Stable public goods cooperation and dynamic social interactions in yeast


R. Craig MacLean, Department of Zoology, University of Oxford, Oxford OX1 3PS, UK.
Tel.: +44 1865 281 264; fax: +44 1865 310 447; e-mail:


Despite long-standing theoretical interest in the evolution of cooperation, empirical data on the evolutionary dynamics of cooperative traits remain limited. Here, we investigate the evolutionary dynamics of a simple public goods cooperative trait, invertase secretion, using a long-term selection experiment in Saccharomyces cerevisiae. We show that average investment in cooperation remains essentially constant over a period of hundreds of generations in viscous populations with high relatedness. Average cooperation remains constant despite transient local selection for high and low levels of cooperation that generate dynamic social interactions. Natural populations of yeast show similar variation in social strategies, which is consistent with the existence of similar selective pressures on public goods cooperation in nature.


There is growing recognition that cooperation between individuals is an important element of the population biology of micro-organisms (Crespi, 2001; Velicer, 2003; West et al., 2006, 2007). Microcosm studies have succeeded in quantifying the costs and benefits of cooperation in microbial systems (Turner & Chao, 1999; Velicer et al., 2000; Rainey & Rainey, 2003; Harrison & Buckling, 2005; MacLean & Gudelj, 2006; Diggle et al., 2007) in testing mechanisms that generate selection for cooperation (Fabrizio et al., 2004; Griffin et al., 2004; Brockhurst et al., 2006, 2007; MacLean & Gudelj, 2006; Brockhurst, 2007; Diggle et al., 2007; Ross-Gillespie et al., 2007), and in studying the de novo evolution of cooperative traits (Rainey & Rainey, 2003; Turner & Chao, 2003; Brockhurst et al., 2006; Fiegna et al., 2006; Kerr et al., 2006; Sandoz et al., 2007). In this paper, we investigate the evolution of a simple cooperative trait of yeast, invertase secretion, using a long-term selection experiment.

The secretion of costly extracellular digestive enzymes that produce public goods provides a simple and elegant example of the evolutionary dilemma of cooperation in micro-organisms (Crespi, 2001; Greig & Travisano, 2004; West et al., 2006). In yeast, the disaccharide sucrose can be transported into the cell slowly and expensively by active transport (Batista et al., 2004), or the enzyme invertase, which is the product of the SUC gene, can be secreted between the cell membrane and the cell wall to catalyse the extracellular hydrolysis of sucrose to form fructose and glucose, monosaccharides which are transported into the cell quickly and cheaply by facilitated diffusion (Dickinson & Schweizer, 2004). The benefit of invertase secretion is clear: populations of invertase-secreting cells have a maximal growth rate on sucrose which is approximately 20% higher than the maximal growth rate of populations of isogenic cells that lack the invertase gene (R. Craig MacLean, unpublished data). This benefit cannot be achieved without a cost to individuals: secreting invertase is associated with a direct fitness cost that reflects the metabolic cost of enzyme secretion (R. Craig MacLean, unpublished data). In an elegant experiment, Greig & Travisano (2004) demonstrated that a cheat strain that refrains from invertase secretion can invade into populations of an invertase-secreting cooperator strain in a density-dependent manner: at a low population density, when interactions between colonies of yeast cells are weak and relatedness is high, the fitness of the cheat mutant is less than that of the cooperator; at a high population density, when interactions between colonies of yeast cells are strong and relatedness is low, the fitness of the cheat mutant is greater than that of the cooperator. Collectively, these results demonstrate that invertase secretion conforms to the basic assumptions of social evolution theory regarding cooperative traits. The availability of detailed data on the genetics and biochemistry of sucrose metabolism (Carlson et al., 1981, 1985; Ye et al., 1999; Batista et al., 2004; Elbing et al., 2004), combined with the ease of genetic manipulation of SUC genes (Greig & Travisano, 2004) makes invertase secretion an excellent model system for studying the evolution of cooperation.

