Multicellularity makes somatic differentiationevolutionarily stableMary E. Wahla,b,1 and Andrew W. Murraya,b,2

aDepartment of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138; and bFAS Center for Systems Biology, Harvard University,Cambridge, MA 02138

This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2014.

Contributed by Andrew W. Murray, June 15, 2016 (sent for review September 13, 2015; reviewed by Harmit S. Malik and Boris I. Shraiman)

Many multicellular organisms produce two cell lineages: germ cells,whose descendants produce the next generation, and somatic cells,which support, protect, and disperse the germ cells. This germ-somademarcation has evolved independently in dozens of multicellulartaxa but is absent in unicellular species. A common explanation holdsthat in these organisms, inefficient intercellular nutrient exchangecompels the fitness cost of producing nonreproductive somatic cellsto outweigh any potential benefits. We propose instead that theabsence of unicellular, soma-producing populations reflects theirsusceptibility to invasion by nondifferentiating mutants that ulti-mately eradicate the soma-producing lineage. We argue that multi-cellularity can prevent the victory of such mutants by giving germcells preferential access to the benefits conferred by somatic cells. Theabsence of natural unicellular, soma-producing species previouslyprevented these hypotheses from being directly tested in vivo: toovercome this obstacle, we engineered strains of the budding yeastSaccharomyces cerevisiae that differ only in the presence or absenceof multicellularity and somatic differentiation, permitting direct com-parisons between organisms with different lifestyles. Our strains im-plement the essential features of irreversible conversion from germline to soma, reproductive division of labor, and clonal multicellularitywhile maintaining sufficient generality to permit broad extension ofour conclusions. Our somatic cells can provide fitness benefits thatexceed the reproductive costs of their production, even in unicellularstrains. We find that nondifferentiating mutants overtake unicellularpopulations but are outcompeted by multicellular, soma-producingstrains, suggesting that multicellularity confers evolutionary stabilityto somatic differentiation.

evolution | multicellularity | differentiation | synthetic biology | yeast

Somatic differentiation, a permanent change in gene expressioninherited by all of a cell’s descendants, produces somatic cells

from a totipotent germ line. Although somatic cells may divideindefinitely, they cannot beget the complete organism and are thusconsidered nonreproductive. Generating such sterile cells has clearfitness costs that must be offset by somatic functions that improvethe viability or fecundity of germ cells. The absence of a soma inunicellular species (1), as well as the persistence of undifferentiatedmulticellular groups among the volvocine algae (2) and cyano-bacteria (3), has fueled speculation that multicellularity must arisebefore somatic differentiation can evolve (4–7). It has been arguedthat somatic differentiation is not observed in unicellular speciesbecause the fitness benefits of somatic cells can never exceed thecost of making them (6–8): although soma can contribute motilityand protective structures to multicellular organisms, somatic cellsin a unicellular species can only benefit the germ line by secretinguseful products into a shared extracellular milieu. However,nutrient exchange between members of microbial consortia (9,10) demonstrates the potential for productive interactions be-tween cell types in the absence of physical adhesion. Benefitsassociated with somatic differentiation, including reproductivedivision of labor (11) and suppression of germ-line mutations

through lineage sequestration (12) or reduced oxidative stress(13), are thus likely accessible to unicellular species.We propose the alternative hypothesis that unicellular somatic

differentiation can offer fitness benefits in a population of ge-netically identical cells but remains rare because it is not anevolutionarily stable strategy (14). Commonly occurring mutantsthat do not differentiate (“cheats”) could take advantage of somaticcell products in the shared media without paying the reproductivecosts of differentiation, thus increasing in frequency until their ge-notype prevails. We also posit that if multicellularity results fromcells of a single lineage failing to disperse (rather than cells aggre-gating from different lineages), differentiating populations canoutcompete cheats: although cheats initially arise through mutationin a group with somatic cells (which the cheats can exploit), lineage-restricted propagation forces the cheat’s descendants to be confinedto multicellular groups composed entirely of cheats, which thuscannot benefit from the local accumulation of somatic cell products(15–17). This hypothesis invokes the demonstrated ability of pop-ulation structure to maintain altruistic traits (15, 18, 19).To test the evolutionary stability of germ-soma differentiation,

we designed strains of the budding yeast Saccharomyces cerevisiaethat produce soma, are multicellular, or combine both traits: onestrain is a multicellular, differentiating organism and the other tworepresent both possible intermediates in its evolution from anondifferentiating, unicellular ancestor (Fig. 1A). Using syntheticstrains that differ from one another at only a few, well-defined loci

Significance

Unicellular species lack the nonreproductive somatic cell typesthat characterize complex multicellular organisms. We considertwo alternative explanations: first, that the costs of lost re-productive potential never exceed the benefits of somatic cells inunicellular organisms; and second, that somatic cells may profit aunicellular population but leave it vulnerable to invasion bycommon mutants. We test these hypotheses using engineeredyeast strains that permit direct comparisons of fitness and evo-lutionary stability between lifestyles. We find that the benefits ofsomatic cell production can exceed the costs in unicellular strains.Multicellular, soma-producing strains resist invasion by non-differentiating mutants that overtake unicellular populations,supporting the theory that somatic differentiation is stabilized bypopulation structure imposed by multicellularity.

