Emergence of coordinated cell division during the evolution of multicellularity

Cell division timing shapes the morphology and size of nascent multicellular organisms | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Cell division timing shapes the morphology and size of nascent multicellular organisms View ORCID Profile Luis F. Cedeño-Pérez , View ORCID Profile Rozenn M. Pineau , View ORCID Profile Thomas C. Day , View ORCID Profile William C. Ratcliff , View ORCID Profile Peter L. Conlin doi: https://doi.org/10.1101/2025.04.23.650085 Luis F. Cedeño-Pérez 1 School of Biological Sciences, Georgia Institute of Technology , Atlanta, GA, USA 2 Interdisciplinary Graduate Program in Quantitative Biosciences, Georgia Institute of Technology , Atlanta, GA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Luis F. Cedeño-Pérez Rozenn M. Pineau 3 University of Chicago , Chicago, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Rozenn M. Pineau Thomas C. Day 4 School of Biological Sciences, University of Southern California , Los Angeles, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Thomas C. Day William C. Ratcliff 1 School of Biological Sciences, Georgia Institute of Technology , Atlanta, GA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for William C. Ratcliff For correspondence: pconlin3{at}gatech.edu ratcliff{at}gatech.edu Peter L. Conlin 1 School of Biological Sciences, Georgia Institute of Technology , Atlanta, GA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Peter L. Conlin For correspondence: pconlin3{at}gatech.edu ratcliff{at}gatech.edu Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract Upon making the transition from unicellularity to multicellularity, many previously optimized cellular traits experience the renewed scrutiny of natural selection due to their novel effects on emergent multicellular phenotypes. Yet we lack a comprehensive understanding of how and why specific cellular traits influence multicellular phenotypes and fitness. The snowflake yeast model system provides a tractable entry point for such investigations. The effects of several cell-level traits (cellular aspect ratio, cell volume, bud neck strength) on multicellular cluster size have been characterized, but we found that these properties were insufficient to explain the difference in cluster size between the two strains that serve as the ancestors of the ongoing Multicellularity Long-Term Evolution Experiment (MuLTEE). Using time-lapse microscopy and single cell tracking, we identified the timing of cell division as a cellular trait that strongly influences multicellular morphology and size in snowflake yeast. The “petite” ancestor divides asynchronously, with a 25% longer first division, while the “grande” ancestor divided synchronously. Using network theoretical and biophysical models, we showed that strains exhibiting a first division delay generate more highly-branched network topologies, accelerating the accumulation of crowding-induced mechanical stress, resulting in clusters that fracture at smaller sizes. Conversely, synchronously dividing strains produce more symmetric, larger clusters. Synchronous cell division can provide benefits through both faster growth and larger size, suggesting multiple potential selective pathways for its evolution. Furthermore, we explore how accelerated first division can produce even larger groups and how another unexpected mechanism for modifying cluster size, apoptosis rate, may interfere with these effects. Our results identify cell division timing as a previously underappreciated axis of phenotypic variation that strongly influences multicellular morphology. This suggests that temporal regulation of cell division represents an evolutionarily accessible mechanism for early control of morphogenesis in nascent multicellular organisms with permanent intercellular bonds. Introduction The evolution of multicellularity represents one of life’s major evolutionary transitions, in which formerly free-living, lower-level units come together to form a new, higher-level entity ( Michod, 2000 ; Szathmáry & Smith, 1995 ; West et al., 2015 ). The first step in this transition is generally believed to be the formation of small, undifferentiated groups of cells ( Bourke, 2011 ; Herron et al., 2022 ). Once this new level of hierarchical organization has formed, if membership in the multicellular group affects a cell’s probability of reproduction or survival, the group’s existence has the potential to alter the relationship between cellular traits and fitness ( Maliet et al., 2015 ; Shelton & Michod, 2014 ). These shifts in cellular trait optima are ultimately responsible for the opening of previously inaccessible evolutionary paths to new multicellular phenotypes, new niches, and the adaptive radiations associated with the evolution of multicellularity in several clades ( Stoy et al., 2024 ). But what types of cellular traits have meaningful fitness consequences in the context of a small, undifferentiated group? How does more sophisticated control over group-level properties, like morphogenesis, get off the ground? Many aspects of multicellular morphology and behavior have been shown to arise from subtle changes in the underlying cellular traits. Properties such as cellular growth rate, adhesion, polarity, and division patterning can strongly influence group morphology, mechanical stability, and life cycles ( Duran-Nebreda & Solé, 2015 ; Staps et al., 2019 ). In the choanoflagellate Choanoeca flexa , for example, synchronized shape changes in individual cells drive collective sheet bending through direct microvilli tip adhesion ( Brunet et al., 2019 ; Fung et al., 2023 ). And changes in cell size and division patterning underlie transitions from unicellular ancestors to multicellular groups in the volvocine algae ( Featherston et al., 2018 ; Hanschen et al., 2016 ; Kirk, 2005 ; Maliet et al., 2015 ). In filamentous cyanobacteria, the spatial arrangement of dividing cells determines filament integrity and the emergence of specialized cell types ( Flores & Herrero, 2010 ; Herrero et al., 2016 ; Muñoz-García & Ares, 2016 ). And in Dictyostelium discoideum , which forms multicellular bodies via the aggregation rather than by clonal development, heritable variation in cell-fate bias and adhesion shapes the geometry and robustness of multicellular fruiting bodies, with direct consequences for group fitness ( Roisin-Bouffay et al., 2000 ; Siu et al., 2004 ; Strassmann & Queller, 2011 ). These examples highlight how small differences in cellular traits can produce large differences in group⍰level phenotype. Despite this broad recognition, the specific cellular traits that most strongly determine multicellular morphology can be difficult to identify, particularly in systems where multiple traits covary. The snowflake yeast system provides an ideal model for investigating how cellular traits and behaviors affect group-level properties. This multicellular organism arose through directed evolution ( Ratcliff et al., 2012 ), where daily selection for larger cluster size occurs through sedimentation, a process termed settling selection. The organism consists of clonal groups of cells that remain attached after division, forming three-dimensional clusters with a branched morphology. Unlike most extant multicellular organisms, snowflake yeast reproduces through group-level fragmentation driven by mechanical stress ( Jacobeen, Graba, et al., 2018 ), providing a simple and tractable system for studying the connection between cellular properties and group-level traits. Previous work has identified several determinants of cluster size at fracture, including cellular aspect ratio, cell-to-cell bond strength, and cell size ( Bozdag et al., 2023 ; Jacobeen, Graba, et al., 2018 ). Although we know that multicellular traits must arise from cellular modifications, we lack a systematic understanding of how specific changes to cellular behaviors reshape multicellular organization, especially during the critical early stages of multicellular evolution Here we used soft-agar time lapse microscopy and single-cell lineage tracing to collect detailed phenotypic measurements of individual cells during the growth and development of a multicellular snowflake yeast cluster. We applied these methods to two simple, undifferentiated multicellular strains and discovered a previously unappreciated difference in the synchrony/asynchrony of cell division between the two strains. Through computational simulations and network modeling, we demonstrate that asynchrony driven by a first division delay, but not asynchrony driven by increased variation in doubling time, results in altered cluster topology and smaller fragmentation sizes compared to synchronously dividing strains. We then demonstrate that the differences in morphology can have significant fitness consequences. We then confirm that clusters that began with asynchronous divisions rapidly revolved synchrony and maintained it for thousands of generations using strains from the ongoing Multicellularity Long-Term Evolution Experiment (MuLTEE) ( Bozdag et al., 2021 , 2023 ; Pineau et al. 2024 ; Tong et al 2025 ). Finally, we explore how different patterns of asynchrony affect group properties, revealing that when cells accelerate their first division, it generates even larger clusters than synchronous division. Together, these results illustrate how modifications to cellular behavior reshape network topology, generating heritable variation in multicellular phenotypes upon which selection can act. Results Motivating observation The two strains that function as the ancestors of the ongoing Multicellularity Long-Term Evolution Experiment (MuLTEE) genetically differ only in the presence or absence of a complete, functional mitochondrial genome ( Bozdag et al., 2021 ). Both strains, which we refer to respectively as “grande” (respiration-competent) and “petite” (respiration-deficient), are homozygous diploid derivatives of Saccharomyces cerevisiae strain Y55. Multicellularity was engineered by homozygous replacement of the ACE2 transcription factor with a kanamycin resistance marker ( ace2::KANMX/ace2::KANMX ), resulting in the formation of small, undifferentiated multicellular clusters. The petite strain was then generated by selecting for a spontaneous mutant incapable of respiration ( Figure 1A ). Download figure Open in new tab Figure 1. Cell properties cannot explain differences in cluster size between ancestral strains. A) Representative images of cluster sizes for the grande and petite ancestors. Scale bar, 100 μm. B) Cluster size distributions measured after 2 days of growth for both ancestral strains. Median cluster volumes: grande, 98,756 μm 3 ; petite, 55,187 μm 3 ( δ = 0.59, p < 2.2×10 −16 , Wilcoxon test). C) Representative images of cells from ancestral strains. Scale bar, 10 μm. D) Cell aspect ratio distributions. Median aspect ratios: grande, 1.19; petite, 1.21 ( δ = -0.14, p < 2.2×10 −16 , Wilcoxon test). E) Cell diameter distributions. Median diameters: grande, 5.07 μm; petite, 4.81 μm ( δ = 0.33, p < 2.2×10 −16 , Wilcoxon test). All data were collected from 10 technical replicates grown for 2 days with a 1:100 dilution after the first day. The grande strain produces significantly larger clusters than the petite strain ( Figure 1A,B ; δ = 0.59, p < 2.2×10 −16 , Wilcoxon test), yet they exhibit relatively minor and opposing differences in cellular aspect ratio ( Figure 1C,D ; δ = -0.14, p < 2.2×10 −16 , Wilcoxon test) and cell size ( Figure 1C,E ; δ = 0.33, p < 2.2×10 −16 , Wilcoxon test). Previous studies using snowflake yeast suggested that changes in maximum cluster size are largely explained by these two cell-level phenotypes; increases in either trait are expected to yield increases in maximum cluster size ( Bozdag et al., 2023 ; Jacobeen, Graba, et al., 2018 ). This indicates that our current understanding of the determinants of cluster size is incomplete and that additional factors must contribute to the observed size differences. To address this discrepancy, and to develop a comprehensive understanding of how cell-level properties influence group-level phenotypes in the snowflake yeast system, we devised a soft-agar time lapse microscopy method to collect data on the following parameters: cellular aspect ratio, cell size, budding angle, and cellular doubling time. To explore the possibility of differences in these phenotypes related to replicative age, we also employed lineage tracing using single-cell tracking to record mother-daughter relationships. Single-cell phenotyping by time lapse microscopy Initial characterization of the grande and petite strains by time lapse microscopy showed differences in cellular aspect ratio and cell diameter that were qualitatively similar to our prior static measurements. Cellular aspect ratio was 1.72% greater in petite cells ( Figure 1D , Figure S1A ; Median aspect ratio of petites = 1.18, grande = 1.16; δ = 0.01, p = 0.89, Wilcoxon test) and cell diameter was 20.12% greater in grande cells ( Figure 1E , Figure S1B ; Median cell diameter for grandes = 5.97, petite = 4.97; δ = 0.80, p = 1.25×10 −11 , Wilcoxon test). We also observed that petite cells bud at wider angles compared to grande cells (Median bud angle = 47.4º in petites vs. 35.2º in grande; δ = -0.11, p = 0.027, Wilcoxon text). A difference in the angle of attachment between mother and daughter cells could impact the packing density of cells in the cluster, making it a reasonable candidate trait for investigation. Unfortunately, because our observations were limited to only the first 2-3 cell divisions, we cannot conclude whether or not this difference would persist after further cell divisions have occurred. The petite strain has a slower overall growth rate than the grande strain (135 minutes vs. 105 minutes) and appears to spend longer in G1 phase ( Figure S1B ; δ = -0.63, p < 2.2×10 −16 , Wilcoxon text). Additionally, inspection of the cell division patterns of the petite and grande strains revealed another difference: the cells within clusters of the grande strain divide synchronously while petite cell divisions are more asynchronous ( Figure 2A, B ; Supplementary Videos 1, 2 ; Supplementary Table 1 ). By plotting the time at which cell divisions occurred relative to the time of the first observed cell division within each cell lineage, we saw periodic bursts of cell division through time for the grande strain, while the timing of cell divisions in the petite strain was more uniform ( Figure 2C ). Download figure Open in new tab Figure 2. Sources of cell division asynchrony in Ancestral strains of the MuLTEE. A-B ) Soft-agar time-lapse of the petite and the grande strains (scale bar: 10 µm). C) Histogram of division times normalized to the first division in each branch, demonstrating synchronous cell divisions in the grande strain and non-synchronous divisions in the petite strain. Histogram bin size = 10 minutes. D ) Doubling time distributions of the grande (n=203) and petite (n=191) ancestor strains. Petite strain exhibits greater variation in doubling times compared to the grande ancestor ( p < 2.2×10-16, Fligner-Killen test). Horizontal black line represents the median doubling time. E ) Scatter plots of cell doubling times for mother-daughter pairs for the grande (n = 53) and petite ancestor (n = 46). The petite ancestor shows asynchronous division with daughters dividing slower than mothers (p = 2.6×10-6, r = 0.69; Wilcoxon signed-rank test), the grande ancestor is dividing synchronously as mother and daughter take the same time to divide (p = 0.006, r = 0.4; Wilcoxon signed-rank test). Dashed diagonal line represent the identit line. F ) Doubling time distributions of the first and subsequent cell divisions. The petite ancestor exhibits a longer delay in first division (Median of 150 and 117.5 minutes for first and subsequent divisions, respectively; δ = 0.46, p = 3.7×10-7, Wilcoxon test) compared to the grande ancestor (median of 105 minutes for both first and subsequent divisions; δ = 0.02, p = 0.79; Wilcoxon test). Measurements included 147 first and 56 second divisions for the grande ancestor and 135 first and 56 second divisions for the petite ancestor. Horizontal black line represents the mean doubling time. G ) Heatmap showing how delay and doubling time variation independently contribute to the combined standard deviation of first and second division times. A 100% increase in delay produces the same increase in asynchrony as increasing the standard deviation from 0 to 30. Purple and yellow asterisks represent the delay and standard deviation of the grande and petite doubling time distributions, respectively. To explain the pattern of asynchronous cell division, we looked at differences in the underlying cellular doubling time distributions of grande and petite cells. Cellular doubling time was calculated as the interval between successive bud emergences, which corresponds to the period between consecutive S-phase initiations in yeast ( Hartwell, 1974 ; Johnston et al., 1977 ). We first determined whether the petite strain simply exhibited greater variation in cellular doubling time than the grande strain since random sampling from a broader distribution of doubling times could give the appearance of asynchronous division. As predicted, the standard deviation of cellular doubling times for the petite strain was significantly greater than that of the grande strain ( Figure 2D ; p < 2.2×10 −16 , Fligner-Killen test). Another possible explanation for asynchrony is that mother cells and their daughters may divide at different rates. To test this, we measured the similarity of cellular doubling times for mother-daughter pairs of cells for each strain (specifically, we calculated the median difference in doubling time of every daughter cell’s first division and that of their mother’s next division). We found that daughters divided 5 minutes slower than their mothers in the grande strain ( Figure 2E ; p = 0.006, r = 0.4, Wilcoxon signed-rank test), but petite daughter cells divided 35 minutes slower than their mothers, suggesting a seven-fold increase in the first division delay ( Figure 2E ; p = 2.6×10 −6 , r = 0.69, Wilcoxon signed-rank test). To expand our dataset, we compared the distribution of all first divisions (corresponding to the daughter above) vs. all subsequent divisions (corresponding to mother above) regardless of whether we had mother-daughter pairs. All divisions beyond the first are grouped together as “subsequent divisions” because only the first division follows a distinct distribution ( Figure S2A,B ). With this larger dataset, we observed a 32.5 minute difference for first and subsequent divisions of the petite strain ( Figure 2E ; δ = 0.46, p = 3.7×10 −7 , Wilcoxon test), consistent with the difference observed using mother-daughter pairs. In contrast, no difference was observed between first and subsequent divisions for the grande ancestor ( Figure 2F ; 0 minutes; δ = 0.02, p = 0.79, Wilcoxon test). In the discussion that follows, we will thus refer to the grande strain as “synchronous” and the petite as “asynchronous”. The pattern observed in the grande strain is consistent with prior work suggesting that yeast lacking ACE2 failed to exhibit the daughter-specific G 1 delay that is typically observed for S. cerevisiae ( Di Talia et al., 2009 ; Laabs et al., 2003 ). In wild-type cells, Ace2 is thought to play a central role in the regulation of cell size at the Start checkpoint by repressing expression of the G1 cyclin CLN3 . Deletion of ACE2 eliminates repression of CLN3, resulting in daughter cells that proceed through G1 at the same rate as mother cells. However, both the grande and petite strains of our multicellular yeast lack the ACE2 gene (as mentioned above, we deleted this gene to engineer the multicellular phenotype because incomplete cell separation is another consequence of ACE2 deletion) ( Oud et al., 2013 ; Ratcliff et al., 2015 ). Thus, some additional factor must be at play in the petite yeast to reintroduce the daughter-specific delay. For instance, delayed G1-to-S transition has been previously reported in mitochondrial DNA-deficient yeast ( Crider et al., 2012 ; Gorospe et al., 2023 ; Leite et al., 2023 ). Alternatively, it may be the case that the slower growing petite strain produces smaller daughter cells that need more time to grow before they reach the critical size necessary for Start ( Hartwell & Unger, 1977 ; Hatzis & Porro, 2006 ; Lord & Wheals, 1980 ; Thompson & Wheals, 1980 ). Further work will be required to determine the cause of the first division delay in our system. To visualize the joint effects of increasing doubling time distribution variation and first division delay on asynchrony, we created synthetic doubling time distributions drawn from log-normal distributions for the first and subsequent divisions. Log-normal distributions were chosen because they capture the skewed nature of cellular doubling time distributions, with shorter tails for fast divisions and longer tails for slow divisions ( Blank et al., 2018 ; Weber et al., 2014 ). In total, we simulated 441 combinations of variation and delay, sampling 10,000 data points per parameter combination. We found that increased asynchrony, defined here as the standard deviation of the overall doubling time distribution, can be caused by either a delay of the first cell division or by increased variance in doubling time ( Figure 2G ). Increasing standard deviation by 1 unit has the same effect on asynchrony as increasing delay by 3.3%. We also observed that the petite strain exhibits both higher variation and delay than the grande strain ( Figure 2G , yellow vs. purple asterisk). Thus, in the text that follows we will examine whether these factors, either individually or in combination, could be responsible for the observed disparity in cluster size of the petite and grande strains. Cellular behavior alters group morphology in a simple network model Two distinct features are apparently contributing to the asynchronous pattern of cell division that we observe in the petite yeast: higher variance in doubling time and a delayed first cell division. To determine whether either, or both, of these factors could be contributing to the unexplained discrepancy in cluster size that we observed between the petite and the grande strains, we developed a simple network model of snowflake yeast, where cells are represented as nodes and mother-daughter connections as edges. Each cluster begins as a single cell and all daughters remain attached to their mothers after division. To enable first division delay and doubling time