The motivation for this study was to study the evolutionary dynamics of cooperation over hundreds of generations under conditions that have been previously shown to select against complete defection from cooperation over a timescale of 10–20 generations (i.e. low population density and high relatedness). We passaged five replicate populations of yeast on agar plates for 40 growth cycles, which correspond to approximately 800 generations of selection. This method maintains a low population density and limits dispersal to neighbouring sites at each transfer to a fresh agar plate, so that relatedness remains high. To study how cooperation evolves under these conditions, we quantified the mean and genetic variance of invertase secretion over time.

Materials and methods

Strains and culture conditions

Previously described (Greig & Travisano, 2004) isogenic strains SUC2 (a/αleu2/leu2 his5/his5 ura3/ura3 SUC2/SUC2) and suc2, an isogenic diploid strain in which both copies of SUC2 have been replaced by KanMX, were used in this study. The KanMX marker used to replace SUC2 is selectively neutral in the absence of sucrose, as demonstrated by using competition experiments between SUC2 and suc2 under conditions where the expression of invertase is fully repressed (Greig & Travisano, 2004). hoΔ is a SUC2 diploid strain in which both copies of HO have been replaced by the KanMX marker and the met marker has been introduced. Competition medium consisted of sucrose-limited minimal media (2% sucrose, 0.5% Ammonium sulphate, 0.17% Yeast Nitrogen Base, uracil 50 mg L−1, histidine 20 mg L−1 and leucine 50 mg L−1) supplemented with agar (1.6%).

Selection experiment

A mixture of 99%SUC2 and 1%suc2 cells were inoculated onto five sucrose agar plates (total cell number ≈3000 per plate). After 2 days of growth, selection lines were transferred to fresh media by first replica plating onto a fresh sucrose plate with a sterile piece of velveteen. A second sterile velveteen was then used to remove excess cells from the freshly inoculated sucrose plate. It is difficult to precisely estimate the number of cells that are transferred to a fresh agar plate using this method and we suspect that there is considerable variation in efficiency of cell transfer between replicate plates. However, visual observations show that the number of cells transferred using this method must be > 1000, because the number of visible microcolonies on agar plates that are inoculated using this method are too numerous to count. This protocol was continued with three growth cycles per week for a total of 40 transfers to fresh media. Samples of each population were frozen at −80 °C in 30% glycerol (w/v) at regular intervals.

Measuring cheat prevalence

Quantitative PCR was used to determine the relative abundance of the SUC2 and suc2 strains in samples taken from the selection experiment (see below for detailed qPCR methods) at regular intervals up until cycle 15 and on samples taken from cycles 24, 29 and 41. To confirm that this method can accurately measure the relative abundance of these ancestral genotypes, we estimated the relative abundance of SUC2 and suc2 in 10 populations containing the suc2 strain at a frequency between 1% and 99%. The estimated frequency of suc2 by using qPCR in this experiment was essentially identical to the known frequency of suc2 (r2 = 0.992, F1,8 = 1085, < 0.0001).

Invertase assays of clones from selection lines

Pre-assay cultures of each clone were first prepared by overnight growth of isolated colonies in 2 mL of YPD (1% yeast extract, 2% peptone, 2% dextrose) incubated at 30 °C with continuous shaking at 150 rpm. Pre-assay cultures were then spun down and the pellet was resuspended in sterile saline (0.9%). One hundred microlitres of resuspended pellet was then used to inoculate each of two small sucrose agar plates. These plates were incubated for approximately 48 h at 30 °C. Cells were then harvested from plates and washed twice in sterile saline. To assay invertase, 50 μL of washed cells was added to 100 μL of sodium acetate buffer (0.2 m, pH 5.2) and 50 μL of 0.5 m sucrose. This reaction mixture was incubated at 37 °C for 20 min. Reactions were stopped by adding 300 μL of 0.2 m K2HPO4 and incubating in boiling water for 3 min. The amount of glucose formed was then measured by using a glucose peroxidase assay kit (Megazyme, Wicklow, Ireland). Invertase activity was calculated as the amount of glucose formed divided by the incubation time and the number of cells used in the assay, as measured by using optical density. This assay was carried out in blocks and invertase activity of clones from selection lines was standardized relative to the invertase activity of the ancestral SUC2 strain, which was assayed in six replicate cultures in each block. Extensive preliminary assays on 16 pure cultures of suc2 assayed in four independent blocks (two to five cultures per block) revealed that this strain does not produce any detectable invertase activity in this assay (J. Hedge, unpublished data).