Author contributions: M.E.W. and A.W.M. designed research; M.E.W. performed research;M.E.W. analyzed data; and M.E.W. and A.W.M. wrote the paper.

Reviewers: H.S.M., Fred Hutchinson Cancer Research Center; and B.I.S., University of Cal-ifornia, Santa Barbara.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1Present address: Microsoft New England Research and Development Center, Cambridge,MA 02142.

2To whom correspondence should be addressed. Email: awm@mcb.harvard.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1608278113/-/DCSupplemental.

8362–8367 | PNAS | July 26, 2016 | vol. 113 | no. 30 www.pnas.org/cgi/doi/10.1073/pnas.1608278113

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

ensures that no undesired variables confound the direct comparisonof fitness and evolutionary stability and permits the study of a uni-cellular, differentiating lifestyle not found in nature. Experiments onliving organisms also avoid the potential pitfall of biologically un-realistic parameter regimes in simulations or purely analyticalmodels. Our results show that production of soma can be advan-tageous even in unicellular populations, although these are sus-ceptible to invasion by nondifferentiating cheats, which multicellularpopulations effectively repel. We therefore speculate that multi-cellularity facilitates the evolution of somatic differentiation prin-cipally by providing stability against common mutants that preventdifferentiation.

Construction of a Synthetic Differentiation SystemWe mimicked somatic differentiation by engineering fast-growing“germ” cells that can give rise to slower dividing, differentiated“somatic” cells that secrete invertase (Suc2), an enzyme that hy-drolyzes sucrose (which laboratory yeast strains cannot take updirectly) into the monosaccharides glucose and fructose, which anycell in the shared medium can then import (20, 21) (Fig. 1B). Thesomatic cells thus perform a digestive function: liberating mono-saccharides, the sole importable carbon source during growth in

sucrose minimal media. In nature, differentiation is often achievedthrough multiply redundant gene regulatory networks that stabilizecell fate (22): to simplify our system, we instead made differenti-ation permanent and heritable by making the expression of somaticcell-specific genes depend on a site-specific recombinase that ex-cises a gene needed for rapid cell proliferation (Fig. 1C).Germ and somatic cells must be present at a suitable ratio for

growth on sucrose: germ cells have the higher maximum growthrate, but monosaccharides become limiting when somatic cells arerare. We predicted that the ratio between cell types would reach asteady-state value reflecting the balance between unidirectionalconversion of germ cells into somatic cells and the restricted di-vision of somatic cells. We designed tunable differentiation anddivision rates to allow us to regulate the ratio between cell typesand thus control the growth rate of the culture as a whole.Both features depend on a single, genetically engineered locus

(Fig. 1C). In germ cells, this locus expresses the fluorescent proteinmCherry and a gene that accelerates cell division, the cycloheximideresistant (cyh2r) allele of ribosomal protein L28 (23); in somaticcells, the locus expresses a different fluorescent protein (mCitrine)and invertase. The germ-line form is converted to the somatic formby a version of Cre recombinase engineered by Lindstrom et al. (24)

Relative growth rate of soma

PENO2 loxP loxPmCherry mCitrine SUC2cyh2r Term Term

Cre recombinase-mediatedgene excision

re

plication conversion

GermUnlimited growth

SomaLimited growth

Secretes invertase

A B

E

F- β-e

stra

diol

+ β-

estr

adio

l

D Bright-field mCherry mCitrine Composite

10 10 μm

C

PENO2 mCitrine SUC2 TermloxP

0 50 100 150 200 250 300[Cycloheximide] (nM)

0.70

0.75

0.80

0.85

0.90

0.95

1.00

0 0.3 1 3 10 30 100 100010-5

10-4

10-3

10-2

10-1

100

[β-estradiol] (nM)

Conversion rate (per generation)?

?