variation, cells reproduce according to user-defined “first division” and “subsequent division” doubling time distributions that can exhibit differences in mean and standard deviation. To isolate the effects of cell division timing on group structure, we ignore the potential impacts of nutrient diffusion ( Bozdag et al., 2021 ; Lavrentovich et al., 2013 ; Wong et al., 2025 ) or growth limitations that may arise due to crowding ( Libby et al., 2014 ; Nanda et al., 2024 ). With this simple model of cluster growth, we can study a number of topological properties relevant to fitness in the snowflake yeast system. First, the diameter of a cluster can be approximated using the “network diameter”, defined as the maximum of shortest paths measured for all possible pairs of nodes in the network. Furthermore, we can obtain a proxy for the local packing density of cells (and the associated physical strain on any given cell-cell connection) by calculating the “edge degree” for each edge in the network as the sum of other edges emanating from the two connected nodes, excluding their mutual connection. As we will detail in the next section, the accumulation of physical strain plays a critical role in determining the timing and location of fragmentation in snowflake yeast ( Jacobeen, Pentz, et al., 2018 ). Using three distinct doubling time distributions, we simulated network growth from a single cell to 1300 nodes, which represents the maximum number of cells measured in a grande ancestor cluster ( Figure S3 ). These distributions included: the empirical petite ancestor data (high variation, first division delay), a synthetic distribution matching the petite ancestor but without delay (high variation, no delay), and the grande ancestor (low variation, no delay) ( Figure 3A ). Network diameter was significantly smaller only for networks generated using the petite strain’s doubling time distribution ( Figure 3B ; for network size of 100: p < 2.2×10 −16 for petite vs grande and petite vs petite w/o delay, and p = 0.07 for petite w/o delay vs grande, Wilcoxon test). Similarly, we observe a significantly higher maximum edge degree in networks generated using the petite strain’s doubling time distribution ( Figure 3C ; for network size of 100: p < 2.2×10 −16 for petite vs grande and petite vs petite w/o delay, and p = 0.24 for petite w/o delay vs grande, Wilcoxon test). This suggests that the delayed first division can lead to more highly branched networks with shorter diameters. Higher variance in doubling time alone (petite without delay distribution), however, does not produce a shift in the mean of either phenotype. Download figure Open in new tab Figure 3. Cell behavior affects group morphology by changing network topology. A ) Doubling time distributions used in simulations: petite ancestor (high variation with first division delay), modified petite ancestor (high variation, no delay), and grande ancestor (low variation, no delay). B ) Mean network diameter by network size. Inset: Diameter-defining paths (red nodes and edges). C ) Mean maximum edge degree by network size. Edge degree is defined as the sum of node degrees for both nodes forming an edge minus 2 (excluding their mutual connection). Inset: Maximum edge degree connections (edge of interest in red, contributing edges in orange). D ) Mean number of mothers with ≥2 unbudded daughter cells as a function of network size. Inset: Motifs of mother cells with ≥2 unbudded daughters (mothers in red, daughters in orange). For panels B-D : Each trajectory shows the average of 300 independent simulations with network properties measured after every 5 nodes added. Square markers indicate network sizes shown in insets. Example networks shown have diameters of 12 and 13 (petite vs. grande, panel B ), edge degrees of 14 and 12 (panel C ), and 13 and 3 motifs (panel D ), respectively. By comparing network topologies, we also observed that mother cells with two or more unbudded daughter cells were significantly more common in the petite networks compared to both the grande strain and the petite without delay ( Figure 3D , red nodes connected to orange; δ = 0.93 vs.grande, δ = 0.78 vs. petite w/o delay, p < 2.2×10 −16 for both comparisons, Wilcoxon test). These “unbudded daughter” motifs, which arise when a mother divides faster than its daughter, are of interest because they typify the phenotypic consequences of asynchrony: they produce an increase in edge degree without increasing the overall network diameter. Grande networks also produced fewer unbudded daughter motifs than the petite without delay condition ( Figure 3D ; δ = 0.46, p < 2.2×10 −16 , Wilcoxon test), suggesting that both first division delay and higher variation in doubling time contribute to the appearance of these motifs. However, because the asynchrony of the petite without delay condition is essentially unbiased, these unbudded daughter motifs are offset by an equal number of “filamentous” motifs where daughters divide faster than their mothers (explored further in Figure 8 ). Linking morphological changes to multicellular life cycle consequences To understand how these morphological differences translate into consequences for the snowflake yeast life cycle, we extended our network model to capture both the growth and reproduction of clusters. In snowflake yeast, cluster reproduction occurs via fragmentation into two independently viable clusters when a mother-daughter connection (i.e., an edge in the language of our network model) is broken due to crowding-induced mechanical stress ( Jacobeen, Graba, et al., 2018 ). Fragmentation due to the accumulation of mechanical stress has been shown to be one of the primary barriers to evolving large group size. Accordingly, several key adaptations during the first 5,000 generations of the MuLTEE successfully prevent crowding (the evolution of increased cellular aspect ratio) or mitigate the effects of mechanical stress (the evolution of thicker, stronger cell-cell connections) ( Bozdag et al., 2023 ). A dynamic model of fragmentation requires the specification of both when and where fragmentation occurs. Biophysical modeling has been successful in predicting the timing of fragmentation, but not the spatial location ( Day et al., 2022 ; Jacobeen, Pentz, et al., 2018 ). While solving this prediction problem in an explicit biophysical model is outside of the scope of this work, we reasoned that we could use our edge degree metric to approximate the local accumulation of crowding-induced mechanical stress on cell-cell connections. Specifically, we assume that each edge in the network has a maximum edge degree threshold beyond which it can no longer maintain the cell-cell connection. Fragmentation therefore occurs when an edge has a degree that exceeds this threshold (timing) and the edge that is severed is the one that crossed the degree threshold (location). These simple assumptions localize fracture events to the most densely connected regions of the network and the size at fracture can be tuned by adjusting the edge degree threshold. To select the edge degree threshold for life cycle simulations, we calibrated our model using the grande ancestor’s fracture size (Edge degree = 15, Fracture size = 338; Figure S3 ). Using this dynamic network modeling framework, we simulated population growth across eight cluster generations for each strain type shown in Figure 3A , tracking key properties at each fragmentation event. First, we measured the average cluster size at fracture. Both synchronized strains (petite without delay and grande) achieved substantially larger fracture sizes than the petite ancestor ( Figure 4A ), with the petite without delay reaching a larger size than the grande ancestor (further explored in Figure 5 ). We then compared our results to the existing biophysical model where cells are represented as ellipsoids of defined diameter and aspect ratio and mechanical stress relate to the cumulative overlap of cells in space ( Bozdag et al., 2023 ; Day et al., 2022 ; Jacobeen, Graba, et al., 2018 ). To incorporate the differences in cell division timing, we adapted the model to add cells in the same temporal sequence as our network simulations, with fragmentation checks occurring only after all cells from each time point were added ( Figure S4A-C ). We again found that clusters generated using the doubling time distribution of the petite were smaller than those generated using the grande strain’s doubling time distribution (with petite clusters fracturing at 312.4 cells and grande clusters at 335.4 cells; Figure S4D ). The effect of doubling time delay is qualitatively similar to the effect observed in the dynamic network model, but smaller in magnitude (23 cells difference in the biophysical model compared to 121 cells in the network model). Furthermore, it provided independent support of the idea that asynchronous division accelerates mechanical stress accumulation through more localized crowding around older cells ( Figure S4E ). Download figure Open in new tab Figure 4. Delayed first division results in smaller, more branched multicellular groups at fracture and asynchronous cell division results in smaller propagules. A ) Cluster fracture size distributions for the three strains. Median sizes: petite = 218.3, petite without delay = 354.0, grande = 339.4. All means significantly differ (pairwise Wilcoxon tests with Bonferroni correction, p 1 indicate higher diameter than in an exponential network, <1 indicate lower diameter. Medians: petite = 0.912, petite without delay = 0.997, grande = 0.990. All means significantly differ ( p < 2e-16, Wilcoxon test with Bonferroni correction). C ) Propagule proportion after fracture (propagule nodes divided by nodes of the cluster before fracture). Medians: petite = 0.393, petite without delay = 0.393, grande = 0.453. Grande ancestor differs from both of the other strains ( p < 2.2×10 −16 , for grande vs petite and grande vs petite w/o delay, and petite vs petite w/o delay p = 1, Wilcoxon test with Bonferroni correction). For all these results 100 simulations were performed by allowing the clusters to grow up to 8 fragmentation events (8 cluster generations), and the plots show the mean value obtained from all the clusters in one simulation. Download figure Open in new tab Figure 5. Effects of delay and variation in doubling time distributions on cluster properties. Effects of 4 different levels of delay and variation on mean cluster fracture size ( A, B ), mean network normalized diameter ( C, D ), and mean propagule proportion ( E, F ). Left column highlights changes in first division delay and right column highlights changes in variation of the doubling time distribution. Each dot is the grand mean of 30 simulations. Each simulation consisted of 50 sequential cluster fragmentation events where fragmentation is followed by growth of either the propagule or parent at random. The difference in cluster size at fracture is driven by differences in network topology. Cells with multiple undivided daughters contribute disproportionately to reaching the edge degree threshold without increasing overall group size. To analyze the effects of network topology on fragmentation, we normalized network diameter to that of an exponentially growing network, enabling comparison across different cluster sizes ( Figure S5 ). This normalized metric reveals underlying structural patterns: values below 1 indicate networks enriched by undivided nodes (with degree 1), while values above 1 reflect configurations with fewer undivided nodes and more degree-2 nodes. The petite ancestor exhibited a normalized diameter below 1 ( Figure 4B ), consistent with its first division delay pattern. In these clusters, the delayed initial division followed by rapid subsequent divisions creates highly connected older cells. In contrast, strains without first division delay showed diameters closer to 1, indicating network topologies more closely resembling exponential growth patterns. All properties shown in Figure 4 remained constant through time in a modified simulation that tracked a single randomly-chosen cluster after fragmentation over 100 cluster generations ( Figure S6 ), further indicating that changes to the doubling time distribution result in characteristic phenotypic consequences. We also predicted that changes to the timing of cell divisions could affect the size of propagules produced by fragmentation. Propagule size has important direct and indirect fitness effects for nascent multicellular organisms ( Kondrashov, 1994 ; Libby et al., 2014 ; Pichugin & Traulsen, 2020 ; Roze & Michod, 2001 ). We define the propagule as the smaller of the two clusters resulting from a fragmentation event and “propagule proportion” as the ratio of cells in the propagule to the total number of cells in the pre-fragmentation cluster. Accordingly, this metric ranges from 0.5, for symmetric division, to near 0 when propagules contain just a few cells and, in our model, only an exponentially dividing network will achieve perfect symmetric division. Notably, the petite and petite without delay achieve similar propagule proportions while the grande ancestor achieves more symmetrical fragmentations ( Figure 4C , p < 2.2×10-16, for grande vs petite and grande vs petite w/o delay, Wilcoxon test), suggesting that high variation in doubling time distribution alone is sufficient to explain the decrease in propagule proportion for the petite strain. This differs from our earlier findings on network diameter and maximum edge degree where high variance in cellular doubling time did not contribute to shifts in the mean of cluster-level phenotypes. Propagule proportion is uniquely affected by increases in variation because any deviation from perfect exponential growth will disrupt the symmetry of fragmentation. To systematically dissect how division timing shapes network topology, we simulated cluster growth using synthetic doubling time distributions drawn from log-normal functions. We varied first division delay and temporal variation, sampling from a range of four distinct values for each parameter which had comparable effects on asynchrony ( Figure 2F ). Our analysis revealed that lower first division delay was associated with higher size at fracture and higher mean normalized diameter, while differences in doubling time variation had no significant effect on these properties ( Figure 5A-D ). The lower size at fracture of the grande ancestor compared to the petite without delay ( Figure 4A ) can be attributed to the small negative effect of decreasing variation when there is no first division delay ( Figure 5B ; delay = 0, light blue; grande variation = 15 vs. petite w/o delay variation = 30). Notably, propagule proportion responded to both timing parameters, with symmetric fragmentation emerging only when both variation and delay were minimal ( Figure 5E, F ). This explains why only the grande ancestor achieves near-symmetric division with propagule proportions approaching 0.5 ( Figure 4B ), as eliminating delay alone is insufficient to ensure symmetrical fragmentation. Partitioning the selective effects of first division delay Having established that first division delay significantly affects the size of clusters at fragmentation, we next investigated the impact this change could have on relative fitness under experimental conditions similar to those used in the MuLTEE. Each passage of the MuLTEE involves a 24-hour growth phase in liquid media followed by a brief phase of “settling selection” in which a subsample of the population is transferred to an Eppendorf tube where the culture is then vortexed and left to settle for 3 minutes before the top ∼95% of the volume is discarded ( Bozdag et al., 2023 ). Since the lack of a first division delay is associated with both faster individual growth and larger cluster size at fragmentation (compare petite vs. petite without delay, Figure 3A ), selection during either the growth phase (where fitness is largely determined by differences in cellular growth rate) or during settling selection (where fitness is largely determined by differences in cluster size) should favor the loss of first division delay ( Pentz et al., 2023 ). To quantify these effects and determine their relative contributions to fitness we simulated competition experiments using our dynamic network model with periodic settling selection based on simulated volumes from our biophysical model. To investigate the fitness effects of first division delay, we competed networks generated using the doubling time distribution of the petite ancestor strain vs. that of the petite without delay strain. Prior to competition, populations of each strain were grown independently until they reached stable cluster size distributions. Competition assays were then initiated by randomly sampling 100,000 cells from these stable distributions, maintaining an approximate 1:1 ratio of strains on a per cell basis. These cocultures were then passaged for 20 transfers or until we observed the extinction of one of the two strains. Each passage consisted of growth to a carrying capacity of 10 million cells (this was shown to be a sufficiently large population to accurately measure fitness; Figure S7 ) followed by two sequential 10-fold dilutions: the first to mimic the random subsampling of the total culture volume and the second to mimic the passaging of biomass that reached the bottom 5% of the tube after 3 minutes of settling selection ( Bozdag et al., 2023 ). Sedimentation rates were calculated using Stokes’ law for all clusters surviving the first 10-fold dilution ( Bozdag et al., 2023 ; Pineau et al., 2024 ; Ratcliff et al., 2013 ). Fitness during the growth and settling phases were estimated separately by measuring selection rates ( Lenski et al., 1991 ; Travisano & Lenski, 1996 ). Selection rates from the growth phase of the first transfer were excluded from this analysis, as the initial lower concentration of the fast-growing strain allowed for additional generations of growth compared to subsequent transfers ( Figure S8 ). Under conditions that closely mimicked those of the MuLTEE, we found that while the petite without delay strain was advantaged during both the growth and settling phases of the experiment, the selection rate during growth ( r G ) was 36% higher than it was during the settling phase ( Figure 6A ; r G, Standard = 0.641, r S, Standard = 0.470; δ = 0.74, p = 1.07×10 −10 , Wilcoxon test). If we eliminate the growth advantage by setting the overall cellular growth rates of the strains with and without delay to be equal, we observe the same fitness benefit during settling selection ( Figure 6B ; r S, Standard = 0.470, r S, Equal = 0.496; δ = 0.154, p = 0.185, Wilcoxon test). This suggests that the benefits of synchronous division during settling selection are conferred by the changes in group morphology and size, not due to a cross-level byproduct of increased cellular growth rate. We then manipulated the strength of settling selection by simulating a modification of the standard settling selection protocol: initiating clusters at the top of the tube rather than at random positions in the media column. Under these altered conditions, the selection rate during settling was approximately 632% higher than the selection rate during growth ( Figure 6C ; r G , Equal = 0.640, r S, Equal = 4.05; δ = -1, p < 2.2×10 −16 , Wilcoxon test), revealing how environmental context can dramatically alter the relative importance of group-level traits in evolutionary outcomes. Again, the fitness benefit during settling selection was not dependent upon differences in cellular growth rate ( Figure 6D ; r S, Standard = 4.05, r S, Equal = 4.14; δ = -0.002, p = 0.986, Wilcoxon test). Download figure Open in new tab Figure 6. Simulated selection rates for growth and settling phases under different selection regimes and growth rate conditions. Selection rates were calculated from competition simulations using the doubling time distributions of the petite ancestor and the petite without delay strain to assess how environmental conditions and cellular growth rates influence selection during the growth and settling phases of the experiment. Two environmental conditions were tested: clusters starting at random positions (matching experimental conditions, A and B ) or at the top of the tube to maximize settling selection effects ( C and D ). Two growth rate scenarios were examined: standard growth rates ( A and C ), where the petite without delay has faster overall growth rate, and equal growth rates ( B and D ), where the net growth rate of the petite without delay was made equal to the petite ancestor by adding 15 minutes to the average doubling time. Each data point represents the mean selection rate per phase (averaged across up to 20 transfers) for each of the 50 replicate competition experiments. Changes in cellular doubling time distributions over 5,000 generations of evolution Our simulated competitions clearly predict that any mutation that eliminates the first division delay observed in the petite ancestor strain should be selectively favored under the experimental conditions of the MuLTEE. To test this prediction, we performed time-lapse microscopy on the descendants of the petite ancestor after 200, 400, 600, and 1000 days of evolution (corresponding to ~5,000 generations). Consistent with our predictions, when we compared the first and subsequent doubling time distributions of our evolved strains, we found that the first division delay had been lost in all 5 replicate populations by day 200 (~1,000 generations) and this remained true through day 1000 ( Figure 7 ; p > 0.05 after Bonferroni correction for all evolved strains; Wilcoxon test). As was the case for our ancestral strains, when we restrict our analyses to mother-daughter pairs, we detect some small but significant differences for 6 of the evolved strains after day 200 (PA1 t400, PA2 t400, PA2 t600, PA3 t600, PA5 t600, and PA1 t1000 showed slight delays in daughter cell division; Figure S9 ; p < 0.02 after Bonferroni correction, Wilcoxon rank-sum test). However, the differences in mother vs. daughter doubling time were significantly reduced relative to the petite ancestor ( p < 0.05 after Bonferroni correction for all evolved strains; one-sided Wilcoxon test) and all detected delays were less than 2 minutes on average by t600. While it is tempting to interpret this result as evidence of selection favoring synchronous cell division, it is also possible that synchronous growth is a mere side effect of fast growth and that it appears as consequence of mutations that compensate for the growth defects imposed by the