SUC2 copy number assays

Genomic DNA was extracted from YPD cultures of individual clones or from whole population samples. SUC2 copy number was determined by using quantitative PCR using primers specific to SUC2 and a single copy gene, HO. This method gave equivalent results to standardizing SUC2 copy number against the total amount of DNA added to qPCR reactions. To confirm that this method can successfully measure SUC2 copy number, we estimated the copy number of 10 mixtures of SUC2 and suc2 cells in which the average copy number of the population varied between 1% and 99%. Observed and expected copy numbers are essentially identical (r2 = 0.993, F1,8 = 1360, P < 0.0001).

Hyper-cooperator re-invasion experiment

A hyper-cooperator strain and the ancestral SUC2 strain were each invaded from an initial frequency of 1% into populations of a neutrally marked SUC2 strain (Baganz et al., 1997), hoΔ, using the same protocol that was used to establish selection lines consisting of a mixture of SUC2 and suc2. Culture conditions were identical to the main selection experiment, except that methionine was added to culture medium at a concentration of 20 mg L−1 to compensate for the met marker in the hoΔ strain. There were four replicates of each invasion experiment and the experiment was carried out for four cycles. The frequency of the invading strains was determined by using quantitative PCR with primers specific to HO, standardized against total genomic DNA as measured by using optical density at a wavelength of 260 nm. Again, preliminary studies revealed that our quantitative PCR-based method for estimating the frequency of HO gives equivalent estimates to measuring the frequency of the strain by colony counting (r2 = 0.87, F1,3 = 12.4, P = 0.035).

Quantitative PCR

DNA for use in quantitative PCR was extracted using a Wizard genomic DNA extraction kit (Promega, Southampton, UK) as per the manufacturer’s instructions. DNA was amplified using SYBR® Green Master Mix (Applied Biosystems International, Warrington, UK) or TaqMan Universal PCR master mix (Applied Biosystems International), depending on whether or not a dual-labelled probe was used in the amplification reaction. Amplification reactions contained each primer at a concentration of 900 nm and a dual labelled probe (where appropriate) at a final concentration of 62.5 nm. Primer and probe sequences are given in Table 1. Amplifications were carried out in an ABI 7000 sequence detection under the following reaction conditions: 10 min at 95 °C followed by 40 cycles or 95 °C for 30 s followed by 60 °C for 30 s. The relative copy number of a particular sequence in a given amplification reaction was determined by comparison with standard curves of DNA extracted from known reference strains.

Table 1.   Oligonucleotides used in this study.
GenotypeTarget sequenceForward primer (5′–3′)Reverse primer (5′–3′)Probe (5′–3′)


Cheater dynamics

At the outset of the experiment, each population contained a neutrally marked obligate cheater mutant (suc2) at a frequency of 1%, so that we could experimentally verify that our experimental set-up imposes at least some selection for cooperation. The obligate cheat mutant steadily declined in frequency between cycles 1 and 15 (Fig. 1) and assays at cycles 24, 29 and 41 failed to detect any evidence of the suc2 strain.

Figure 1.

 Selection against cheats. Plotted points show the mean (±SEM; n = 4) log ratio of cooperators/cheats during the initial phase of the selection experiment. All populations initially contained the suc2 cheater strain at a frequency of 1%. We failed to detect the cheater strain in samples of populations taken at cycles 24, 29 and 41 by using quantitative PCR. This result was confirmed by plating out glycerol stocks of these samples on agar plates containing geneticin, which selects for the suc2 strain.