Unicellular Multicellular

Diff

eren

tiate

dU

ndiff

eren

tiate

d Proteolytically-cleaved linker

=

Fig. 1. Engineering yeast as a model for somatic differentiation. (A) Alternative evolutionary trajectories for the evolution of development. Multicellularityand somatic differentiation have separate biological bases and thus probably evolved sequentially. In some clades, the persistence of likely evolutionaryintermediates suggests that multicellularity arose first (solid arrows); no examples of evolution through a unicellular differentiating intermediate (dashedarrows) are known. Somatic cells are colored yellow. (B) Schematic of germ-soma division of labor model engineered in our differentiating strains. (C) Celltype-defining locus in the differentiating strains. Each pair of cell type-specific proteins (mCherry and Cyh2r for germ cells, mCitrine and Suc2 for somatic cells)is initially expressed as a single polypeptide with a ubiquitin linker (red triangles). Cellular deubiquitinating enzymes cleave this linker posttranslationally at itsC terminus (34), allowing the two resulting peptides to localize and function independently: for example, Suc2 enters the secretory pathway, whereasmCitrine remains in the cytoplasm. A transcriptional terminator (Term) blocks expression of the somatic cell proteins in germ cells. Cre recombinase-mediatedgene excision between loxP sites removes the terminator, as well as the germ cell-specific genes. (D) The unicellular differentiating strain (yMEW192) duringgrowth in yeast extract-peptone-dextrose media (YPD) before (Top) or 5 h after addition of 1 μM β-estradiol (Bottom). The white arrow indicates a buddedcell that has recently undergone conversion and contains both mCherry and mCitrine proteins and thus fluoresces in both channels (cf Fig. S1). (E) Conversionrates estimated by flow cytometry during growth in YPD containing β-estradiol. Error bars represent 95% confidence intervals (CIs) of the mean, determinedusing data obtained from three biological replicates. (F) The growth disadvantage of somatic cells relative to germ cells was measured by growing them incoculture in YPD plus cycloheximide and determining the change in the ratio between cell types over multiple rounds of cell division. Error bars represent95% CIs of the mean, determined using data obtained from three biological replicates.

Wahl and Murray PNAS | July 26, 2016 | vol. 113 | no. 30 | 8363

EVOLU

TION

INAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

to be active only in the presence of β-estradiol. Adding β-estradiolto a growing culture induced conversion of germ to somaticcells, apparent as the onset of mCitrine expression and slow lossof mCherry fluorescence by dilution (Fig. 1D, Fig. S1, and MovieS1). As the β-estradiol concentration increased, conversion ratesranged from undetectable levels (<10−3 conversions per cell pergeneration) to ∼0.3 conversions per cell per generation (Fig. 1Eand Fig. S2A). Expression of the codominant, cycloheximide-sensitive WT allele of CYH2 from its native locus permittedcontinued growth following cyh2r excision, but at a reduced ratethat depended on cycloheximide concentration. The growth ratedeficit of somatic cells ranged from undetectable (<1%) tonearly 30% as the cycloheximide concentration increased (Fig.1F and Fig. S2B).The combination of irreversible differentiation and restricted

somatic cell division caused cultures to approach a steady-stateratio between the two cell types over time (Fig. 2A). The steady-state fraction of somatic cells increased with the conversion rateand decreased with the somatic cells’ growth disadvantage, asexpected (Fig. 2B). The steady-state ratio between cell typescould be tuned over four orders of magnitude by altering thecycloheximide and β-estradiol concentrations (Fig. 2B).To investigate the ability of invertase secretion from somatic

cells to support germ cell proliferation, we determined how theculture’s growth rate on sucrose depended on the ratio betweencell types. In the presence of cycloheximide, cultures containingboth cell types at an intermediate ratio grew more quickly insucrose media than cultures of either cell type alone (Fig. 2C),confirming that somatic cells benefit their germ line throughinvertase secretion.Clonal multicellularity arises through the maintenance of

contact between daughter cells following cytokinesis (25). In bud-ding yeast, clonal multicellularity can be produced, either throughengineering (26) or evolution (16, 27), by mutations that preventdegradation of the septum, the specialized part of the cell wall thatholds mother and daughter cells together after their cytoplasmshave been separated by cytokinesis (28). Deletion of CTS1, a chi-tinase gene required for septum degradation, causes formation of“clumps” (groups of daughter cells attached through persistentsepta) that typically contain 4–30 cells during growth in well-mixedliquid medium (Fig. 3A) (29). In the presence of β-estradiol, dif-ferentiating strains that lack CTS1 (Δcts1) produced clumps thatfrequently contained both germ and somatic cells, as shown byfluorescence microscopy (Fig. 3A) and flow cytometry (Fig. 3 B andC). Combining our gene excision-based differentiation system with

CTS1 deletion thus allowed us to produce strains exhibiting all ofthe life strategies needed to compare the evolutionary stability ofunicellular and multicellular differentiation.Somatic cells in unicellular strains can only benefit germ cells by

secreting useful products into a shared medium; nondifferentiatingcheats (e.g., Cre− germ cells) and germ cells in well-mixed mediahave equal access to somatic cell products, but cheats do not paythe reproductive toll of differentiation. We therefore predictedthat cheats would enjoy a fitness advantage over germ cells,allowing them to invade unicellular, differentiating populations(Fig. 4A, Top). In multicellular species, however, significant local

[β-estradiol] (nM)

Frac

tion

of s

omat

ic c

ells

1 10 100 10000.0

0.2

0.4

0.6

0.8

1.025 nM CHX

100 nM CHX300 nM CHX

0 5 10 15 20 25 30 350.0

0.2

0.4

0.6

0.8

1.0

Culture doublings

Frac

tion

of s

omat

ic c

ells

A B C

Fraction of somatic cells

Gro

wth

rate

(dou

blin

gs/h

r.)

