Cooperation underlies diverse phenomena including the origins of multicellular life, human behaviour in economic markets and the mechanisms by which pathogenic bacteria cause disease. Experiments with microorganisms have advanced our understanding of how, when and why cooperation evolves, but the extent to which microbial cooperation can recapitulate aspects of animal behaviour is debated. For instance, understanding the evolution of behavioural response rules (how should one individual respond to another's decision to cooperate or defect?) is a key part of social evolution theory, but the possible existence of such rules in social microbes has not been explored. In one specific context (biparental care in animals), cooperation is maintained if individuals respond to a partner's defection by increasing their own investment into cooperation, but not so much that this fully compensates for the defector's lack of investment. This is termed ‘partial compensation’. Here, I show that partial compensation for the presence of noncooperating ‘cheats’ is also observed in a microbial social behaviour: the cooperative production of iron-scavenging siderophores by the bacterium Pseudomonas aeruginosa. A period of evolution in the presence of cheats maintains this response, whereas evolution in the absence of cheats leads to a loss of compensatory behaviour. These results demonstrate (i) the remarkable flexibility of bacterial social behaviour, (ii) the potential generality of partial compensation as a social response rule and (iii) the need for mathematical models to explore the evolution of response rules in multi-player social interactions.
Cooperation is a crucial foundation of the behavioural and ecological diversity we see in nature (Dugatkin, 1997; Székely et al., 2010). In fact, many of the major transitions in evolution (Maynard Smith & Szathmary, 1995) – from single replicator molecules to genomes, from unicells to multicellular organisms and from solitary individuals to societies – depend fundamentally on cooperation between individual entities. Understanding cooperation also has practical significance for the fields of conservation (Banks et al., 2007; Blumstein, 2010) economics (Bowles, 2008), and clinical microbiology (Foster, 2005; Buckling et al., 2007; Diggle, 2010). For instance, cooperative production of iron-scavenging siderophores by the opportunistic human pathogen Pseudomonas aeruginosa (Fig. S1) is a key determinant of virulence and the links between cooperation and virulence in this bacterium have been extensively studied (reviewed in (Buckling et al., 2007; Harrison & Buckling, 2009a)).
The ability to manipulate population structure and environment and to study evolution in real time, with replication, means that microbes have been widely used in empirical tests of evolutionary theory (Buckling et al., 2009). Social evolution theory is no exception, because many microbes are surprisingly social – as well as cooperating to produce shared nutrient-scavenging molecules, microbes can form multicellular structures such as biofilm, communicate with one another in a process called quorum sensing, move in coordinated swarms to capture prey and even exhibit reproductive cooperation or altruistic ‘suicide’ [reviewed in (Velicer, 2003; West et al., 2006; Foster, 2010)]. Bacteria, social amoebae, yeast and bacteriophage have all been employed to test key hypotheses about the evolution of cooperation – mainly those rooted in the dominant paradigm of inclusive fitness theory (Hamilton, 1964; Frank, 1998). Microbial model systems offer great opportunities to investigate the effect of relatedness (Turner & Chao, 1999; Griffin et al., 2004; Mehdiabadi et al., 2006; Diggle et al., 2007; Gilbert et al., 2007; Rumbaugh et al., 2009), the scale of competition (Griffin et al., 2004; Brockhurst et al., 2006; MacLean & Gudelj, 2006), dispersal regime (Kerr et al., 2006; Kümmerli et al., 2009a), population viscosity (Kümmerli et al., 2009b) and the costs and benefits of cooperation and cheating (Brockhurst et al., 2008; Jiricny et al., 2010). Furthermore, microbial studies have allowed researchers to address questions such as the existence of facultative vs. fixed cheating strategies (Santorelli et al., 2008; Gore et al., 2009) and the genetic factors that influence the evolution of cooperation, such as pleiotropy (Foster et al., 2004; Harrison & Buckling, 2009b) and mutation rate (Harrison & Buckling, 2005).
Behavioural ecologists may justifiably ask, however, how well microbes can represent the types of cooperation present in animals. Often, this reflects a concern that expression of traits by bacteria is not as dynamic and responsive to external social stimuli as is animal behaviour. On the one hand, it remains clear that bacterial gene expression is tightly regulated in response to internal and external cues – siderophore expression, for example, responds to intracellular iron levels and to population density (Ratledge & Dover, 2000; Visca et al., 2002; Tiburzi et al., 2008; Kümmerli et al., 2009c) and cells can monitor siderophore availability in their growth medium and adjust their production accordingly (Kümmerli & Brown, 2010). On the other hand, microbial study has not addressed a question central to the study of cooperator-cheat interactions, namely how do cooperative individuals respond to the presence of cheats, and what is the optimal response to cheats as favoured by selection? Only one study (Kümmerli et al., 2009c; described below) has attempted to measure how cooperative (siderophore-producing) bacteria react to the presence of cheating (siderophore-negative) cells.