Evolutionary dynamics of investment in the common good

To investigate the overall effect of selection on cooperation, we assayed the invertase activity of six randomly chosen clones from samples of our selection lines taken at cycles 18, 29 and 41 (Fig. 2). In no case did the average invertase activity of any given population differ significantly from 1, as determined by two-tailed t-tests (Table 2) and we failed to detect any among-population variation in mean cooperation, as determined by using one-way anovas (Table 3). Across all populations, average invertase activity remained more or less constant (Fig. 2; Table 2). The across-population average invertase secretion at cycle 29 was somewhat lower than that of the ancestral SUC2 cooperator (Table 2; mean 0.96, SEM 0.01, n = 5), but this difference does not achieve statistical significance after correcting for multiple comparisons (α = 0.05/3 = 0.017). To assay for diversification in cooperation within populations, we assayed duplicate cultures of clones sampled from cycles 29 and 41. We failed to detect any statistically significant within-population genetic variation in cooperation, as determined by using one-way anovas (Table 4).

Figure 2.

 Evolutionary dynamics of investment in cooperation. Filled symbols show the mean invertase activity of randomly sampled clones sampled taken from selection lines at cycles 18, 29 and 41 (standard errors omitted for clarity). Open symbols show the across-population mean (±SEM; n = 5).

Table 2.   Average cooperation across populations.
CyclePopulationPopulation mean invertase activitySEMd.f.TPGrand meanSEMd.f.tP
  1. Within- and across-population average invertase activities over time are shown. P-values were calculated using a two-tailed t-test.

Table 3.   Among-population variation in cooperation.
CycleAmong-population d.f.Within-population d.f.Among-population MSWithin-population MSFP
  1. Variation in the average invertase activity among populations is shown. P-values were calculated using a one-way anova treating population as a fixed effect.

Table 4.   Within-population genetic variation in cooperation.
CyclePopulationAmong-clone d.f.Within-clone d.f.Among-clone MSWithin-clone MSFP
  1. Within-population genetic variation in invertase activity among randomly chosen clones is shown. P-values were calculated using a one-way anova treating clone as a random effect.


Evolutionary dynamics of SUC2 copy number

Invertase genes (SUC) are among the most polymorphic genes in Saccharomyces cerevisiae: most strains carry a single chromosomal invertase (SUC2), but approximately 8% of strains carry additional telomeric copies of invertase (Naumov et al., 1996). This high degree of polymorphism in natural populations suggests that SUC copy number may have evolved in our experiments. To investigate this possibility, we quantified mean SUC copy number in our selection lines relative to a single copy reference gene, HO (Fig. 3). Average SUC2 copy number was less than one in several populations, as determined by using one-way anovas (Table 5), but only one of these differences remains statistically significant after correcting for multiple comparisons (α = 0.05/15 = 0.0033). Across all populations, mean SUC2 copy number declined during the course of the experiment, as determined by using a linear regression of across-population average copy number against cycle number (slope of copy number against cycle −0.0024; SE 0.000371, F1,2 = 36, P = 0.0226). This tendency is also reflected in the data on the average copy number in individual populations: by the end of the experiment, average copy number was less than one in four of five populations (median copy number 0.91).

Figure 3.

 Evolutionary dynamics of the SUC2 gene. Filled symbols show the estimated mean SUC2 copy number of individual selection lines at different time points. Open symbols show the across-population average SUC2 copy number (±SEM; n = 5).