Approach to steady-state cell type ratio Fraction of somatic cells at steady-state

0.0 0.2 0.4 0.6 0.8 1.00.00

0.05

0.10

0.15

0.20

0.25

Culture growth rate in 0.5% sucrose mediacontaining 150 nM cycloheximide

Fig. 2. Stable maintenance of both cell types facilitates growth in sucrose + cycloheximide media. (A) Representative time courses of the fraction of somaticcells in cultures of the unicellular differentiating strain (yMEW192 and yMEW192 convertant) initiated at various cell type ratios and passaged in YPDcontaining 10 nM β-estradiol and 600 nM cycloheximide. (B) The steady-state fraction of somatic cells was determined as in A for various cycloheximide andβ-estradiol concentrations. Error bars represent 2 standard deviations, calculated using data obtained from three biological replicates. (C) Dependence ofgrowth rate on fraction of somatic cells in cultures of the unicellular differentiating strain growing in 0.5% sucrose minimal media containing 150 nM cy-cloheximide and β-estradiol concentrations of 0, 0.3,1, 3, 10, and 50 nM, respectively. Error bars represent 95% CIs of the mean, determined using dataobtained from three biological replicates.

76.4%

A Bright-field mCherry mCitrine Composite

10 μm

0.0

0.2

0.4

0.6

0.8

1.0

1 nM 3 nM 10 nM300 pM[β-estradiol]

Frac

tion

of c

lum

ps

Germ only

Both types

Somatic only

0 1 2 3 4 50

2

4

6

8

10

mCherry Fluorescence (A.U.)

mCi

trin

e Fl

uore

scen

ce (A

.U.)

Clump fluorescence distribution Clump composition by cell typeB C

5.4%

18.2%

Fig. 3. Δcts1 strains form multicellular clumps containing both cell types.(A) Representative image of clump size and cell type composition in a well-mixedliquid culture of the Δcts1 multicellular differentiating strain (yMEW208) at asteady-state cell type ratio in sucrose minimal media containing 10 nM β-estra-diol and 150 nM cycloheximide. (B) Representative distribution of clumpmCherry and mCitrine fluorescence for the multicellular differentiating strain(yMEW208) at a steady-state cell type ratio in sucrose minimal media containing10 nM β-estradiol and 150 nM cycloheximide. Gating of clumps by cell typecomposition (lower sector: germ cells only; middle sector: both cell types; topsector: somatic cells only) is shown in red. (C) The fraction of clumps containingone or both cell types was determined as in B for cultures of the multicellulardifferentiating strain (yMEW208) at a steady-state cell type ratio in sucroseminimal media containing 150 nM cycloheximide at the indicated β-estradiolconcentrations. Error bars represent 95% CIs of the mean, determined with dataobtained from six biological replicates.

8364 | www.pnas.org/cgi/doi/10.1073/pnas.1608278113 Wahl and Murray

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

accumulation of somatic cell products (in our experiment, mono-saccharides) within multicellular groups can give differentiatinglineages an advantage over cheats as long as the benefit of betternutrition overcomes the cost of producing slower-replicating, so-matic cells (26). This benefit is present even in well-mixed culturesbecause of the unstirred liquid within and surrounding the clump. Inclonally multicellular species, such as our Δcts1 strain, novel cheatsarising by mutation will eventually be segregated into cheat-onlygroups by cell division and group fragmentation. We hypothesizedthat cheats would then experience reduced access to somatic cellproducts, potentially negating their growth advantage over germ cells(Fig. 4A, Bottom).To test the prediction that cheats could overcome unicellular but

not multicellular differentiation, we introduced cheats into unicel-lular or multicellular differentiating cultures and investigated theirfate. We mixed differentiating cultures expressing a third fluores-cent protein, Cerulean, with cheats that lack this third color andcannot differentiate because they lack the recombinase whose ac-tion gives rise to somatic cells (Cre−). We monitored the relativefrequency of the two strains over a series of growth and dilutioncycles. In sucrose media, cheats invaded unicellular, differentiatingpopulations but were outcompeted in multicellular, differentiatingpopulations (Fig. 4B). The growth advantage of multicellular, dif-ferentiating strains was nullified in monosaccharide-containing