The question of how individuals should respond to cheating appears in two main forms in the literature on cooperation. First, much of the literature based on game theory explores labile response strategies in repeated interactions [reciprocity and tit-for-tat, and their variants: (Camerer, 2003)]. Second, theoreticians and empiricists have studied how partners in a cooperative task should respond to changes in each other's effort over the course of the task, that is, how does one individual respond to its partner ‘slacking off’ and contributing less to the common good, and what response rule represents an evolutionarily stable strategy (ESS)? The subset of the cooperation literature where this question has been developed and addressed in detail is that which deals with cooperation between parents over offspring care. Game theoretical models have shown that cooperation over care (as opposed to one individual ceasing to care and leaving its partner holding the babies) is an ESS if a decrease in care by one parent leads the other to increase its own care effort, but not so much that it completely compensates for the lost care; this result holds regardless of whether we consider responses over evolutionary (Houston & Davies, 1985) or behavioural (McNamara et al., 1999, 2003) time. Such ‘partial compensation’ has been found to be the rule in a meta-analysis of responses to loss of partner effort in birds (Harrison et al., 2009).
Two published studies have reported that wild-type P. aeruginosa cells increase their levels of siderophore expression on exposure to either a cheating mutant (Kümmerli et al., 2009c) or a second species (Harrison et al., 2008). However, both of these studies compared levels of siderophore production in populations that had undergone a period of exponential growth, from a very small starting density, in the presence or absence the cheat or competitor. Therefore, in these studies any physiological effect of cheat presence on siderophore expression by cooperator cells is likely confounded with the effects of exponential growth on siderophore expression. This is a key point, because growth phase and population density are both known to affect siderophore expression (Kümmerli et al., 2009c); furthermore, natural populations of cells in the environment or in established infections likely do not grow exponentially. Therefore, it would be more informative to study responses to cheats in populations that have reached stationary phase. Although there is some turnover of cells in such populations, this approach should give an approximate reflection of how the average cell responds physiologically to its neighbours' investment in siderophores.
I grew wild-type (cooperator) P. aeruginosa populations, and populations of an isogenic mutant that does not produce the primary siderophore pyoverdine (cheat) to stationary phase in liquid medium. I then separated the cells in each population from their growth medium (containing pyoverdine) by centrifuging the cultures and removing the liquid supernatant from the pelleted cells. Cooperator and cheat cells were then re-suspended in fresh medium, mixed in varying ratios and incubated overnight to allow the cooperator cells to start producing pyoverdine again. As these populations were already at carrying capacity, population size did not change in this ‘recovery’ period; therefore, comparing measurements of pyoverdine production taken after recovery in pure and mixed populations should give a reliable indication of how cooperator cells respond to the presence of cheat cells. This showed that wild type, cooperative P. aeruginosa cells show partial compensation for the presence of cheats: they increase their own production of siderophores, but this does not completely compensate for the presence of cheats. I then demonstrated that this response is preserved in a lab population that had undergone a period of coevolution with cheating clones, but was lost in a lab population that was evolved in the absence of cheats.
These results confirm that bacterial social behaviour is dynamic and flexible, and are consistent with the hypothesis that partial compensation for cheating could be important in a variety of social contexts. P. aeruginosa siderophores could therefore represent a useful model system for testing hypotheses about the evolution of behavioural response rules. The role of partial compensation, and of response rules in general, in maintaining cooperative behaviours in various contexts is ripe for wider theoretical and empirical exploration. In particular, responses to cheating in multi-player (as opposed to dyadic) interactions would benefit from further consideration.