petite mutation. Regardless of the ultimate cause of this change in the MuLTEE, our model shows how manipulation of mother-daughter doubling times can provide a lever for the control of multicellular morphology that can affect fitness. Download figure Open in new tab Figure 7. Recovery of synchronous cell division in the evolved derivatives of the petite strain. Doubling time distributions of the first cell division (light green) and subsequent cell divisions (dark green). The petite ancestor strain (t0) is the only one in which the first division is significantly slower than the subsequent division ( p = 7.92×10-6 after Bonferroni correction, Wilcoxon test). All of the evolved strains have no statistically significant delay in the first division ( p > 0.06 for all evolved strains after Bonferroni correction, Wilcoxon test). Supplementary Table 1 and 2 report the total number of timelapses collected and cell doublings times measured. Further increases in group size are enabled by accelerated first division Up until this point, we limited our focus to asynchronous cell divisions caused by a delay in the first cell division because this was the pattern observed in our multicellular yeast ancestors. However, any difference in the mean doubling times for the first and subsequent cell division will lead to asynchrony. Here we relax this limitation to consider alternative scenarios. We hypothesized that faster first divisions would result in larger fragmentation sizes, as clusters would expand outward more rapidly, forming more filament-like branches, allowing the cluster to accumulate more cells before reaching the edge degree fragmentation threshold. To test this hypothesis, we simulated cluster growth using synthetic log-normal doubling time distributions with varying levels of first cell division delay, ranging from positive values (as observed in the evolution of multicellular yeast) to negative values representing accelerated first divisions ( Figure 8A ). Download figure Open in new tab Figure 8. Accelerated first cell division creates a different network topology that allows for the formation of bigger clusters. A ) Synthetic doubling time distributions for strains with fast first division (−50% and -25% delay), synchronous cell divisions (0% delay), and delayed first division (25% and 50% delay). Standard deviation = 15 for all distributions. Simulations tested 21 delay values from - 50% to +50% in 5% increments. B ) Contour plot of mean normalized network diameter. C ) Contour plot of mean maximum edge degree. D ) Mean fracture size across different edge degree thresholds. E ) Contour plot of mean filament branch motifs. A filament branch is defined as at least 3 contiguous connected cells with a node degree of 2 or less. F ) Contour plot of mean mother cells with ≥1 delayed daughter cell. Blue and black dots in E and F mark the fast first division network (panel G ) and synchronous network (panel H ), respectively. G ) Example network (200 cells) with fast first division (−30% delay); blue nodes represent filamentous branches where daughter cells divided faster than their mothers. H ) Example synchronous network (200 cells); blue nodes show filamentous branches, while red (mothers) and orange (daughters) highlight motifs of mother cells with ≥2 undivided daughters. Color scales are logarithmic in panels D, E , and F . For panels B, C, E , and F : 300 simulations per delay value, grown from 1 to 1300 cells. For panel D : 30 simulations per edge degree (5-15) and delay combination, tracking fracture size over 50 generations with random parent/propagule selection after fragmentation. Extending our previous results, we find that delay is negatively correlated with normalized network diameter ( Figure 8B ; the normalization function makes this effect independent of network size) and positively correlated with maximum edge degree ( Figure 8C ). Accordingly, clusters grow to a larger size at fracture as delay decreases and the impact of this delay compounds as the edge degree threshold increases ( Figure 8D ). Finally, as hypothesized, networks generated with faster first divisions exhibited a greater number of “filamentous” network motifs, where daughter cells divide faster than their mothers ( Figure 8E ) while the previously characterized “unbudded daughter” motif was more prevalent in slow first division networks ( Figure 8F ). While networks with very severe differences in first and subsequent division times may exhibit only one type of these motifs (e.g., exclusively filamentous motifs when delay is strongly negative; Figure 8G ), both motifs will often occur in a single network when variance in doubling time is sufficiently high relative to delay ( Figure 8H ). This combination of both types of motifs causes the mean of the normalized network diameter of synchronous networks to remain close to 1, similar to that of exponentially dividing networks, since the opposing effects of both motifs counterbalance each other. Apoptosis provides an alternative path to large group size Given the substantial increase in cluster size our model suggests possible via accelerated first division, it is somewhat surprising that we do not observe accelerated first division evolving in the MuLTEE ( Figure 7 ; Figure S9 ). However, we do observe the repeated, parallel evolution of another cell-level trait that could have similar impacts on cluster morphology, albeit via an entirely different mechanism: elevated rates of apoptosis ( Ratcliff et al., 2012 ; Tong, 2022 ). Elevated rates of apoptosis are thought to have been favored in snowflake yeast because dead cells serve as weak points, facilitating fragmentation and releasing cells from the interior of large clusters where spatial constraints or nutrient diffusion may limit their growth ( Conlin & Ratcliff, 2016 ; Libby et al., 2014 ; Ratcliff et al., 2012 ). Another side-effect of cell death is that after a cell dies, it leaves no further offspring. This otherwise mundane and obvious consequence of cell death, taken in the context of our network model of snowflake yeast growth, implies that, for a network of any given size, network diameter will increase as a function of the apoptosis rate. Furthermore, because fragmentation is driven by local packing density of cells, cell death could even increase the size at which fragmentation occurs by creating pockets of less densely packed cells (that is, higher apoptosis rates may lead to a reduction in the average edge degree). We examined, and largely confirmed, these predicted effects on network diameter and edge degree by simulating the growth of exponentially dividing networks (delay = 0, standard deviation = 15) with a per division apoptosis probability ranging from 0.0-0.30. As predicted, mean network diameter monotonically increases with increasing apoptosis rates, suggesting that cell death results in less densely branched networks ( Figure S10A ) Contrary to our expectations, mean edge degree peaks at intermediate values of apoptosis (0.15) rather than at apoptosis probability = 0.0. Beyond an apoptosis probability of 0.15, however, edge degree monotonically decreases, consistent with the idea that higher rates of apoptosis will lead to a reduction in average edge degree ( Figure S10B ). Taken together, these results begin to point to a potential alternative hypothesis for the evolution of elevated apoptosis rates in the snowflake yeast system: cell death may directly contribute to increased cluster size. Next, we looked at the joint effects of cell division timing delay and apoptosis rate. We found no evidence of an interaction between apoptosis rate and delay on network diameter; increasing death probability always increases network diameter, regardless of the delay value ( Figure S11A ). On the other hand, the effect of increasing death probability on maximum edge degree depends on the delay value. Specifically, when delay is high, increasing death probability decreases max edge degree but when delay is low it can have the opposite effect ( Figure S11B ). Apoptosis rate has a balancing effect on the changes in mean edge degree caused by cell division delays; increasing the apoptosis rate reduces the range of possible max edge degree values achieved by different delays. These results could explain why we don’t observe the evolution of fast first division in the snowflake yeast system given the considerable predicted impact of high rates of apoptosis on cluster size. The evolution and maintenance of higher rates of apoptosis limits the benefits that a fast first division could provide to the organism. Discussion Our results demonstrate how the timing of cell division shapes group morphology and size with consequences for multicellular fitness. Asynchronous cell division due to a first division delay produces more branched architectures enriched with “unbudded daughter” motifs, making the older cells become more crowded and driving early fragmentation relative to synchronously dividing networks. However, not all types of asynchronous cell divisions lead to a decrease in size at fracture. Asynchrony due to high variance in doubling time does not translate to a shift in mean size at fracture. And asynchrony due to an accelerated first division actually increases size at fracture by creating “filamentous branches”, where daughter cells divide faster than their mother cell, reducing the crowdedness in the older connections of the network. Our modeling framework captures key features of cluster growth and fragmentation, but additional biophysical properties need to be studied to understand their impact on group morphology. A cellular property that has been understudied for its contributions on cluster size at fracture is bud site distribution. We noted that petite cells bud at wider angles than grande cells (47.4° vs. 35.2°), but a full investigation of the driving factor of this budding angle difference and effects of this phenotype to group properties needs to be performed. Bud site selection has been thoroughly studied in unicellular yeast, and it has been found that ploidy affects budding distributions ( Chiou et al., 2017 ; Oliferenko et al., 2009 ; Park & Bi, 2007 ). In multicellular yeast, this budding pattern can be modified by the expression of BUD4, which can lead to a change between axial or bipolar budding, causing a cluster size difference of nearly 10 times more cells ( Nanda et al., 2024 ). In addition, future work should also be aimed at further exploring the potential for interactions between the cellular properties that have been characterized to affect size at fracture. As here we show that apoptosis rate reduces the effect on cluster size that first division delay would otherwise have in the absence of apoptosis (Figure S11). Furthermore, we found that for the petite and grande strains, division synchrony, cell diameter, and cellular aspect ratio explain 35%, 25% and -13% of the size difference between the strains, respectively, but combined they explain 43% of the total size difference (Supplementary Table 3). Understanding how cellular innovations produce novel phenotypes at the group-level brings us closer to the goal of a comprehensive understanding of the evolution of multicellularity and development. Experimental evolution allows us to directly observe these modifications and track how