media, where somatic cells should confer no fitness advantage toclumps (Fig. 4B). Moreover, the growth advantage of differentiatingstrains depended strongly on the conversion rate (β-estradiol con-centration): higher conversion rates were advantageous only in themulticellular differentiating case (Fig. 4B), where they increased thefraction of somatic cells overall, as well as the fraction of clumpscontaining at least one somatic cell (Fig. 3C). In unicellular culturesgrown on sucrose, or in cultures grown on glucose (unicellular ormulticellular), increasing the conversion rate increased the growthadvantage of cheats by reducing the growth rate of the germ cellpopulation (Fig. 4B).Our results show that unicellular, soma-producing strains are

evolutionarily unstable to invasion by nondifferentiating cheats. Thisfinding is unlikely to depend on the specific molecular mechanismsthat effect differentiation or allow somatic cells to assist germ cells:any form of differentiation in a single-celled species would requirethat the two cell types exchange resources through a shared medium.From the spontaneous mutation rate in budding yeast [∼ 3 × 10−10

per base pair per cell division (30)] and the size of the recombinasegene, we estimate mutations that inactivate our engineered re-combination system would occur in about one out of every 107 celldivisions; this is likely an underestimate of the frequency of inacti-vating mutations in natural differentiation, which typically requiresmore loci and thus presents a larger target for mutation (22). Be-cause this frequency is high relative to typical microbial populationsizes, cheats would thus likely arise and reach high frequenciesshortly after the appearance of unicellular somatic differentiation,explaining the absence of extant species with this life strategy. Theshort persistence time of unicellular differentiating species makesthem an unlikely intermediate in the evolution of development rel-ative to undifferentiated multicellular species, which have persistedfor hundreds of millions to billions of years in some clades (1–3). Wecannot, however, rule out the possibility that the transient existenceof a unicellular differentiating species might suffice for the secondaryevolution of multicellularity, which has been observed experimentallyin small populations on the timescale of weeks (16, 31). In any case,the differentiating phenotype cannot be stably maintained againstcheats until clonal multicellularity evolves.Our study demonstrates that synthetic biology can directly test

hypotheses about evolutionary transitions, complementing retro-spective inference through comparison of existing species, exper-imental evolution, mathematical modeling, and simulation. Wenote that other major evolutionary transitions, including theappearance of body plans (spatially ordered arrangements of celltypes) and life cycles (temporal sequences of growth and dis-persion), could be studied through experimental evolution orfurther engineering of the strains described above. Furthermore,our differentiating strain provides an experimentally tractableversion of a common simplifying assumption in population ge-netics: independent deleterious mutations are modeled as havingthe same fitness cost (32). In our strains, cyh2r excision is a formof irreversible mutation that always produces the same fitnessdisadvantage, but the magnitude of the fitness cost can be ex-perimentally varied by altering the cycloheximide concentration.Thus, engineered organisms, including those we have developed,can permit robust experimental testing of a wide variety ofoutstanding hypotheses in evolutionary biology.

Materials and MethodsConversion Time Courses and Conversion Rate Assays.We followed the dynamicsof expression of fluorescent proteins during conversion from germ line to somaby adding β-estradiol at the indicated concentration to cultures of pure germcells (yMEW192), which were pregrown in log phase in YPD [yeast extract,peptone, dextrose (glucose)]. Cultures were maintained in log phase for addi-tional growth while samples were collected for analysis by flow cytometry. Atthe indicated time point, cultures were washed twice with PBS to removeβ-estradiol and resuspended in YPD for additional growth before a final flowcytometry analysis. More details of the design and data from this experimentare shown in Fig. S1.

A

B

Δcts1 0.5% sucrose

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

Gro

wth

adv

anta

ge o

f diff

eren

tiatin

g st

rain

Δcts1 0.5% glu/fru

CTS1 0.5% sucrose

CTS1 0.5% glu/fru

300 pM β-estradiol1 nM β-estradiol3 nM β-estradiol

10 nM β-estradiol

Uni

cellu

lar

Mul

ticel

lula

r

Non-differentiating cheatGerm cell Somatic cellLegend:

Fig. 4. Multicellular differentiating strains resist invasion by Cre− cheats.(A) Hypothesized evolutionary outcomes for novel nondifferentiating mu-tants (red) in unicellular (Top) and multicellular (Bottom) populations. Germcells are blue and somatic cells are yellow. (B) Growth advantages of dif-ferentiating strains (unicellular: yMEW192, multicellular: yMEW208) relativeto cheaters (unicellular: yMEW193, multicellular: yMEW209) in minimalmedia containing 150 nM cycloheximide. Error bars represent 95% CIs ofmean growth advantages, determined using data obtained from three bi-ological replicates. Growth advantages of the multicellular strain growing insucrose minimal medium at 1, 3, and 10 nM β-estradiol are significantlygreater than zero (P < 10−3, one-tailed t test); growth advantages of theunicellular strain growing in sucrose are significantly less than zero at allβ-estradiol concentrations (P < 10−6, one-tailed t test).