Materials and methods
Strains and growth conditions
Pseudomonas aeruginosa strain PAO1 (ATCC 15 692) was used as the wild type, siderophore-producing strain (cooperator) and an isogenic deletion mutant that lacks the pyoverdine biosynthetic locus pvdD was used as the nonproducing cheat [PAO1∆pvdD (Ghysels et al., 2004)]. All experiments were conducted in casamino acids broth (CAA; 5 g casamino acids, 1.18 g K2HPO4*3H2O, 0.25 g MgSO4*7H2O, per litre), which represents an iron-poor environment. CAA can be made strictly iron-limited by the addition of 70 μg mL−1 human apotransferrin (Sigma) and 20 mM NaHCO3. Apotransferrin binds free Fe(III), preventing nonsiderophore-mediated uptake of iron by bacterial cells; this binding is dependent on the presence of bicarbonate anions. All bacterial growth took place at 37 °C; 2-mL cultures were incubated with continuous shaking to aerate cultures while microplate cultures were incubated statically. Preliminary work confirmed that PAO1∆pvdD does indeed cheat on the wild type, as expected given work on other pairs of wild type : cheat clones in this species (Griffin et al., 2004; Harrison et al., 2006). After 24 h' growth in 1.5-mL iron-limited CAA, monocultures of the wild type achieved an average density of c. 9 × 108 cells, while monocultures of the pyoverdine mutant achieved a much lower average density of c. 3 × 108 cells (n = 5 in each case, t8 = 8.05, P < 0.01), demonstrating the benefit of pyoverdine production in this medium. In pure cultures, the average fitness of the mutant relative to the wild type (calculated for randomly paired mutant and wild-type cultures) was 0.30 ± 0.029, whereas in mixed cultures containing approximately 1/3 mutants the average mutant relative fitness was 1.36 ± 0.197. (The fitness v (Jiricny et al., 2010) reflects whether the mutant increases in frequency relative to the wild type, in which case v > 1, or decreases, in which case v < 1; if the relative frequencies do not change then v = 1). This demonstrates that in mixed culture, the benefit of pyoverdine production is shared with and exploited by nonproducing cells.
Pyoverdine production assay
Pyoverdine production of bacterial cultures was assayed by a pyoverdine-specific excitation-emission assay: the relative fluorescence of 100 μL of culture at 460 nm following excitation at 400 nm was measured using a SpectraMax M2 spectrophotometer (Ankenbauer et al., 1985; Cox & Adams, 1985; Jiricny et al., 2010).
Production of evolved and coevolved cooperator clones
I wished to evolve the cooperator alone (evolution treatment) and in the presence of the cheat (coevolution treatment). Following the protocol of Griffin et al. (2004), experimental metapopulations of cooperators (evolution treatment) and cooperators + cheats (coevolution treatment) were founded and allowed to evolve for approx. Eighty generations under a regime of low relatedness and global (inter-deme) competition. Metapopulations were four wells of a 24-well microplate, each containing 2 mL CAA + apotransferrin and NaHCO3 as above. These conditions should select for the maintenance of both cooperators and cheats in the coevolution treatment over evolutionary time (Griffin et al., 2004). A schematic of the selection regimes is provided as electronic (Fig. S2). At the end of the evolution period, individual cooperator clones were isolated from each metapopulation by plating on CAA agar and selecting yellow colonies; these were stored in glycerol at −80 °C for later use. Fourteen clones from one evolved metapopulation and fourteen clones from one coevolved metapopulation were used for the experiments reported.
Response of cooperators to presence of cheat cells
I wished to determine the response of cooperator populations to the presence of cheating, in the absence of any significant population growth. Therefore, I designed a method that measures pyoverdine response by stationary-phase cells. The ancestral cooperator and cheat clones were grown for 24 h in CAA, by which time stationary phase will have been achieved. The CAA was not made further iron-limited to allow cheat populations to reach densities comparable with that of the cooperator (both reached an OD600 of ~ 1.0). Cultures were centrifuged at 5000 g for 10 min to pellet the cells and the supernatants (containing siderophores) discarded. Each culture was re-suspended in fresh CAA made iron-limited by the addition of apotransferrin and NaHCO3 as above. Aliquots of the cultures were then mixed to produce mixtures comprising 100%, 70%, 55%, 38% and 0% cooperator cells. An aliquot of each mixture was diluted and plated on CAA agar to provide confirmation of starting density and per cent cooperator cells. Ten 200-μL aliquots of each mixture were then placed in wells of a 96-well microplate and incubated for 18 h at 37 °C. (Preliminary work indicated that this amount of time allowed pyoverdine production to reach a stable level but that growth of cells in this period was negligible, that is, this time period should give a sufficiently accurate reflection of the physiological response of cells, rather than responses due to population growth or diminution). Following the 18-h ‘recovery’ period, each population was shaken for 15 s, assayed for OD600 and pyoverdine production and an aliquot of each population diluted and plated on CAA agar to confirm that density and per cent cooperator cells had not changed during the recovery period. To account for slight variation in OD600 between replicates, pyoverdine production was standardized by dividing by OD600. This procedure was repeated for the coevolved and evolved cooperator clones using 100%, 50% and 0% cooperator cell treatments and three-fold replication. As a control, aliquots of stationary-phase cooperator populations were also re-suspended in iron-limited CAA at 70% and 55% of the initial density, divided into ten 200 μL aliquots, incubated for 18 h at 37 °C and then assayed for OD600 and pyoverdine (PVD) production and an aliquot diluted and plated out. For the comparison of pure vs. mixed cultures, fluorescence data were standardized by dividing by the number of cooperator cells present (as estimated by plating), rather than by OD.