Table 5.   Population-level SUC2 copy number.
CyclePopulationHOSUC2SEPopulation SUC2 copy numberd.f.FPMean copy numberSEMd.f.tP
  1. Relative abundance of SUC2 and HO genes in DNA samples taken from populations at cycles 24, 29 and 41 is shown. Copy number was estimated as SUC2/HO and the standard error of this estimate comes from a pooled error variance. Where the sample size of the number of estimates of HO and SUC2 abundance was uneven, a pooled error variance could not be calculated. We tested for a deviation from a population copy number of one by using one-way anovas, treating gene (HO or SUC2) as a fixed effect. We tested for deviations from an across-population mean copy number of one using two-tailed t-tests.

46.245.06 0.815300.003     
532.529.5 0.9150.200.67     

The ascendance of hyper-cooperators

During the course of the selection experiment, we observed the formation of conspicuous high-density clusters of cells in our selection lines that persisted over multiple growth cycles (Fig 4a). To investigate the genetic composition of these clusters, we assayed the invertase activity of genotypes isolated from high-density clusters in three separate selection lines. Genotypes from high-density clusters tend to have a ‘hyper-cooperator’ phenotype characterized by elevated invertase secretion (Fig. 4b; F1,35 = 64, < 0.0001). To gain insight into the evolutionary dynamics of these groups of hyper-cooperators, we longitudinally sampled one very conspicuous high-density cluster of cells that first became apparent at around cycle 20 (Fig. 4c). The average relative invertase activity of genotypes isolated from this cluster was significantly greater than one (mean 1.19, SEM 0.025, t23 = 9.41, < 0.0001), but remained constant over time (F3,20 = 0.52, P = 0.67). We quantified the copy number of SUC2 in six hyper-cooperator mutants, but found no evidence for SUC2 duplication (not shown).

Figure 4.

 The ascendance of ‘hyper-cooperators’. Panel a shows a photograph of one of our populations taken at cycle 36. The conspicuous high-density cluster of cells is shown with an arrow. Bars in panel b show the mean (±SEM; n = 6) invertase activity of clones sampled from high-density clusters (black bars) and randomly sampled clones (grey bars) from three populations sampled at cycle 41. Panel c shows the invertase activity of clones sampled from the cluster of cells shown in panel a at multiple time points. Filled symbols show the invertase data for individual clones and open symbols show the mean invertase activity (±SEM; n = 6). Bars in panel d show the mean frequency (±SEM; n = 4) of a hyper-cooperator mutant and a marked SUC2 control strain (hoΔ) after four cycles of invasion into populations of ancestral SUC2. All populations initially contained the invader strain at a frequency of 0.01.

Fitness consequences of elevated invertase secretion

What are the fitness consequences of elevated invertase secretion? To address this question, we investigated the ability of a hyper-cooperator genotype (relative invertase activity 1.17; SE 0.038; t4 = 4.42, P = 0.012) and a marked SUC2 strain (hoΔ) to invade into populations of the ancestral SUC2 strain from an initial frequency of ≈1% under the same conditions that existed during the selection experiment. In the marked strain (hoΔ), a well-characterized gene with no known function other than mating type switching (HO) has been replaced by the KanMX marker (Baganz et al., 1997). This marker is neutral under the conditions of our experiment, as demonstrated by the lack of change in the frequency of the marked strain after four cycles of competition (Fig. 4d; mean frequency 0.6%, SE 0.3%, t3 = 1.1, P = 0.37). The hyper-cooperator strain failed to invade from rare, as demonstrated by the lack of difference in the frequency of the hyper-cooperator strain, relative to the marked strain, after four cycles of competition (Fig. 4d; F1,6 = 0.5, P = 0.82).