Wahl and Murray PNAS | July 26, 2016 | vol. 113 | no. 30 | 8365

EVOLU

TION

INAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

Wemeasured the rate of germ line to soma conversion by adding β-estradiolto pure cultures of germ cells (yMEW192) pregrown in log phase in YPD(Fig. S2A). Pure cultures of germ cells were maintained in parallel to permitcalculation of the number of germ cell generations elapsed through cell densitymeasurement with a Coulter Counter (Beckman Coulter). Samples collected atindicated time points were washed twice in PBS solution to remove β-estradioland then resuspended in YPD for continued growth, allowing recently con-verted somatic cells to dilute out remaining mCherry so that cell types could beunambiguously distinguished by flow cytometry. Linear regression of the logfraction of cells that were germ cells vs. the number of germ cell generationselapsed was used to determine a 95% CI of the conversion rate for each ex-periment using the fit and confint functions in MATLAB (Mathworks). Theconversion rates reported in Fig. 1E correspond to the weighted arithmeticmean and corresponding variance calculated with data obtained from three ormore independent experiments. More details of the design and data from thisexperiment are shown in Fig. S2.

Fitness Assays for Relative Growth Difference Between Germ and Somatic Cells.The growth rate difference between cell types was determined using a fitnessassay protocol described previously (33). Briefly, pure cultures of germ(yMEW192) and somatic (yMEW192 convertant) cells pregrown in log phase inYPD containing cycloheximide were combined and maintained in log phasethrough multiple growth and dilution cycles in the indicated media (Fig. S2B).Samples taken at each dilution step were used to determine the fraction ofeach cell type by flow cytometry. Pure cultures of germ (yMEW192) cells weremaintained in log phase in parallel to determine the number of germ cellgenerations elapsed through cell density measurements with a BeckmanCoulter Counter. The 95% CI for the slope of the linear regression line oflog(somatic cells/germ cells) vs. the number of germ cell generations elapsedwas determined as described above. The growth rate differences reported inFig. 1F represent the weighted arithmetic means and corresponding variancesof the slopes calculated in three independent experiments.

Steady-State Ratio Assays. Pure cultures of germ cells (yMEW192) and somaticcells (yMEW192) were mixed to initiate cultures from a range of cell typeratios. Cultures were washed with PBS and resuspended in YPD containingβ-estradiol and cycloheximide at the indicated concentrations. Samples weretaken at the indicated time points to determine the number of culturedoublings since initiation (determined from Coulter Counter measurementsof culture density) and the fraction of somatic (mCitrine+ mCherry−) cells inthe culture. Cultures typically converged on a steady-state ratio after 30–40culture doublings.

Growth Rate Assays in Sucrose Minimal Media. Cultures of converting strains attheir steady-state ratios were washed twice with PBS and resuspended at∼105 cells/mL in minimal media containing 0.5% sucrose and the concen-trations of cycloheximide and β-estradiol in which they had been growingpreviously. Pure cultures of the Cre− strain yMEW193 and the yMEW192convertant were also used to represent pure cultures of germ and somatic

cells of the differentiating strain, respectively. After an acclimation period ofapproximately 10 h, culture density measurements were taken regularly with aCoulter Counter, and a 95% CI of the slope determined by linear regression oflog(culture density) vs. time. The reported growth rates (Fig. 2C) were obtainedfrom the weighted arithmetic means of slopes and their corresponding vari-ances estimated from three or more independent experiments.

Determination of the Distribution of Cell Type Composition in Clumps. Culturesof Δcts1 strains growing in minimal media containing 0.5% sucrose and theindicated concentrations of cycloheximide and β-estradiol were analyzed byflow cytometry. Clumps in these cultures typically ranged in size from 4 to 30clumps; forward and side scatter gating could therefore not be used toeliminate events consisting of two or more clumps. Instead, cultures werediluted and vortexed thoroughly before flow cytometry to reduce theprobability of multiple clumps being counted as a single event. The efficacyof this strategy was checked by combining pure cultures of germ and so-matic cells in the absence of β-estradiol and confirming that the fraction ofmCherry+ mCitrine+ events in the mixed culture was less than 1%. Clumpscontaining only germ cells had negligible fluorescence in the mCitrinechannel and thus formed a distinguishable population; likewise, clumpscontaining only somatic cells had negligible fluorescence in the mCherrychannel. Clumps with nonnegligible fluorescence in both channels wereconsidered to contain both cell types.

Competition Assays. In competition assays, Cerulean+ differentiating strains(unicellular: yMEW192; multicellular: yMEW208) were mixed with Cerulean−

Cre− reference strains (unicellular: yMEW193; multicellular: yMEW209) at a1:1 ratio and passaged through five cycles of growth and dilution in in-dicated media. Samples taken at each dilution step were analyzed by flowcytometry to determine the ratio between Cerulean+ and Cerulean− events(cells or clumps). The 95% CI of the slope of the linear regression line oflog(Cerulean+ events/Cerulean− events) vs. the number of culture doublingselapsed was determined as described above. The growth advantages of thedifferentiating strain reported in Fig. 4B represent the weighted arithmeticmeans and corresponding variances of the slopes calculated in three or moreindependent experiments.