The results for cooperator response to the presence of cheat cells were unchanged by the elimination of one outlier from the data set. Data were analysed using Minitab 15 and R 2.14.0 (http://www.r-project.org; the car package (Fox & Weisberg, 2011) was used to conduct anova using Type II sums of squares).
Cooperators partially compensate for siderophore loss
I first assessed the response of cooperator cells to the presence of cheat cells, in the absence of population growth. Cells of the ancestral, wild-type cooperator clone (PAO1 WT) and a cheat clone (the isogenic pyoverdine knockout, PAO1 ∆pvdD) were recovered from stationary-phase populations, stripped of all exoproducts (including siderophores) by centrifuging and removing culture supernatants and re-suspended in fresh iron-limited medium to make mixed populations with varying frequencies of each clone. Pure cooperator and cheat populations were also created. Each population was divided into ten replica populations and these were incubated overnight to allow ‘recovery’ of pyoverdine production by cooperator cells. Each population was then assayed for pyoverdine production. Cell density and proportion of cooperator cells maintained constant over the recovery period (data not shown). As shown in Fig. 1 (white bars), pyoverdine production by the pure cheat populations was approximately 1% of that of the pure cooperator populations. In the mixtures (Fig. 1, grey bars), pyoverdine signal was positively correlated with the percentage of cooperator cells in the population (anova: F2,27 = 495.03, P < 0.001). Under the null hypothesis of no physiological response to the presence of cheats, the expected pyoverdine signal in each of the mixed populations can be predicted from the pyoverdine signals in the pure populations (Fig. 1, stars). In each cooperator/cheat mixture, the pyoverdine signal was significantly higher than expected under this null hypothesis, but significantly less than that of the pure cooperator population [t-tests, P < 0.001; all remained significant after correction for false discovery rate (Benjamini & Hochberg, 1995)]. Therefore, averaged across the population, cooperator cells upregulate pyoverdine production in the presence of cheats but do not fully compensate for cheat presence.
It is possible that the inability of cooperators to partially compensate for loss to cheats is simply the result of resource limitation, that is, cells up-regulate pyoverdine production as much as possible, but the maximum recovery is constrained by available resources and hits a ceiling when cells become exhausted. However, this hypothesis would predict a similar proportional increase in pyoverdine production in mixed populations regardless of the relative frequencies of cooperators and cheats. Figure 1 clearly shows that cooperators approximately double their pyoverdine expression when at a frequency of 38% but only increase it by approximately 25% when at a frequency of 55% and approximately 10% when at a frequency of 70%. If the increase of 100% at low frequency represents a maximal response constrained by exhaustion, then we would predict a similar proportional increase and full compensation for cheating loss at the two higher frequencies. The experimental results are therefore not consistent with the hypothesis that the pattern of partial compensation is simply a constraint imposed by resource limitation.
It is possible that the observed pattern has nothing to do with the presence of cheats per se, but is simply a function of reduced cooperator cell numbers. To test this, the 70% and 55% treatments were repeated without adding cheat cells, that is, stationary-phase populations were re-suspended in fresh iron-limited medium at 70% and 55% carrying capacity, allowed to recover for 18 hours and then assayed for cell density and pyoverdine production. Analysis of total pyoverdine in these cultures showed that re-suspended cells increased their level of production but that total pyoverdine was still less than in cultures re-suspended at 100% carrying capacity (data not shown; 38% treatment was not carried out as populations diluted to this density started to grow in the 18 h recovery period). However, comparison of pyoverdine production per cooperator cell in these cultures with pyoverdine production per cooperator cell in mixed populations showed that individual cells produce more pyoverdine when cheat cells are present and increase the total population size to carrying capacity. (anova: no main effect of cooperator density F1,36 = 1.89, P = 0.18; significant main effect of cheat presence/absence F1,36 = 28.63, P < 0.001; significant interaction F1,36 = 4.68, P = 0.037; see Fig. S3). Therefore, while cells up-regulate pyoverdine production in response to a reduction in their numbers, they up-regulate it even more when cheats are also present and act as a sink for pyoverdine, possibly by an integrated response to both local cell density and rate of ferrisiderophore return.
