Interspecific mutualisms have been playing a central role in the functioning of all ecosystems since the early history of life. Yet the theory of coevolution of mutualists is virtually nonexistent, by contrast with well-developed coevolutionary theories of competition, predator-prey and host-parasite interactions. This has prevented resolution of a basic puzzle posed by mutualisms: their persistence in spite of apparent evolutionary instability. The selective advantage of 'cheating', that is, reaping mutualistic benefits while providing fewer commodities to the partner species, is commonly believed to erode a mutualistic interaction, leading to its dissolution or reciprocal extinction. However, recent empirical findings indicate that stable associations of mutualists and cheaters have existed over long evolutionary periods. Here, we show that asymmetrical competition within species for the commodities offered by mutualistic partners provides a simple and testable ecological mechanism that can account for the long-term persistence of mutualisms. Cheating, in effect, establishes a background against which better mutualists can display any competitive superiority. This can lead to the coexistence and divergence of mutualist and cheater phenotypes, as well as to the coexistence of ecologically similar, but unrelated mutualists and cheaters.

Mutualisms are ubiquitous in nature, as is their exploitation by both conspecific and heterospecific cheaters. Yet, evolutionary theory predicts that cheating should be favoured by natural selection. Here, we show theoretically that asymmetrical competition for partners generally determines the evolutionary fate of obligate mutualisms facing exploitation by third-species invaders. When asymmetry in partner competition is relatively weak, mutualists may either exclude exploiters or coexist with them, in which case their co-evolutionary response to exploitation is usually benign. When asymmetry is strong, the mutualists evolve towards evolutionary attractors where they become extremely vulnerable to exploiter invasion. However, exploiter invasion at an early stage of the mutualism's history can deflect mutualists' co-evolutionary trajectories towards slightly different attractors that confer long-term stability against further exploitation. Thus, coexistence of mutualists and exploiters may often involve an historical effect whereby exploiters are co-opted early in mutualism history and provide lasting 'evolutionary immunization' against further invasion.

No abstract