they map onto group phenotypes and fitness. The MuLTEE has proven to be a powerful experimental system for studying these dynamics because it allows us to directly observe (and manipulate) how cellular changes translate into multicellular phenotypes in a way that would be difficult to establish in more complex organisms. Here we add to the growing catalog of cellular innovations we’ve observed in the MuLTEE, including cellular elongation ( Bozdag et al., 2023 ), strengthened intercellular connections ( Jacobeen, Graba, et al., 2018 ), and genome duplication ( Tong et al., 2025 ), each representing distinct mechanisms by which selection on multicellular traits drives the evolution of cellular properties. Together, these examples demonstrate how experimental evolution in simple multicellular systems can illuminate the mechanistic bases of major evolutionary transitions, revealing how modifications to individual cell behavior give rise to novel multicellular phenotypes. Our results suggest that temporal regulation is an evolutionarily accessible mechanism for control of morphogenesis in nascent multicellular organisms with permanent intercellular bonds, providing a foundation upon which more elaborate patterning mechanisms might later evolve. In this sense, cell division timing may represent a simple but powerful lever through which natural selection can shape multicellular organization during the earliest stages of the transition to multicellularity. Methods Soft-agar time lapse preparation The yeast strains were platted on YPD agar from 20% glycerol stock isolates stored at -80ºC and grown for 3 days at 30ºC in a static incubator. A single colony was grown for 24 hours in 10ml YPD in 25 × 150 mm culture tubes for 24 h at 30°C with 225 rpm shaking and then 1ml sample was transferred to 10ml of fresh media for a 4-hour growth for the strain to reach exponential growth phase. One milliliter of culture was centrifuged, the supernatant removed, and clusters were broken using a pipette tip. The sample was resuspended in 1 mL of 2x medium, subjected to 5 minutes of settling selection, and the top 100 μL collected. A final dilution was performed to achieve ∽200 clusters per 100 μL, based on microscopic estimation. Soft agar (0.7%) was prepared by melting in a microwave and allowed to cool slightly. Equal volumes (100 μL each) of cell suspension and soft agar were mixed and transferred to a microscopy plate well. The agar was allowed to solidify for at least 30 minutes before imaging. The time lapses where imaged using a ×20 Nikon objective and images where collected every 5 minutes for 15 hours in total. Time lapse analysis The timelapse analysis was performed using Fiji ( Schindelin et al., 2012 ), TrackMate ( Tinevez et al., 2017 ) and Cellpose 2.0 ( Pachitariu & Stringer, 2022 ) for cell tracking. Manual correction was needed to ensure correct detection of the first frame the buds appeared to guarantee accurate measurement of cell doubling times. Then custom Python scripts were used to obtain the doubling time of the cells and the mother-daughter doubling time difference to quantify synchronous cell divisions. Cell diameter and cell aspect ratio was obtained from the last segmentation before the cell divides. The budding angle was calculated as the angle defined by the points of the mother’s cell distal pole, the center of the mother cell ellipse, and the proximal pole of the daughter cell (as a proxy for the bud scar). To obtain the polarity of the cell (define the poles), a cell’s mother location needs to be known, making the proximal pole the pole of the daughter cell that is closer to its mother’s center. Measuring cell properties from time lapse microscopy To confirm that cell properties were maintained during the time lapses, cell growth was calculated from the time lapse data. This information was obtained by processing the spots and edges files output from TrackMate’s time lapse analysis to link all masks along each cell’s growth trajectory and counting the number of divisions by classifying mother and daughter cells after each cell division. From this dataset, the second division of the ancestors was used to calculate the median cell diameter (long-axis; TrackMate reports the semi-axis length, so the reported values were multiplied by 2) and aspect ratio for each cell. The second division was chosen because the median of the cell properties from the masks in the second division showed less variation than the final mask of the first cell division, which could be affected by errors in cell segmentation due mainly to cell overlap. The median cell properties from all cells were then used to calculate the overall median cell diameter and aspect ratio per strain for comparison with the static measurements. Measuring cluster and cell properties from static images Ancestral strains were grown in a YPD agar plate and 5 colonies of each were grown in different tubes of liquid YPD for 24 hours. After the first growth cycle a random sample of 100 μl was transferred to a new tube for another day to grow, and from the remanent of the tube after day 1 and with the tube of day 2, the measurements were performed. For the cluster size measurements, a sample of 1 ml of the culture was taken, and then two dilutions were performed in series in water, a 1:100 and a 1:10 for a 1:1000 total dilution. Then 500μl was transferred to 16 well-plate for imaging of the whole well using a 4x Nikon objective. The images were analyzed with an ImageJ script to calculate the cluster properties. To measure the cell shapes, the cell wall was stained using calcofluor. For this process a sample of 100μl was collected from each tube, and it was washed in PBS two times, and then it was incubated in 1ml PBS with 1μl of calcofluor and incubated for 10 minutes in the dark and washed one last time before being imaged with a 40x Nikon objective. The images were processed with Cellpose 2.0 ( Pachitariu & Stringer, 2022 ) to create the masks of the cells and manual correction was performed to delete the masks of all cells that were not fully grown or haven’t budded yet, and after the masks and images were processed in ImageJ to obtain the measurements of the cells. Finally, to obtain the distribution of cells in the cluster of the aerobic ancestor, the same cells stained with calcofluor where imaged using a 20x Nikon objective by adding 3 μl into the microscope slide and gently flattening them into a monolayer of cells. Enough images were taken to capture at least 100 clusters per replicate. To measure the number of cells per cluster, cell segmentation was performed with cellpose2 to segment all the cells, and the cluster segmentation and assignment of cell masks into a cluster was performed with a custom Fiji script. Network model (using discrete or synthetic continuous data) The complete network model was implemented in Python 3.10 using the NetworkX library ( Hagberg et al., 2008 ). Clusters are represented as a dynamical network graph G = (N, E), in which cells are nodes (N) and incomplete cell divisions as edges (E). Cell divisions are modeled by adding a new node connected to the nodes representing the respective mother cell. Upon creation, the new node and the mother node were attributed a division time which was sample from empirical data or from a log-normal distribution. After addition of all new cells, the network was evaluated for potential fragmentation events, which were determined by using our created metric the edge degree: Where deg( n 1 ) and deg( n 2 ) are the node degrees of the nodes forming an edge. Edge degree was evaluated for every node n in the network. If more than 1 edge was flagged for fragmentation, one was selected at random. Some of the codes used GNU Parallel ( Tange, 2018 ) to coordinate workflow of scripts and to increase the execution speed. Log-normal distribution estimation To create the log-normal distributions we defined the parameters for a normal distribution by selecting the mean ( μ ) and standard deviation ( σ ), and the parameters for the log-normal distribution were obtained by the formulas: Normalized network diameter estimation The normalization of the network diameter was used to compare the network topology of the simulated networks to an exponentially growing network independent of the size (supplementary Figure 5). To find the formula for the normalized network diameter, an exponential network was grown, and a linear regression was fit to the data of the log scaled number of nodes and network diameter, and from the regression parameters the formula was estimated to be: Settling selection simulations To mimic competition experiments between different strains, we first simulated the growth phase of the cluster followed by calculating the sedimentation velocities in order to perform settling selection. For the growth phase, clusters were sampled from an initial cluster size distribution and initiated in a 1:1 ratio of both competing strains. These initial clusters were grown for 24h or until the carrying capacity of the simulation is reached (carrying capacity is 10 million cells unless specified otherwise). Next, the settling selection is performed consisting first of a 10% dilution of the total biomass (total number of cells), and then the selection of the first 10% of clusters that reach the end of the length of a tube (2.5 cm), when starting the cells at a random position or at the top of the tube, and the surviving cells are going to be transferred to the next growth phase for the defined amount of cycles or until one population goes extinct. The growth phases were performed in Python using the network model described above, and the settling simulations were performed using a spatial model in MATLAB to calculate the diameter of the clusters ( Bozdag et al., 2023 ). To estimate which clusters will reach the bottom of the tube and survive during settling selection, we calculate the cluster diameter from the simulated clusters and use Stokes’ law to estimate the velocity of sedimentation ( Ratcliff et al., 2013 ; Pineau et al., 2024 ). Each cluster diameter is obtained by calculating the radius of gyration, obtained by calculating the root-mean-square distance between the cells and the center of the cluster, and then transformed to the sphere diameter using the formula where Rg is Radius of gyration ( Rubinstein & Colby, 2003 , p. 64). The sedimentation velocity is calculated as: where d i is the diameter of the cluster, g is the gravity constant ( g = 9.81 m s −2 ), ρ d is the mass density of yeast cells ( ρ d = 1112.6 kg m 3 ; Bryan et al., 2010 ), ρ w is the mass density of water at 25°C ( ρ w = 997 kg m 3 ), and η is the dynamic viscosity of water at 25°C ( η = 9.81 × 10 −4 Pa · s ). Biophysical model In this biophysical model, cells are represented as ellipsoids with defined diameter and aspect ratio parameters. New cells are added by sampling both the budding location and growth angle, if the sampled position is too close to an already present budded cell a new position is going to be sampled, there is a maximum of 10 attempts to find a location for the new cell that is not so crowded, but if the position can’t be found then the last sampled position is going to be used. Mechanical stress is going to be estimated as the overlap between cells, therefore cluster fragmentation occurs