Constraints on Evolution

Mutation provides grist for the mill of evolution, but these genetic changes aren't usually directly exposed to selection. Natural and sexual selection act on a creature's characteristics — its phenotype — and so can only see mutations that affect those characteristics. A complete picture of evolution will have to include an account of how an organism's genes — its genotype — give rise to its phenotype, as well as some way of mapping the phenotype to fitness. Although these are enormous questions, each the basis of a field in its own right (development and ecology, respectively), evolutionary biologists have to consider the ways in which theses processes can constrain and affect the evolutionary pathways that are available.

Isaac Salazar-Ciudad and Miquel Marin-Riera took advantage of a well-established computational model of tooth development to investigate this question. The model incorporates various genetic and cellular interactions that determine the shape of a tooth; these parameters represent the genotype in the model, and variations in them are effectively mutations. The developmental model translates these genotypic variations into realized phenotypes — teeth of different shapes. By randomly changing the parameters and allowing teeth to reproduce based on their fitness, the researchers simulated populations of evolving teeth.

The duo tested three different ways of evaluating fitness, all of which were based on comparison with a predetermined optimum. In the first phenotype-to-fitness map, a tooth's fitness was determined by how exactly its shape matched an optimum tooth in every detail — every ridge, bump, and edge. The second map was less strict; fitness depended on how similar the tooth was to the optimum in only a few characteristics rather than the precise shape. The third map wasn't based directly on the shape. Instead, an 'orientation patch count' (OPC) was calculated for each tooth and compared with the optimum. The OPC is a measure of the overall ruggedness of the tooth and also serves as a good predictor of diet in mammals; those with a low OPC tend to eat animals, while higher OPC values are associated with plant-eaters. Even though the OPC is indirectly based on the shape of the tooth, it's effectively a phenotype-to-fitness map based on a tooth's function rather than its shape.

Populations using a phenotype-to-fitness map based on shape usually weren't able to evolve the optimal phenotype in the simulations. The first map, in which fitness was most strictly tied to form, severely limited the population's ability to explore different phenotypes. Populations had to be very large to have a chance of finding the optimum tooth shape, and even then it was only possible if they started relatively close to it. Even the somewhat less strict second map, based on only a handful of characteristics, was only able to find the optimum shape in around 40% of the simulations. Only when very few (2-4) characteristics were used to determine the fitness did the simulated populations reliably find their way to the optimum.

The function-based map, on the other hand, enabled populations to always evolve to the optimum. This is because the OPC sums many phenotypic features into a single value, so the relationship between fitness and phenotype in the function-based map is simple. Teeth of different shapes can be functionally quite similar, so small changes in the phenotype don't necessarily affect the fitness; this gives the population the freedom to explore different phenotypes and eventually discover the optimum. By contrast, the relationship between fitness and phenotype can be quite complex with shape-based maps, depending on how many characteristics are taken into account. Changes in the phenotype have a much stronger effect on fitness, so populations can end up trapped at a point where any change would reduce fitness.

With this work, Salazar-Ciudad and Marin-Riera have shown how the translation from genotype to phenotype to fitness can constrain an evolving system. If the overall genotype-to-fitness map is too complex, populations can't explore the landscape of possibilities and end up trapped away from the optimum; a degree of simplicity is required in either the genotype-phenotype or the phenotype-fitness map in order to give the population some flexibility. What this means in practice is that complex developmental systems will find it difficult to evolve towards a specific shape. Instead, these systems are likely to be crafted by selection based on coarse criteria and evolve to fulfil a function. In other words, evolution doesn't tinker and tweak to optimize every facet of development, but just settles for whatever will get the job done.

Ref
Salazar-Ciudad I & Marín-Riera M. (2013) Adaptive dynamics under development-based genotype-phenotype maps. Nature 497(7449):361-4. doi: 10.1038/nature12142

[Disclaimer: I know one of the authors of the paper.]