Coevolution with cheats maintains partial compensation
Partial as opposed to full compensation may represent an adaptation (or preadaptation) to the presence of cheats [as is the case in dyadic games: (McNamara et al., 1999, 2003)]. If this is the case, then a period of coevolution with cheats might reinforce this response, whereas a period of evolution in the absence of cheats might lead to no or full compensation. To test this hypothesis, I initiated metapopulations with a 50/50 mix of cooperator and cheat cells in iron-limited medium and evolved these under conditions of low relatedness and inter-deme (global) competition as described by Griffin et al. (2004) (coevolution treatment). In Griffin et al.'s experiments, this treatment resulted in the maintenance of mixed cooperator/cheat populations. I also initiated and evolved pure cooperator metapopulations under the same conditions, to provide a control for any effects of adaptation to laboratory conditions (evolution treatment). Figure S2 gives a schematic of the experimental protocol. After c. 80 generations of evolution, I isolated fourteen individual pyoverdine-producing clones from one coevolved and one evolved population. These were assayed for pyoverdine production after re-suspension of stationary-phase populations alone and in a c. 50/50 mixture with cells of the ancestral cheating clone exactly as described for the ancestral cooperator, above. Therefore, this experiment did not address any possible effects of coevolution between cooperators and cheats (Zhang et al., 2009). Consistent with previous observations (Harrison & Buckling, 2005), cooperator clones isolated from coevolved populations on average produced significantly less pyoverdine in pure culture than did cooperator clones isolated from evolved populations (approximately 70%; t26 = 5.87, P < 0.001).
Figure 2 shows the raw data for pyoverdine production by each of the 28 clones in pure culture, actual pyoverdine production in a c .50/50 mix with the cheat and predicted pyoverdine production in mixtures under the null hypothesis of no response to cheats. It can be seen that pyoverdine production in evolved clone + cheat mixtures is very close to that predicted under no response, while in coevolved clone + cheat mixtures, responses are generally positive but do not fully compensate for cheat presence (although there is apparently one case of no response and two possible cases of full compensation). To analyse these data, while removing noise due to differences between clones in the overall levels of pyoverdine production, I calculated for each clone the amount of pyoverdine measured in mixed culture as a percentage of that expected if cells do not respond at all to the presence of the cheat. Therefore, a value of 100% represents no physiological response, a value of 200% would represent full compensation in a 50/50 mix, values between 100% and 200% represent partial compensation and values < 100% represent downregulation of pyoverdine in response to cheats. Mean values from three replica populations of each clone were used in the analysis; these data are summarized in Fig. 3. I used anova to test for an effect of evolution (no exposure to cheats) vs. coevolution (continued exposure to cheats) on the per cent predicted pyoverdine production. To eliminate any effects of slight variation from a 50/50 ratio, the measured frequency of cooperators in the population was included as a covariate; I also included the mean amount of pyoverdine produced by the clones in pure populations as this also varied. The results are shown in Table 1. On average, clones from the coevolution treatment showed a greater response to cheats than did those that from the evolution treatment (F1,21 = 9.00, P = 0.007). This difference came because of a change in behaviour following evolution in the absence of cheats: the partial compensation response of coevolved clones was not significantly different from that of the ancestor in c. 50/50 mixtures with cheats (t-test against ancestral mean of 137%, shown as dotted line in Fig. 3; t13 = 1.02, P = 0.328), but on average clones that had evolved in the absence of cheats showed no response to cheats (t-test against hypothesized mean of 100%; t13 = 0.26, P = 0.797). This is consistent with the hypothesis that the presence of cheats drives the maintenance of a partial cooperation response, though further work with replicated selection lines is required to confirm this finding.
