 28–65), then more hydrodynamic cluster shape (days 65–227).
-
2014
Theory, synthesis, a classroom kit, and a move to Georgia Tech
Libby, Ratcliff, Travisano & Kerr (PLoS Computational Biology) formalized clusters as tree graphs and explained why that geometry favours apoptosis, predicting an evolved rate near 10−2. Libby & Ratcliff (Science) named ratcheting as the mechanism that stabilizes transitions. Ratcliff & Travisano (BioScience) synthesized the program. The 2012 experiment became a three-week undergraduate lab, distributed free of charge (The American Biology Teacher). Ratcliff moved to Georgia Tech the same year.
-
2015
ACE2 identified, heritability measured
Ratcliff, Fankhauser, Rogers, Greig & Travisano showed by knockout and complementation that loss of ACE2 is necessary and sufficient for the snowflake phenotype, with non-synonymous ACE2 mutations in five of ten independently evolved lineages. Broad-sense heritability of cluster size at reproduction was 0.84, and a Pascal's-triangle model of the body plan fit strains from 7 to 227 days of evolution (r² = 0.94). Nature Communications 6:6102. Because a single gene deletion produces the phenotype, every MuLTEE population was later founded from a constructed ace2Δ ancestor.
-
2017
First NSF award for multicellular complexity
NSF IOS-1656549, “Origin of multicellular complexity in experimentally-evolved Saccharomyces cerevisiae,” Ratcliff PI with Yunker as co-PI, $638,834, February 2017 to January 2020.
-
2018
Reproduction turns out to be fracture
Jacobeen et al. (Nature Physics) showed that snowflake yeast reproduce when growth-induced internal stress breaks the cluster: total volume contracts to 94% at fracture. Over seven weeks, about 291 generations, cells evolved an 8% larger aspect ratio while cluster volume fraction fell from 0.32 ± 0.04 to 0.22 ± 0.03, giving a 1.7-fold size increase with no change in the chitin bonds themselves.
-
2018
The oxygen experiment begins
Bozdag founded twenty populations from a single isogenic ace2Δ Y55 ancestor to test whether oxygen availability, not the organism, was what capped multicellular size: anaerobic petites, mixotrophic yeast, and strictly aerobic yeast at intermediate and at high oxygen, all under identical daily growth-and-settling selection. Fifteen of these populations, the anaerobic, mixotrophic and obligately aerobic lines, are the fifteen still being transferred today.
-
2018–2020
The experiment is continued past its endpoint
The designed run ended at 145 transfers. Bozdag continued daily transfers past 300 and 400, through a stretch of roughly 250 days in which anaerobic cluster size did not increase at all, on the reasoning behind Lenski's E. coli experiment running since 1988: a plateau in a selection experiment does not necessarily mean adaptation has stopped. Past day 300 the anaerobic populations resumed increasing in size and became macroscopic. Bozdag and Ratcliff then agreed the experiment should be maintained and developed as a long-term evolution experiment.
-
2019
NSF CAREER award
IOS-1845363, “Examining the Role of Nascent Multicellular Life Cycles on the Evolution of Organismal Complexity,” Ratcliff sole PI, $1,136,348, May 2019 to April 2025.
-
2020
NIH R35 MIRA awarded
NIGMS R35 to Ratcliff as sole PI, August 2020 to May 2024. The 2024 renewal was organized around three MuLTEE priorities: cellular differentiation, synchronized cell division, and whole-genome duplication.
-
2021
Oxygen suppression of macroscopic multicellularity
The first 145 transfers published: 20 populations, about 812 generations, four metabolic and oxygen regimes. Intermediate oxygen permitted only an 8.9% size increase, while anaerobic and high-oxygen populations grew 93% and 97% larger. The relationship between oxygen and multicellular size is not monotonic. Bozdag, Libby, Pineau, Reinhard & Ratcliff, Nature Communications 12:2838, with Bozdag as first author.
-
2023
Macroscopic multicellularity, and the experiment gets its name
After 600 rounds of selection, about 3,000 generations, all five anaerobic populations had become macroscopic: mean cluster radius rose from 16 µm to 434 µm, roughly a 20,000-fold increase in volume, with toughness up about 10,000-fold. Bozdag, Zamani-Dahaj, Day et al., Nature 617:747–754, published online 10 May 2023, with Bozdag as first author and Ratcliff as senior author. The experiment was named the MuLTEE during preparation of this paper, five years after its first transfer. The same year it ran as a three-quarter-page colour photo feature in The New York Times.
-
2024
Four mechanism-and-consequence papers
Montrose et al. (Science Advances) traced cell elongation to reduced Hsp90 expression. Pineau et al. (Nature Ecology & Evolution) reported small and large lineages coexisting for about 4,300 generations. Day et al. (Physical Review X 14:011008) generalized entanglement to any growing branched living matter. Bingham & Ratcliff (PNAS) showed that multicellularity lowers effective population size.
-
2024
LTREB proposal for the 2025–2035 decade
Submitted in September 2024 with Bozdag as PI and Ratcliff as co-PI, requesting support to maintain the evolving populations, the frozen archive, systematic phenotyping and genomics, and an open-access portal. The experiment's current long-term support is NSF award 2452109.
-
2025
Genome duplication, and a stocktaking of long-term experiments
Tong et al. (Nature 639:691–699) reported convergent tetraploidy arising within the first 50 days and persisting for the next 950, close to 5,000 generations, across ten replicate populations. Stroud & Ratcliff (Nature 639:589–601) placed the MuLTEE among the field's long-running systems. Also 2025: Narayanasamy et al. on metabolically driven flows, Cedeño-Pérez et al. on coordinated cell division, Wong et al. (PLOS Biology) on oxygen-binding proteins, and Yoon et al. (Nature Communications) on super-resolution imaging.
-
Present
Past 9,000 generations, and still transferring
The archive holds more than 3,000 frozen samples across the 15 lines. Populations are cryopreserved every 25 days, roughly every 125 generations. Bozdag continues to run the daily transfers, with Dung T. Lac managing the archive and strain distribution. The characterized leading edge of the experiment is day 1,000, the timepoint sequenced and phenotyped in Tong et al. 2025. The living populations are well beyond it.
Publications
The eight papers that use MuLTEE populations, plus the snowflake yeast work behind them.
15 lines · 3 treatmentsPopulations
What each of the fifteen lines has done since it was founded in 2018.
3,000+ frozen samplesData & Strains
Request any population from any timepoint in the archive, plus genomes and protocols.