Additional materials and methods are provided as SI Text. Strain geno-types are found in Table S1, plasmid details in Table S2.

ACKNOWLEDGMENTS. We thank Derek Lindstrom and Dan Gottschling forproviding the estradiol-inducible Cre construct PSCW11-cre-EBD78; Alex Schier,Michael Desai, Michael Laub, Cassandra Extavour, David Haig, and members ofthe A.W.M. and David Nelson laboratories for helpful discussions during man-uscript preparation; and Beverly Neugeboren, Linda Kefalas, and Sara Amaralfor research support. This work was supported in part by National ScienceFoundation and Department of Defense National Defense Science and Engi-neering graduate research fellowships and by NIH/National Institute of GeneralMedical Sciences Grant GM068763.

1. Grosberg RK, Strathmann RR (2007) The evolution of multicellularity: A minor majortransition? Annu Rev Ecol Evol Syst 38(1):621–654.

2. Herron MD, Hackett JD, Aylward FO, Michod RE (2009) Triassic origin and early ra-diation of multicellular volvocine algae. Proc Natl Acad Sci USA 106(9):3254–3258.

3. Schirrmeister BE, Antonelli A, Bagheri HC (2011) The origin of multicellularity in cy-anobacteria. BMC Evol Biol 11:45.

4. Kirk DL (2005) A twelve-step program for evolving multicellularity and a division oflabor. BioEssays 27(3):299–310.

5. Bell G, Mooers AO (1997) Size and complexity among multicellular organisms. Biol JLinn Soc Lond 60(3):345–363.

6. Bonner JT (2003) On the origin of differentiation. J Biosci 28(4):523–528.7. Solari CA, Kessler JO, Goldstein RE (2013) A general allometric and life-history model for

cellular differentiation in the transition to multicellularity. Am Nat 181(3):369–380.8. Willensdorfer M (2009) On the evolution of differentiated multicellularity. Evolution

63(2):306–323.9. Hom EF, Murray AW (2014) Plant-fungal ecology. Niche engineering demonstrates a

latent capacity for fungal-algal mutualism. Science 345(6192):94–98.10. Hekstra DR, Leibler S (2012) Contingency and statistical laws in replicate microbial

closed ecosystems. Cell 149(5):1164–1173.11. Bell G (1985) The origin and early evolution of germ cells as illustrated by the vol-

vocales. The Origin and Evolution of Sex (MBL Lectures in Biology), eds Monroy A,Halvorson HO (A. R. Liss, New York), pp 221–256.

12. Buss LW, Mayr Ea (1987) The Evolution of Individuality (Princeton Univ Press, Princeton).13. Goldsby HJ, Knoester DB, Ofria C, Kerr B (2014) The evolutionary origin of somatic

cells under the dirty work hypothesis. PLoS Biol 12(5):e1001858.14. Smith JM, Price GR (1973) The logic of animal conflict. Nature 246(5427):15–18.15. Smith JM (1964) Group selection and kin selection. Nature 201(4924):1145–1147.

16. Ratcliff WC, Denison RF, Borrello M, Travisano M (2012) Experimental evolution ofmulticellularity. Proc Natl Acad Sci USA 109(5):1595–1600.

17. Ratcliff WC, et al. (2013) Experimental evolution of an alternating uni- and multi-cellular life cycle in Chlamydomonas reinhardtii. Nat Commun 4:2742.

18. Gilbert OM, Foster KR, Mehdiabadi NJ, Strassmann JE, Queller DC (2007) High re-latedness maintains multicellular cooperation in a social amoeba by controllingcheater mutants. Proc Natl Acad Sci USA 104(21):8913–8917.

19. Momeni B, Waite AJ, Shou W (2013) Spatial self-organization favors heterotypic co-operation over cheating. eLife 2:e00960.

20. Greig D, Travisano M (2004) The Prisoner’s Dilemma and polymorphism in yeast SUCgenes. Proc Biol Sci 271(Suppl 3):S25–S26.

21. Gore J, Youk H, van Oudenaarden A (2009) Snowdrift game dynamics and facultativecheating in yeast. Nature 459(7244):253–256.

22. Davidson EH (2006) The Regulatory Genome: Gene Regulatory Networks in Developmentand Evolution (Elsevier/Academic Press, Amsterdam).

23. Stöcklein W, Piepersberg W, Böck A (1981) Amino acid replacements in ribosomalprotein YL24 of Saccharomyces cerevisiae causing resistance to cycloheximide. FEBSLett 136(2):265–268.