Table 1. Analysis of variance results for per cent predicted pyoverdine (PVD) production
Sum of squares
Clones from an evolved population showed significantly less PVD response to the presence of cheats than did clones from a population that had coevolved with cheats. This analysis eliminated variance due to slight variations in cooperator frequency under test conditions and the amount of PVD produced by the clones when in pure culture. Although there was no main effect of the level of PVD produced in pure cultures, there was a significant interaction of this variable with treatment; this was because in the evolution treatment there was a negative correlation between PVD production in pure culture and per cent predicted PVD in mixed culture (see Fig. S1).
PVD production in pure culture
Treatment*PVD production in pure culture
Cooperator frequency*PVD production in pure culture
These results clearly confirm that bacterial social behaviour is dynamic and flexible, and that coevolution with cheats maintains a response of partial compensation. Although this study focuses on stationary-phase cells to disentangle bacterial ‘behaviour’ from the effects of population growth, Kümmerli et al. (2009c) work on plasticity of siderophore production in growing populations of cooperators and cheats produced remarkably similar results: while the question of compensation is not explicitly addressed in this study, the data clearly show partial compensation [(Kümmerli et al., 2009c), Fig. 4a]. Further, data collected by these authors suggest that cooperator cells compensate more when fewer siderophores are required to support growth (when iron is less limiting: R. Kümmerli, personal communication).
This study provides results that are consistent with the hypotheses that P. aeruginosa cells sense and respond to the presence of cheats, and that a response of partial compensation is maintained through coevolution with cheats. This result could be seen as consistent with the hypothesis that partial compensation is an adaptive response to cheats. However, the data do not incontrovertibly demonstrate this. First, it is difficult to separate the effects of cheat presence from the effects of simple siderophore dilution; future empirical studies could explicitly compare the evolution of responses to cheats in populations evolved with and without cheats in high or low diffusion environments, in a cross-factored design. In addition, to determine whether cooperators respond specifically to the presence of cheats as well as more generally to siderophore dilution, microbiologists could attempt to determine whether cheat cells release different chemical signals or cues, as compared with cooperators (aside from the lack of pyoverdine). It should be noted, however, that existing models showing the role of partial compensation in maintaining cooperation make no assumptions about the mechanism of the partial compensation response; indeed, they do not even assume that direct observation of or social cues from the cheat are necessary. Partial compensation could have evolved as a response to other, perhaps nonsocial cues – for example, offspring state in the parental care context and siderophore loss in environments with high loss to diffusion in the siderophore case. Whatever the mechanism for, or early adaptive significance of, this response, the end result is the same: cooperators respond in a manner that is predicted to help to make cooperation an ESS in models that consider dyadic interactions. (That there is not always a clear-cut dichotomy between social and nonsocial cues or behaviours is nicely underlined by the current debate surrounding bacterial quorum sensing: see discussion by West et al., 2012 and empirical results from, for example, Yang et al., 2010; Dandekar et al., 2012; Darch et al., 2012). Second, explicit experimental investigation of the fitness consequences of different response rules (partial compensation vs. full or no compensation) when in competition with cheats should now be conducted in order to explore whether partial compensation leads to higher cooperator relative fitness and the evolutionary stability of cooperation. Finally and most fundamentally, existing theory based on dyadic games should be extended by the construction of models that consider the evolution of response rules in multi-player games, and games with differing payoff structures from the parental care case.
Partial compensation as a behavioural mechanism for maintaining cooperation has received no explicit attention outside of the literature on cooperation over offspring care. That it should be observed in a radically different context (bacterial public goods production) is intriguing: testing whether partial compensation occurs in other types of cooperation seems a worthwhile pursuit. It is particularly noteworthy that this behavioural strategy has not been explored in our own species, nor has it been explored in organisms amenable to experimental evolution. The evolution of response rules in cooperation is ripe for new consideration and the discovery that a social bacterium exhibits partial compensation could pave the way for new advances in this area – not only because of the numerous advantages of microbes for studying evolution in detail over many generations (Buckling et al., 2009) but also because it could help researchers to consider the evolution of response rules in multi-player games, that is, beyond the dyad typified by biparental care as a model trait and by much of the existing corpus of work on the evolution of social strategies.
I thank Pierre Cornelis for the bacterial strains used in this experiment, Angus Buckling for providing lab consumables and Samuel Alizon, Sam Brown, Mike Cant, Steve Diggle, Melanie Ghoul, Vincent Jansen, Rolf Kümmerli, Jacek Radwan, Tamás Székely, Stu West, the social evolution group at Oxford and two anonymous referees for helpful discussion of the study. This study was supported by a Fellowship by Examination at Magdalen College, Oxford and by the European Research Council.