Development shapes evolution in silico

Why are there so few extant phyla or basic body-plans? There are only about 35 phyla, and nine of them include 96% of all living species. There used to be many more, back in the early days of the cambrian explosion, so one has to wonder why so many phyla were singled out for extinction, and just a few have gone on to do very well.

The usual way to learn about our evolutionary history is to look at fossils. Unfortunately, fossils don't tell us very much about how the organisms developed, and this is sad because development is believed to have everything to do with organismal evolution. Comparative molecular biology is useful in tracing the actual events as they unfolded in the past, but it is still very difficult to deduce what really happened, and what caused the appearance of one body-plan instead of another.

Borenstein and Krakauer attempts a third path, namely a computational model of development. This has the obvious advantage that everything that happens is knowable; whatever leads to the origin of a new body plan can be studied in detail. The caveat is then that only the elements that has been put into the model can affect what goes on. In other words, much more control and illumination, at the expense of realism. The challenge is to either make the model as realistic as possible, or make it exactly suitable to answer the question that we are interested in. Here the latter approach is taken.

Their goal is to investigate the effects of developmental bias on evolutionary trajectories, and thus to shed some light on the hypothesis that the evolutionary history of phyla can be partly explained by developmental bias without the need for natural selection. No doubt that everyone on their right mind realizes that natural selection played a huge role giving the evolutionary pathways a direction, but this paper imagines how much direction can be explained without the mechanism of natural selection.

The model they propose is a simple genotype-phenotype map where the mapping is given by a developmental matrix. This matrix describes how the binary genotype (i.e. the genetic configuration) maps onto a phenotype (the complete description of the organism, including morphology, physiology, etc.). What that means is that a genotype through this mapping becomes a certain phenotype, and further that there need not be a one-to-one correspondence between the two. This results in canalization, meaning that several distinct genotypes creates identical phenotypes. Changing between these genotypes has no phenotypic effect, so the genotype is robust to mutations.

The developmental model can be thought of as a set of regulatory genes that are either turned on or off (1 or 0), while the resulting phenotype is the subset of transcription factors that the genes can affect to be expressed or not, thus defining the organism. In between these, the developmental matrix describes how the genes interact, setting the stage for epistatic interactions between the genes.

Figure 1. Pairwise Hamming-distances between organisms within and among phyla.

The main result presented in the study is that by letting the developmental map evolve without natural selection, the disparity between diverging groups (phyla) increases, while the genotypic and phenotypic variance becomes smaller within the groups (figure 1: hamming distance is a measure of how many mutations the two genotypes are away from each other). This is nicely illustrated in figure 2, where the phylogeny has been traced first ancestor to all the varieties present after the organisms have evolved. Notice how lineages with common histories have similar or identical phenotypes.

Figure 2. Phylogenetic tree of evolved organisms.

Developmental bias has the effect of reducing the possible phenotypes to a subset of visible phenotypes. Once development has gone down a specific path, it is forever stuck with that type. The changes that can be made after that fact is much more limited than the changes that were possible given the basal body-plan way back in time. Thus, the idea goes, we have many species within each phylum, but they are all stuck with the basic body-plan of that phylum, forever destined to not to roam far in genotype-space.

---

Elhanan Borenstein, David C. Krakauer (2008). An End to Endless Forms: Epistasis, Phenotype Distribution Bias, and Nonuniform Evolution PLoS Computational Biology, 4 (10) DOI: 10.1371/journal.pcbi.1000202