when the overlap between cells exceeds a specified threshold. In the cases where the clusters are built from networks ( Figure 6 and Supplementary Figure 3), the cells are added in order in which the nodes were added to the network. Cluster size difference predictions using a spatial model The goal of this procedure is to approximate the experimental cluster size measurements with the size at fracture predicted by the spatial MATLAB model ( Bozdag et al., 2023 ), specifically examining three key variables: cell aspect ratio, cell diameter, and cell synchrony. For each individual variable, we performed a controlled analysis where we first calculated the overlap threshold needed to trigger fragmentation at the target volume while holding all three variables constant to the grande strain values. After establishing this threshold, we then calculated the predicted fracture size using the specific parameters of each ancestral strain (petite and grande) for the variable being tested, while keeping the other two variables constant at reference values. The contribution of each variable to the observed size difference between ancestors was quantified as the percentage of the experimental difference explained by the simulated difference. For both cell aspect ratio and cell diameter analyses, we implemented synchronous cell division as the standard condition for both strains. Use of LLMs in writing and coding LLM assistance was used for some aspects of writing and coding of the scripts used to generate the results presented in this manuscript. For writing, Claude (Sonnet 3.5 and newer) was used to suggest style edits of early versions of the manuscript. All AI-suggested edits were carefully reviewed to guarantee that the content of the edited text was preserved and that the resulting text still accurately represented the authors’ intent. For later versions of the manuscript, all modifications were performed without any AI assistance. For coding, Claude (Sonnet 3.5 and newer) was used to aid the development of the following: creating bash scripts including the ones used for parallel execution of scripts and the workflow of the settling simulations; modifications to the MATLAB script for the biophysical simulations; modifications to the base network simulations; and refinement of the figures presented in this article. AI assistance was not used in the creation of the base network simulation functions, and the Python and ImageJ scripts used to analyze the microscopy data. All AI-suggested additions to code were carefully reviewed and tested to verify that the new code worked as intended. Supplementary Materials Supplementary Figures 1-11 Supplementary Tables 1-3 Supplementary Videos 1-2 Acknowledgments We thank Peter J. Yunker and Tom E.R. Belpaire for their thorough comments on the manuscript. This work was supported by NIH grant No. 5R35GM138030 to W.C.R. Funder Information Declared National Institutes of Health, https://ror.org/01cwqze88 , 5R35GM138030 Footnotes The manuscript has been thoroughly revised. The simulations and experiments presented in the original submission have been extended to provide stronger support for the central argument of the article and to offer a more complete characterization of the observed phenomenon. References ↵ Blank , H. M. , Callahan , M. , Pistikopoulos , I. P. E. , Polymenis , A. O. , & Polymenis , M. ( 2018 ). Scaling of G1 Duration with Population Doubling Time by a Cyclin in Saccharomyces cerevisiae . Genetics , 210 ( 3 ), 895 – 906 . doi: 10.1534/genetics.118.301507 OpenUrl Abstract / FREE Full Text ↵ Bourke , A. F. G. ( 2011 ). Principles of social evolution . Oxford University Press . ↵ Bozdag , G. O. , Libby , E. , Pineau , R. , Reinhard , C. T. , & Ratcliff , W. C. ( 2021 ). Oxygen suppression of macroscopic multicellularity . Nature Communications , 12 ( 1 ), 2838 . doi: 10.1038/s41467-021-23104-o OpenUrl CrossRef PubMed ↵ Bozdag , G. O. , Zamani-Dahaj , S. A. , Day , T. C. , Kahn , P. C. , Bumetti , A. J. , Lac , D. T. , Tong , K. , Conlin , P. L. , Balwani , A.H. , Dyer , E. L. , Yunker , P. J. , & Ratcliff , W. C. ( 2023 ). De nova evolution of macroscopic multicellularity . Nature , 617 ( 7962 ), 747 – 754 . doi: 10.1038/s41586-023-06052-1 OpenUrl CrossRef PubMed ↵ Brunet , T. , Larson , B. T. , Linden , T. A. , Vermeij , M. J. A. , McDonald , K. , & King , N. ( 2019 ). Light-regulated collective contractility in a multicellular choanoflagellate . Science , 366 ( 6463 ), 326 – 334 . doi: 10.1126/science.aay2346 OpenUrl Abstract / FREE Full Text ↵ Bryan , A. K. , Goranov , A. , Amon , A. , & Manalis , S. R. ( 2010 ). Measurement of mass, density, and volume during the cell cycle of yeast . Proceedings of the National Academy of Sciences , 107 ( 3 ), 999 – 1004 . doi: 10.1073/pnas.0901851107 OpenUrl Abstract / FREE Full Text ↵ Chiou , J. , Balasubramanian , M. K. , & Lew , D. J. ( 2017 ). Cell Polarity in Yeast . Annual Review of Cell and Developmental Biology , 33 ( 1 ), 77 – 101 . doi: 10.1146/annurevcellbio-100616-060856 OpenUrl CrossRef PubMed ↵ Conlin , P. L. , & Ratcliff , W. C. ( 2016 ). Trade-offs drive the evolution of increased complexity in nascent multicellular digital organisms . Multicellularity: Origins and Evolution , 131 – 147 . ↵ Crider , D. G. , García-Rodriguez , L. J. , Srivastava , P. , Peraza-Reyes , L. , Upadhyaya , K. , Boldogh , R. , & Pon , L. A. ( 2012 ). Rad53 is essential for a mitochondrial DNA inheritance checkpoint regulating G1 to S progression . Journal of Cell Biology , 198 ( 5 ), 793 – 798 . doi: 10.1083/jcb.201205193 OpenUrl Abstract / FREE Full Text ↵ Day , T. C. , Hohn , S. S. , Zamani-Dahaj , S. A. , Yanni , D. , Bumetti , A. , Pentz , J. , Honerkamp Smith , A. R. , Wioland , H. , Sleath , H. R. , Ratcliff , W. C. , Goldstein , R. E. , & Yunker , P. J. ( 2022 ). Cellular organization in lab-evolved and extant multicellular species obeys a maximum entropy law . eLife , 11 , e72707 . doi: 10.7554/eLife.72707 OpenUrl CrossRef PubMed ↵ Di Talia , S. , Wang , H. , Skotheim , J. M. , Rosebrock , A. P. , Futcher , B. , & Cross , F. R. ( 2009 ). Daughter-Specific Transcription Factors Regulate Cell Size Control in Budding Yeast . PLoS Biology , 7 ( 10 ), e1000221 . doi: 10.1371/joumal.pbio.1000221 OpenUrl CrossRef PubMed ↵ Duran-Nebreda , S. , & Sole , R. ( 2015 ). Emergence of multicellularity in a model of cell gmwth, death and aggregation under size-dependent selection . Journal of The Royal Society Interface , 12 ( 102 ), 20140982 . doi: 10.1098/rsif.2014.0982 OpenUrl CrossRef PubMed ↵ Featherston , J. , Arakaki , Y. , Hanschen , E. R. , Ferris , P. J. , Michod , R. E. , Olson , B. J. S. C. , Nozaki , H. , & Durand , P. M. ( 2018 ). The 4-Celled Tetrabaena socialis Nuclear Genome Reveals the Essential Components for Genetic Control of Cell Number at the Origin of Multicellularity in the Volvocine Lineage . Molecular Biology and Evolution , 35 ( 4 ), 855 – 870 . doi: 10.1093/molbev/msx332 OpenUrl CrossRef PubMed ↵ Flores , E. , & Herrero , A. ( 2010 ). Compartmentalized function through cell differentiation in filamentous cyanobacteria . Nature Reviews Microbiology , 8 ( 1 ), 39 – 50 . doi: 10.1038/nrmicro2242 OpenUrl CrossRef PubMed Web of Science ↵ Fung , L. , Konkol , A. , Ishikawa , T. , Larson , B. T. , Brunet , T. , & Goldstein , R. E. ( 2023 ). Swimming, Feeding, and Inversion of Multicellular Choanoflagellate Sheets . Physical Review Letters , 131 ( 16 ), 168401 . doi: 10.1103/PhysRevLett.131.168401 OpenUrl CrossRef PubMed ↵ Gorospe , C. M. , Carvalho , G. , Herrera Curbelo , A. , Marchhart , L. , Mendes , I. C. , Nied źwiecka , K. , & Wanrooij , P.H. ( 2023 ). Mitochondrial membrane potential acts as a retrograde signal to regulate cell cycle progression . Life Science Alliance , 6 ( 12 ), e202302091 . doi: 10.26508/lsa.202302091 OpenUrl Abstract / FREE Full Text ↵ Hagberg , A. A. , Schult , D. A. , & Swart , P. J. ( 2008 ). Exploring Network Structure, Dynamics, and Function using NetworkX . Proceedings of the 7th Python in Science Conference (SciPy 2008) , 11 – 15 . http://conference.scipy.org.s3-website-us-east-1.amazonaws.com/proceedings/scipy2008/paper_2/ ↵ Hanschen , E. R. , Marriage , T. N. , Ferris , P. J. , Hamaji , T. , Toyoda , A. , Fujiyama , A. , Neme , R. , Noguchi , H. , Minakuchi , Y. , Suzuki , M. , Kawai-Toyooka , H. , Smith , D. R. , Sparks , H. , Anderson , J. , Bakarić , R. , Luria , V. , Karger , A. , Kirschner , M. W. , Durand , P. M. , … Olson , B. J. S. C. ( 2016 ). The Gonium pectorale genome demonstrates co-option of cell cycle regulation during the evolution of multicellularity . Nature Communications , 7 ( 1 ), 11370 . doi: 10.1038/ncomms11370 OpenUrl CrossRef PubMed ↵ Hartwell , L. H. ( 1974 ). Saccharomyces cerevisiae cell cycle . Bacteriological Reviews , 38 ( 2 ), 164 – 198 . doi: 10.1128/br.38.2.164-198.1974 OpenUrl FREE Full Text ↵ Hartwell , L. H. , & Unger , M. W. ( 1977 ). Unequal division in Saccharomyces cerevisiae and its implications for the control of cell division . The Journal of Cell Biology , 75 ( 2 ), 422 – 435 . doi: 10.1083/jcb.75.2-422 OpenUrl Abstract / FREE Full Text ↵ Hatzis , C. , & Porro , D. ( 2006 ). Morphologically-structured models of growing budding yeast populations . Journal of Biotechnology , 124 ( 2 ), 420 – 438 . doi: 10.1016/j.jbiotec.2006.01.011 OpenUrl CrossRef PubMed ↵ Herrero , A. , Stavans , J. , & Flores , E. ( 2016 ). The multicellular nature of filamentous heterocyst-forming cyanobacteria . FEMS Microbiology Reviews , 40 ( 6 ), 831 – 854 . doi: 10.1093/femsre/fuwo29 OpenUrl CrossRef PubMed ↵ Herron , M. D. , Conlin , P. L. , & Ratcliff , W. C. ( 2022 ). The Evolution of Multicellularity ( 1st ed.). CRC Press . doi: 10.1201/9780429351907 OpenUrl CrossRef ↵ Jacobeen , S. , Graba , E. C. , Brandys , C. G. , Day , T. C. , Ratcliff , W. C. , & Yunker , P. J. ( 2018 ). Geometry, packing, and evolutionary paths to increased multicellular size . Physical Review E , 97 ( 5 ), 050401 . doi: 10.1103/PhysRevE.97.050401 OpenUrl CrossRef Jacobeen , S. , Pentz , J. T. , Graba , E. C. , Brandys , C. G. , Ratcliff , W. C. , & Yunker , P. J. ( 2018 ). Cellular packing, mechanical stress and the evolution of multicellularity . Nature Physics , 14 ( 3 ), 286 – 290 . doi: 10.1038/s41567-017-0002-y OpenUrl CrossRef PubMed ↵ Johnston , G. , Pringle , J. , & Hartwell , L. ( 1977 ). Coordination of growth with cell division in the yeast . Experimental Cell Research , 105 ( 1 ), 79 – 98 . doi: 10.1016/0014-4827(77)90154-9 OpenUrl CrossRef PubMed Web of Science ↵ Kirk , D. L. ( 2005 ). A twelve-step program for evolving multicellularity and a division of labor . BioEssays , 27 ( 3 ), 299 – 310 . doi: 10.1002/bies.20197 OpenUrl CrossRef PubMed Web of Science ↵ Kondrashov , A. S. ( 1994 ). Mutation load under vegetative reproduction and cytoplasmic inheritance . Genetics , 137 ( 1 ), 311 – 318 . doi: 10.1093/genetics/137.1.311 OpenUrl Abstract / FREE Full Text ↵ Laabs , T. L. , Markwardt , D. D. , Slattery , M. G. , Newcomb , L. L. , Stillman , D. J. , & Heideman , W. ( 2003 ). ACE2 is required for daughter cell-specific G1 delay in Saccharomyces cerevisiae . Proceedings of the National Academy of Sciences , 100 ( 18 ), 10275 – 10280 . doi: 10.1073/pnas.1833999100 OpenUrl Abstract / FREE Full Text ↵ Lavrentovich , M. O. , Koschwanez , J. H. , & Nelson , D. R. ( 2013 ). Nutrient shielding in clusters of cells . Physical Review E , 87 ( _6 ), 062703 . doi: 10.1103/PhysRevE.87.062703 OpenUrl CrossRef PubMed ↵ Leite , A. C. , Costa , V. , & Pereira , C. ( 2023 ). Mitochondria and the cell cycle in budding yeast . The International Journal of Biochemistry & Cell Biology , 161 , 106444 . doi: 10.1016/j.biocel.2023.106444 OpenUrl CrossRef PubMed ↵ Lenski , R. E. , Rose , M. R. , Simpson , S. C. , & Tadler , S. C. ( 1991 ). Long-Term Experimental Evolution in Escherichia coli. I. Adaptation and Divergence During 2,000 Generations . The American Naturalist , 138 ( 6 ), 1315 – 1341 . doi: 10.1086/285289 OpenUrl CrossRef PubMed ↵ Libby , E. , Ratcliff , W. , Travisano , M. , & Kerr , B. ( 2014 ). Geometry Shapes Evolution of Early Multicellularity . PLoS Computational Biology , 10 ( 9 ), e1003803 . doi: 10.1371/journal.pcbi.1003803 OpenUrl CrossRef PubMed ↵ Lord , P. G. , & Wheals , A. E. ( 1980 ). Asymmetrical division of Saccharomyces cerevisiae . Journal of Bacteriology , 142 ( 3 ), 808 – 818 . doi: 10.1128/jb.142.3.808-818.1980 OpenUrl Abstract / FREE Full Text ↵ Maliet , O. , Shelton , D. E. , & Michod , R. E. ( 2015 ). A model for the origin of group reproduction during the evolutionary transition to multicellularity . Biology Letters , 11 ( 6 ), 20150157 . doi: 10.1098/rsbl.2015.0157 OpenUrl CrossRef PubMed ↵ Michod , R. E. ( 2000 ). Darwinian dynamics: Evolutionary transitions in fitness and individuality (2. print., and 1. paperback print) . Princeton University Press . ↵ Muñoz-García , J. , & Ares , S. ( 2016 ). Formation and maintenance of nitrogen-fixing cell patterns in filamentous cyanobacteria . Proceedings of the National Academy of Sciences , 113 ( 22 ), 6218 – 6223 . doi: 10.1073/pnas.1524383113 OpenUrl Abstract / FREE Full Text ↵ Nanda , P. , Barrere , J. , LaBar , T. , & Murray , A. W. ( 2024 ). A dynamic network model predicts the phenotypes of multicellular clusters from cellular properties . Current Biology , 34 ( 12 ), 2672 - 2683.e4 . doi: 10.1016/j.cub.2024.05.014 OpenUrl CrossRef PubMed ↵ Oliferenko , S. , Chew , T. G. , & Balasubramanian , M. K. ( 2009 ). Positioning cytokinesis . Genes & Development , 23 ( 6 ), 660 – 674 . doi: 10.1101/gad.1772009 OpenUrl Abstract / FREE Full Text ↵ Oud , B. , Guadalupe-Medina , V. , Nijkamp , J. F. , De Ridder , D. , Pronk , J. T. , Van Maris , A. J. A. , & Daran , J.-M. ( 2013 ). Genome duplication and mutations inACE2 cause multicellular, fast-sedimenting phenotypes in evolved Saccharomyces cerevisiae . Proceedings of the National Academy of Sciences , 110 ( 45 ). doi: 10.1073/pnas.1305949110 OpenUrl Abstract / FREE Full Text ↵ Pachitariu , M. , & Stringer , C. ( 2022 ). Cellpose 2.0: How to train your own model . Nature Methods , 19 ( 12 ), 1634 – 1641 . doi: 10.1038/s41592-022-01663-4 OpenUrl CrossRef PubMed ↵ Park , H.-O. , & Bi , E. ( 2007 ). Central Roles of Small GTPases in the Development of Cell Polarity in Yeast and Beyond . Microbiology and Molecular Biology Reviews , 71 ( 1 ), 48 – 96 . doi: 10.1128/MMBR.00028-06 OpenUrl Abstract / FREE Full Text ↵ Pentz , J. T. , MacGillivray , K. , DuBose , J. G. , Conlin , P. L. , Reinhardt , E. , Libby , E. , & Ratcliff , W. C. ( 2023 ). Evolutionary consequences of nascent multicellular life cycles . eLife , 12 , e84336 . doi: 10.7554/eLife.84336 OpenUrl CrossRef PubMed ↵ Pichugin , Y. , & Traulsen , A. ( 2020 ). Evolution of multicellular life cycles under costly fragmentation . PLOS Computational Biology , 16 ( 11 ), e1008406 . doi: 10.1371/journal.pcbi.1008406 OpenUrl CrossRef PubMed ↵ Pineau , R. M. , Libby , E. , Demory , D. , Lac , D. T. , Day , T. C. , Bravo , P. , Yunker , P. J. , Weitz , J. S. , Bozdag , G. O. , & Ratcliff , W. C. ( 2024 ). Emergence and maintenance of stable coexistence during a long-term multicellular evolution experiment . Nature Ecology & Evolution . doi: 10.1038/s41559-024-02367-y OpenUrl CrossRef ↵ Ratcliff , W. C. , Denison , R. F. , Borrello , M. , & Travisano , M. ( 2012 ). Experimental evolution of multicellularity . Proceedings of the National Academy of Sciences , 109 ( 5 ), 1595 – 1600 . doi: 10.1073/pnas.1115323109 OpenUrl Abstract / FREE Full Text ↵ Ratcliff , W. C. , Fankhauser , J. D. , Rogers , D. W. , Greig , D. , & Travisano , M. ( 2015 ). Origins of multicellular evolvability in snowflake yeast . Nature Communications , 6 ( 1 ), 6102 . doi: 10.1038/ncomms7102 OpenUrl CrossRef PubMed ↵ Ratcliff , W. C. , Pentz , J. T. , & Travisano , M. ( 2013 ). TEMPO AND MODE OF MULTICELLULAR ADAPTATION IN EXPERIMENTALLY EVOLVED SACCHAROMYCES CEREVISIAE . Evolution , 67 ( 6 ), 1573 – 1581 . doi: 10.1111/evo.12101 OpenUrl CrossRef PubMed Web of Science ↵ Roisin-Bouffay , C. , Jang , W. , Caprette , D. R. , & Gomer , R. H. ( 2000 ). A Precise Group Size in Dictyostelium Is Generated by a Cell-Counting Factor Modulating Cell-Cell Adhesion . Molecular Cell , 6 ( 4 ), 953 – 959 . doi: 10.1016/S1097-2765(05)00082-1 OpenUrl CrossRef PubMed Web of Science ↵ Roze , D. , & Michod , R. E. ( 2001 ). Mutation, Multilevel Selection, and the Evolution of Propagule Size during the Origin of Multicellularity . The American Naturalist , 158 ( 6 ), 638 – 654 . doi: 10.1086/323590 OpenUrl CrossRef PubMed Web of Science ↵ Rubinstein , M. , & Colby , R. H. ( 2003 ). Polymer physics . Oxford university press . ↵ Schindelin , J. , Arganda-Carreras , I. , Frise , E. , Kaynig , V. , Longair , M. , Pietzsch , T. , Preibisch , S. , Rueden , C. , Saalfeld , S. , Schmid , B. , Tinevez , J.-Y. , White , D. J. , Hartenstein , V. , Eliceiri , K. , Tomancak , P. , & Cardona , A. ( 2012 ). Fiji: An open-source platform for biological image analysis . Nature Methods , 9 ( 7 ), 676 – 682 . doi: 10.1038/nmeth.2019 OpenUrl CrossRef PubMed Web of Science ↵ Shelton , D. E. , & Michod , R. E. ( 2014 ). Group Selection and Group Adaptation During a Major Evolutionary Transition: Insights from the Evolution of Multicellularity in the Volvocine Algae . Biological Theory , 9 ( 4 ), 452 – 469 . doi: 10.1007/s13752-014-0159-x OpenUrl CrossRef ↵ Siu , C. , Harris , T. , Wang , J. , & Wong , E. ( 2004 ). Regulation of cell-cell adhesion during Dictyostelium development . Seminars in Cell & Developmental Biology , 15 ( 6 ), 633 – 641 . doi: 10.1016/S1084-9521(04)00090-4 OpenUrl CrossRef PubMed Web of Science ↵ Staps , M. , Van Gestel , J. , & Tamita , C. E. ( 2019 ). Emergence of diverse life cycles and life histories at the origin of multicellularity . Nature Ecology & Evolution , 3 ( 8 ), 1197 – 1205 . doi: 10.1038/s41559-019-0940-o OpenUrl CrossRef PubMed ↵ Stoy , K. S. , Bingham , E. , Bumetti , A. , Datta , S. , Pineau , R. , Peterson , A. , & Ratcliff , W. C. ( 2024 ). Adaptive radiation during the evolution of complex multicellularity . Evolutionary Journal of the Linnean Society , 3 ( 1 ), kzaeoo8 . doi: 10.1093/evolinnean/kzaeoo8 OpenUrl CrossRef ↵ Strassmann , J. E. , & Queller , D. C. ( 2011 ). Evolution of cooperation and control of cheating in a social microbe . Proceedings of the National Academy of Sciences , 108 ( supplement_2 ), 10855 – 10862 . doi: 10.1073/pnas.1102451108 OpenUrl Abstract / FREE Full Text ↵ Szathmáry , E. , & Smith , J. M. ( 1995 ). The major evolutionary transitions . Nature , 374 ( 6519 ), 227 – 232 . doi: 10.1038/374227ao OpenUrl CrossRef PubMed Web of Science ↵ Tange , O. ( 2018 ). Gnu Parallel 2018 . Zenodo . doi: 10.5281/ZENODO.1146014 OpenUrl CrossRef ↵ Thompson , P. W. , & Wheals , A. E. ( 1980 ). Asymmetrical Division of Saccharomyces cerevisiae in Glucose-limited Chemostat Culture . Microbiology , 121 ( 2 ), 401 – 409 . doi: 10.1099/00221287-121-2-401 OpenUrl CrossRef ↵ Tinevez , J.-Y. , Perry , N. , Schindelin , J. , Hoopes , G. M. , Reynolds , G. D. , Laplantine , E. , Bednarek , S. Y. , Shorte , S. L. , & Eliceiri , K. W. ( 2017 ). TrackMate: An open and extensible platform for single-particle tracking . Methods , 115 , 80 – 90 . doi: 10.1016/j.ymeth.2016.09.016 OpenUrl CrossRef PubMed ↵ Tong , K. ( 2022 ). Evolution of Cell Differentiation and Whole Genome Duplication in a Multicellularity Long-Term Evolution Experiment [Ph.D., Georgia Institute of Technology] . In ProQuest Dissertations and Theses. ProQuest Dissertations & Theses A&I; ProQuest One Academic (3275496342) . https://www.proquest.com/dissertationstheses/evolution-cell-differentiation-whole-genome/docview/3275496342/se-2?accountid=11107 ↵ Tong , K. , Datta , S. , Cheng , V. , Haas , D. J. , Gourisetti , S. , Yopp , H. L. , Day , T. C. , Lac , D. T. , Khalil , A. S. , Conlin , P. L. , Bozdag , G. O. , & Ratcliff , W. C. ( 2025 ). Genome duplication in a long-term multicellularity evolution experiment . Nature . doi: 10.1038/s41586-025-08689-6 OpenUrl CrossRef ↵ Travisano , M. , & Lenski , R. E. ( 1996 ). Long-Term Experimental Evolution in Escherichia coli. IV. Targets of Selection and the Specificity of Adaptation . Genetics , 143 ( 1 ), 15 – 26 . doi: 10.1093/genetics/143.1.15 OpenUrl Abstract / FREE Full Text ↵ Weber , T. S. , Jaehnert , I. , Schichor , C. , Or-Guil , M. , & Carneiro , J. ( 2014 ). Quantifying the Length and Variance of the Eukaryotic Cell Cycle Phases by a Stochastic Model and Dual Nucleoside Pulse Labelling . PLoS Computational Biology , 10 ( 7 ), e1003616 . doi: 10.1371/joumal.pcbi.1003616 OpenUrl CrossRef PubMed ↵ West , S. A. , Fisher , R. M. , Gardner , A. , & Kiers , E. T. ( 2015 ). Major evolutionary transitions in individuality . Proceedings of the National Academy of Sciences , 112 ( 33 ), 10112 – 10119 . doi: 10.1073/pnas.1421402112 OpenUrl Abstract / FREE Full Text ↵ Wong , W. , Bravo , P. , Yunker , P. J. , Ratcliff , W. C. , & Bumetti , A. J. ( 2025 ). Oxygen-binding proteins aid oxygen diffusion to enhance fitness of a yeast model of multicellularity . PLOS Biology , 23 ( 1 ), e3002975 . doi: 10.1371/joumal.pbio.3002975 OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted May 18, 2026. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Cell division timing shapes the morphology and size of nascent multicellular organisms Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Cell division timing shapes the morphology and size of nascent multicellular organisms Luis F. Cedeño-Pérez , Rozenn M. Pineau , Thomas C. Day , William C. Ratcliff , Peter L. Conlin bioRxiv 2025.04.23.650085; doi: https://doi.org/10.1101/2025.04.23.650085 Share This Article: Copy Citation Tools Cell division timing shapes the morphology and size of nascent multicellular organisms Luis F. Cedeño-Pérez , Rozenn M. Pineau , Thomas C. Day , William C. Ratcliff , Peter L. Conlin bioRxiv 2025.04.23.650085; doi: https://doi.org/10.1101/2025.04.23.650085 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Evolutionary Biology Subject Areas All Articles Animal Behavior and Cognition (7629) Biochemistry (17660) Bioengineering (13881) Bioinformatics (41911) Biophysics (21436) Cancer Biology (18578) Cell Biology (25482) Clinical Trials (138) Developmental Biology (13371) Ecology (19887) Epidemiology (2067) Evolutionary Biology (24302) Genetics (15599) Genomics (22483) Immunology (17728) Microbiology (40364) Molecular Biology (17163) Neuroscience (88537) Paleontology (666) Pathology (2830) Pharmacology and Toxicology (4821) Physiology (7637) Plant Biology (15129) Scientific Communication and Education (2045) Synthetic Biology (4290) Systems Biology (9817) Zoology (2269)