24. Lindstrom DL, Gottschling DE (2009) The mother enrichment program: A geneticsystem for facile replicative life span analysis in Saccharomyces cerevisiae. Genetics183(2):413–422.

25. Queller DC (2000) Relatedness and the fraternal major transitions. Philos Trans R SocLond B Biol Sci 355(1403):1647–1655.

26. Koschwanez JH, Foster KR, Murray AW (2011) Sucrose utilization in budding yeast asa model for the origin of undifferentiated multicellularity. PLoS Biol 9(8):e1001122.

27. Koschwanez JH, Foster KR, Murray AW (2013) Improved use of a public good selectsfor the evolution of undifferentiated multicellularity. eLife 2:e00367.

8366 | www.pnas.org/cgi/doi/10.1073/pnas.1608278113 Wahl and Murray

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

28. Dohrmann PR, et al. (1992) Parallel pathways of gene regulation: Homologous reg-ulators SWI5 and ACE2 differentially control transcription of HO and chitinase. GenesDev 6(1):93–104.

29. Kuranda MJ, Robbins PW (1991) Chitinase is required for cell separation duringgrowth of Saccharomyces cerevisiae. J Biol Chem 266(29):19758–19767.

30. Lang GI, Murray AW (2008) Estimating the per-base-pair mutation rate in the yeastSaccharomyces cerevisiae. Genetics 178(1):67–82.

31. Boraas M, Seale D, Boxhorn J (1998) Phagotrophy by a flagellate selects for colonialprey: A possible origin of multicellularity. Evol Ecol 12(2):153–164.

32. Haldane JBS (1937) The effect of variation of fitness. Am Nat 71(735):337–349.33. Lang GI, Botstein D, Desai MM (2011) Genetic variation and the fate of beneficial

mutations in asexual populations. Genetics 188(3):647–661.34. Bachmair A, Finley D, Varshavsky A (1986) In vivo half-life of a protein is a function of

its amino-terminal residue. Science 234(4773):179–186.35. Adams A, Cold Spring Harbor L (1997) Methods in Yeast Genetics: a Cold Spring Harbor

Laboratory CourseManual (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY).36. Li MZ, Elledge SJ (2012) SLIC: A method for sequence- and ligation-independent

cloning. Methods Mol Biol 852:51–59.37. Sikorski RS, Hieter P (1989) A system of shuttle vectors and yeast host strains designed

for efficient manipulation of DNA in Saccharomyces cerevisiae. Genetics 122(1):19–27.38. Longtine MS, et al. (1998) Additional modules for versatile and economical PCR-based

gene deletion and modification in Saccharomyces cerevisiae. Yeast 14(10):953–961.39. Rizzo MA, Springer GH, Granada B, Piston DW (2004) An improved cyan fluorescent

protein variant useful for FRET. Nat Biotechnol 22(4):445–449.

40. Sheff MA, Thorn KS (2004) Optimized cassettes for fluorescent protein tagging in

Saccharomyces cerevisiae. Yeast 21(8):661–670.41. Shaner NC, et al. (2004) Improved monomeric red, orange and yellow fluorescent

proteins derived from Discosoma sp. red fluorescent protein. Nat Biotechnol 22(12):

1567–1572.42. Abremski K, Hoess R (1984) Bacteriophage P1 site-specific recombination. Purification

and properties of the Cre recombinase protein. J Biol Chem 259(3):1509–1514.43. Yoshimatsu T, Nagawa F (1989) Control of gene expression by artificial introns in

Saccharomyces cerevisiae. Science 244(4910):1346–1348.44. Orr-Weaver TL, Szostak JW, Rothstein RJ (1981) Yeast transformation: A model system

for the study of recombination. Proc Natl Acad Sci USA 78(10):6354–6358.45. Fried HM, Warner JR (1982) Molecular cloning and analysis of yeast gene for cyclo-

heximide resistance and ribosomal protein L29. Nucleic Acids Res 10(10):3133–3148.46. Hauf J, Zimmermann FK, Müller S (2000) Simultaneous genomic overexpression of

seven glycolytic enzymes in the yeast Saccharomyces cerevisiae. Enzyme Microb

Technol 26(9-10):688–698.47. Voth WP, Olsen AE, Sbia M, Freedman KH, Stillman DJ (2005) ACE2, CBK1, and BUD4

in budding and cell separation. Eukaryot Cell 4(6):1018–1028.48. Wach A, Brachat A, Alberti-Segui C, Rebischung C, Philippsen P (1997) Heterologous

HIS3 marker and GFP reporter modules for PCR-targeting in Saccharomyces cerevisiae.

Yeast 13(11):1065–1075.49. Schindelin J, et al. (2012) Fiji: An open-source platform for biological-image analysis.

Nat Methods 9(7):676–682.

Wahl and Murray PNAS | July 26, 2016 | vol. 113 | no. 30 | 8367

EVOLU

TION

INAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020