A preview of an article that is going to be in the June issue of American Naturalist, The Evolution of Phenotypic Polymorphism: Randomized Strategies versus Evolutionary Branching:
A population is polymorphic when its members fall into two or more categories, referred to as alternative phenotypes. There are many kinds of phenotypic polymorphisms, with specialization in reproduction, feeding, dispersal, or protection from predators. An individual’s phenotype might be randomly assigned during development, genetically determined, or set by environmental cues. These three possibilities correspond to a mixed strategy of development, a genetic polymorphism, and a conditional strategy. Using the perspective of adaptive dynamics, I develop a unifying evolutionary theory of systems of determination of alternative phenotypes, focusing on the relative possibilities for random versus genetic determination. The approach is an extension of the analysis of evolutionary branching in adaptive dynamics. It compares the possibility that there will be evolutionary branching, leading to genetic polymorphism, with the possibility that a mixed strategy evolves. The comparison is based on the strength of selection for the different outcomes. An interpretation of the resulting criterion is that genetic polymorphism is favored over random determination of the phenotype if an individual’s heritable genotype is an adaptively advantageous cue for development. I argue that it can be helpful to regard genetic polymorphism as a special case of phenotypic plasticity.
I have cut & pasted the discussion below….
Studying the evolution of phenotype determination in terms of the signs and relative magnitudes of the branching and randomization disruptivities, as illustrated in figure 1, has the advantage of bringing genetic and random determination into a single analysis. This was achieved by limiting consideration to situations close to a phenotypically monomorphic equilibrium. The approach can be regarded as an extension of the analysis of evolutionary branching in adaptive dynamics (Metz et al. 1996; Geritz et al. 1998), which gives a unified perspective on the evolution of genetic polymorphism. Such conceptual unification is helpful and important, but when alternative phenotypes have evolved away from a previously monomorphic state, there is in principle a new situation that may require a separate analysis. In addition, phenotypic polymorphism could come about in different ways, such as by modification of a previously existing polymorphism or by mutations of large effect or drastic changes in environments. Nevertheless, the individual-based simulations I performed showed that an analysis in terms of branching and randomization disruptivities can often succeed in predicting the evolutionary outcome.
One important feature of my analysis is that cases are split into those where a genotype can function as an advantageous cue for determination of alternative phenotypes and those where such a genetic cue instead would be disadvantageous compared to random determination, as well as the intermediate cases where genetic and random determination are neutral relative to each other. This kind of division has general validity and makes the point that genetic determination of the phenotype is conceptually parallel to environmental determination, for which issues of the accuracy and other statistical properties of environmental cues are fundamental. In fact, it could be helpful to regard genetic determination of alternative phenotypes as a special case of adaptive phenotypic plasticity.
It is common to think of alternative phenotypes as threshold traits or developmental switches (Roff 1996; Lynch and Walsh 1998; West-Eberhard 2003), where trait expression is switched when the value of an internal liability passes a threshold. The liability might be the concentration of a hormone, which in turn could be influenced by genes, environmental cues, or random effects. If we ask in which way such a phenotype-determining mechanism can be an adaptation, there is the question of the perfection of alternatives being switched between, on the one hand, and the question of when to switch, on the other hand. For the latter, we can regard random variation in liability as adaptive if suitable phenotype frequencies are produced in this way, and any further improvement in the fit of phenotype to selective circumstances achieved by a switching mechanism is then also a possible adaptation. Environmental cues are, of course, candidates to be employed in such better-than-random switching mechanisms, but, as we have seen, heritable genetic variation is also a candidate. Thus, the alternative phenotypes of a multiple-niche polymorphism, together with the switching mechanism, are elements of an adaptation to spatially varying conditions. For genetic switching, spatial variation in phenotype frequencies from recent selection, unless obliterated by extensive gene flow, will be the source of information in the genetic cue. There is some similarity to the idea of local adaptation when gene flow is limited, but the difference is that the alternative phenotypes and the switching mechanism make up one adaptation to a range of environments rather than that each phenotype separately is an adaptation to a single environment. So, in the context of an evolved switching mechanism, like a nonlinear genotype-phenotype mapping, it becomes natural to view genetic determination of the phenotype as potentially an adaptation, falling within the general framework of phenotypic plasticity.