Despite long-standing theoretical interest in the evolution of cooperation (Hamilton, 1963, 1964; Axelrod & Hamilton, 1981; Bull & Rice, 1991; Maynard Smith & Száthmary, 1995; Taylor & Frank, 1996; Hauert & Doebeli, 2004; Sachs et al., 2004; Nowak, 2006; West et al., 2006), this work presents one of the first long-term experimental investigations of this topic (see also Kerr et al., 2006; Sandoz et al., 2007). The most novel aspect of our approach is that we have followed the long-term evolutionary dynamics of cooperation at both the level of individual genotypes (Figs 1 and 4) and populations as a whole (Figs 2 and 3). Our results clearly show that average investment in cooperation at the population level does not change on a timescale of hundreds of generations in this system (Fig. 3). This result was not driven by a lack of genetic variation in our selection lines; defector genotypes and hyper-cooperator genotypes were present in our populations, but these failed to effectively invade from rare, directly demonstrating selection for an intermediate optimal investment in public goods cooperation.

Dynamic social interactions

Two lines of evidence support the idea that selection on cooperation in this system varies at a very fine grained spatial scale. First, hyper-cooperator genotypes repeatedly invaded into our selection lines, but we failed to detect any general advantage associated with this social strategy in competition experiments against a marked ancestral cooperator. The simplest explanation for these seemingly contradictory results is that the benefits of elevated cooperation were restricted to the particular spatial locations where hyper-cooperators invaded into our populations. Secondly, we found that the average SUC2 copy number declined very gradually throughout the course of the experiment, but a marked suc2 cheat that we introduced into our populations at the outset of the experiment failed to invade from rare. The marked cheat was introduced at an initial frequency of only 1% (approximately 30 colonies) and the simplest explanation to account for these results is that we failed to find any evidence of selection for the obligate suc2 cheat because we measured selection on this genotype over a limited area (1% of the total area occupied by the population). The net result of spatial variation in selection for cooperation is the emergence of dynamic social interactions that maintain a more-or-less constant mean investment in cooperation at the scale of the population despite strong local variation in cooperation.

Previous studies (Greig & Travisano, 2004) have shown that the fitness of invertase cheats increases with population density, because of increased availability of glucose produced by cooperators in dense populations. Selection for hyper-cooperation drives a local increase in population density (i.e. Fig. 4a,b), which creates ecological conditions that favour cheating, suggesting that spatial variation in selection for hyper-cooperators and cheats are linked phenomena and that diversity is maintained by nontransitive local interactions. However, the data presented in this manuscript do not allow for a rigorous test of this idea and to carry out such experiments we would ideally need to be able to measure the distribution of different social strategies over very small spatial scales using sophisticated imaging techniques. This is clearly beyond the scope of this work, but future experiments to quantify local, rather than population-level average, interactions between social strategies could make an important contribution to our understanding of social evolution.

Comparison with natural populations

To what extent are the patterns of variation found in nature consistent with our findings? A survey of 91 strains of S. cerevisiae isolated from natural and industrial sources found that 12% of strains do not secrete invertase because they carry a degenerate suc pseudogene, 80% of strains carry a single active SUC gene and 8% of strains carry at least two SUC genes (Naumov et al., 1996). If we assume that strains of S. cerevisiae with more than one copy of SUC have elevated invertase secretion, these patterns of genetic variation are consistent with selection for an intermediate optimal investment in invertase secretion. However, the genetic basis of variation in invertase secretion clearly differs between our experimental populations and nature; none of the hyper-cooperator strains that we isolated has a duplicate SUC gene. It might be argued that variation in invertase secretion in S. cerevisiae does not reflect the outcome of natural selection, because some of these strains may have been artificially selected for high or low invertase secretion. In Saccharomyces paradoxus, an undomesticated congener of S. cerevisiae, we have found that 90% (n = 27) of strains isolated from oak trees in southern England secrete similar quantities of invertase, whereas the remaining 10% (n = 3) of strains refrain from invertase secretion (J. Hedge and R.C. MacLean, unpublished data). Although it is important to emphasize that alternative selective pressures could give rise to the patterns of variation in cooperation found in natural populations of yeast, the diversity of social strategies found in natural populations is nonetheless consistent with the results of this study.


This work was funded by grants from NERC (UK) to the Centre for Population Biology. We thank D. Greig for strains and two anonymous reviewers for helpful comments on a previous version of this manuscript.