Spatial structure can favour the maintenance of indiscriminate cooperation through increases in the genetic similarity of interacting individuals, but cooperating patches can still be invaded by ‘cheats’ that benefit from the cooperative behaviour, but do not pay the costs of participating in it (Hamilton 1963, 1964; Nowak 2006). However, where explicitly investigated, cooperation is the norm in natural populations of microbes. For example, only eight of 345 natural isolates of Pseudomonas aeruginosa were found to be cheats with respect to the production of extracellular iron-scavenging siderophores (Bodilis et al. 2009). While this could be explained to some extent by assuming cheats rarely encounter patches of cooperators through immigration, a number of recent studies demonstrate how easily cheats can be generated by mutation, and subsequently increase in frequency (Rainey & Rainey 2003; Harrison & Buckling 2005; Sandoz et al. 2007; Kohler et al. 2009; Racey et al. 2010). Cheat invasion can be prevented if a single locus that confers cheating behaviour has a negative effect on another fitness trait (Foster et al. 2004), but such pleiotropy does not seem to be common in microbial cooperative behaviours (West et al. 2007), and certainly does not limit the invasion of siderophore cheats in laboratory adapted strains of Pseudomonas aeruginosa (Harrison & Buckling 2005). Here, we investigate a simple and potentially ubiquitous mechanism limiting cheat invasion in natural populations of microbes: strong non-social selection pressures.
Environmental change, particularly changes in other interacting organisms such as predators (Feldman et al. 2009), parasites (Koskella & Lively 2007), and competitors (Joshi & Thompson 1996), can result in maladaptation of populations to their local environment, and hence strong selection (Bell 2010). Assuming new patches are founded by cooperating cells, as patches of cooperators are more productive and so produce more migrants (Hamilton 1963, 1964; Nowak 2006), beneficial mutations are more likely to arise by chance in the numerically dominant cooperator population, than in rare invading cheats. If these non-social beneficial mutations are under stronger selection than the cheating mutation, cooperative alleles will hitchhike with the beneficial mutations and the cheating mutation will decrease in frequency by clonal interference (Fisher 1930; Schiffels et al. 2011) from the beneficial non-social mutation. Crucially, there are likely to be many selective sweeps due to continual environmental change (especially with a coevolving parasite), or if multiple mutations are required to reach fitness optima. Hence, although a cheat may gain a beneficial non-social mutation, or a cheat may emerge from a cooperator population that has gained one beneficial non-social mutation, additional beneficial mutations are more likely to arise in the numerically dominant cooperator population.
To test our hypothesis of indirect positive frequency-dependent cooperator fitness when adapting to novel environments, we developed a theoretical model and tested it by evolving populations of the bacterium Pseudomonas fluorescens SBW25. Wild-type P. fluorescens produces extracellular siderophores that scavenge iron from the environment and are re-adsorbed by cells (Cornelis 2010), and we establish here that siderophore production is a cooperative trait, as is the case in the closely related bacterium, Pseudomonas aeruginosa (Griffin et al. 2004). We then evolved mixed populations of wild-type P. fluorescens and an isogenic mutant cheat strain at various starting frequencies. Strong directional selection was created by culturing them for c. 40 generations in iron-limited nutrient media, a novel environment for these non-laboratory adapted bacteria. Moreover, some populations were coevolved with a virus (the lytic DNA bacteriophage, SBW25Φ2), resulting in continual strong selection for resistance (Buckling & Rainey 2002; Morgan et al. 2005). Note that competition occurs only within-, and not between-, patches (i.e. soft selection; (Wallace 1975)) in both our model and experiments, although we implicitly assume the operation of global regulation and hence between-patch competition (i.e. hard selection) to explain why cooperators are likely to be the first to colonise empty patches.