Several decades ago, ecological geneticists were concerned with providing an adaptive interpretation of genetic polymorphism (e.g., Dobzhansky 1951; Cain and Sheppard 1954; Fisher 1958; Ford 1965, 1971), but this interest seems to have declined in recent times. Possibly, the currently dominating idea of the gene as the fundamental unit of selection (Dawkins 1976) has been responsible for the decline because that idea might seem to be in conflict with the notion that genetic variation could play a role in an adaptive device for the organism. In ecological genetics, a traditional but also controversial view was that genetic polymorphism can be adaptive if it renders an organism efficient over a range of environments (Dobzhansky 1951; Cain and Sheppard 1954), corresponding to the modern concept of multiple-niche polymorphism. When examining this idea, Fisher (1958) traced it to the discussion by Darwin (1859) of divergence of character under natural selection and went on to suggest that a “theory of games,” with randomized strategies as one feasible evolutionary outcome, could be a way to understand species interactions, such as those between predators and prey. Fisher’s treatment is significant as apparently the first explicit suggestion of an evolutionary game theory. It is also clear that Fisher (1958) interpreted balanced polymorphism as an evolved randomized strategy. Ford (1965) had a similar view of genetic polymorphism as a means to regulate phenotype frequencies, and he argued that it would usually be preferable as a switching device, compared to environmental cues, perhaps because he felt that environmental cues would result in excessive fluctuations in phenotype frequencies. It then appears that traditional ecological genetics did not view genetic determination of the phenotype as essentially different from random determination. It was only with the subsequent analysis of bet hedging that this distinction became clear (Seger and Brockmann 1987), although the interpretation of genetic polymorphism as a special case of phenotypic plasticity, which I have argued for here, was not made.
My treatment did not deal directly with environmental phenotype determination, which is possibly the most widespread and important system, because I wanted a simple and focused treatment of genetic and random determination. Never
theless, environmental determination could be included in the same kind of analysis. By regarding reaction norms as multidimensional “pure strategies,” one can investigate whether some component of such a pure strategy would be exposed to disruptive selection, possibly leading to evolutionary branching and genetic polymorphism or, alternatively, whether a randomized reaction norm might evolve. A similar kind of approach has recently been used to study environmental sex determination (Van Dooren and Leimar 2003; Leimar et al. 2004). Factors such as the accuracy of environmental cues and the degree of correlation between environmental cue values observed by different individuals, together with other aspects of the ecological situation, will determine if some component of a reaction norm would be exposed to branching or randomization disruptivity.
Of the examples I used, the effect of competition between relatives on the evolution of alternative phenotypes and on phenotype determination seems not to have been modeled previously. The conclusion that random determination should be favored was, however, reached by Moran (1992), and Day (2001) found that relatedness decreased branching disruptivity in situations where relatives compete. In practice, compared to purely random determination, other mechanisms based on phenotypic cues like relative size or age might often be more efficient in reducing competition between interacting relatives and could thus be expected instead.
On the basis of the simulations for the examples, it is clear that the mutational process can have a marked influence on the system of phenotype determination that evolves. For instance, if there is little additive genetic variation in the degree of random phenotype determination, genetic determination might instead be the evolutionary outcome. A similar point was made by Van Dooren and Leimar (2003) and Leimar et al. (2004) in the context of sex determination. The explanation is that genetic and random cues can sometimes serve equally or nearly equally well for phenotype determination. As a consequence, in situations where branching and randomization disruptivities are equal, it seems appropriate to regard a genetic determination of the phenotype as a randomization device. Nevertheless, considering general fitness functions, the theory developed here shows that those situations are marginal. Using heritable genetic variation for phenotype determination will generally either be selectively favored and thus be a case of adaptive genetic determination of the phenotype or instead have a selective cost as compared to random determination.
Posted by razib at 10:11 PM