Epistasis in living colour—mapping an evolutionary trajectory

ResearchBlogging.orgReconstruction of evolutionary trajectories will be a favourite topic on this blog, since it’s a very interesting area that is currently growing rapidly. I already wrote about nice new work showing how heat stability can evolve in thermophiles (heat-loving organisms). Now there’s a new paper looking at how blue color vision arose in the lineage leading to human beings. Both studies asked the same basic and important question: how did evolution get from A to B, given that most of the paths leading from A to B are dead ends?

The new story is “Epistatic Adaptive Evolution of Human Color Vision” by Yokoyama and colleagues, published a few weeks ago in PLoS Genetics. The background is quite interesting. Human vision is trichromatic, meaning that three different colour-detecting systems provide information that is used to sense colour. One of those is a blue-detecting system, in which cells (the cones of the retina) use a blue-sensitive pigment. These cells are not sensitive to ultraviolet (UV) light, which is nearby on the spectrum, and so we humans can’t see UV. But interestingly, our ancestors could, and some of our mammalian cousins (such as rodents) and lots of other vertebrates still can. UV sensitivity was lost in our lineage because of the development of a lens that does not let UV get through.

We know from other studies that our blue-sensitive pigment (a protein called short wavelength-sensitive pigment, or SWS1 pigment) is descended from the UV-sensitive pigment of our ancestor. And we know that the difference between the two proteins (the UV-sensitive ancestor and the blue-sensitive current version) is accounted for by 7 mutations. That’s not a lot of mutations, but the changes represent thousands of possible pathways (trajectories) by which evolution could have gotten to the current version of the pigment. To be precise, there are 7! = 5040 different possible trajectories, which run through 127 potential versions of the pigment. Some versions are dead (because they don’t absorb any light or are unstable), meaning that any trajectory that includes them is a dead end.

In order to figure out how this all went down, we would need to look at all 127 possible version of the pigment to see how many of them are functional. And that’s exactly what Yokoyama et al. did.

The first thing the authors report is that almost half of the mutation combinations result in worthless proteins. In fact, one of the single mutations is completely dead. This rules out 4008 of the 5040 possible trajectories, right out of the gate. But it raises a really interesting question about protein evolution in general: how can it be that a mutation that destroys the pigment’s function is currently part of the human version of that protein? Ignoring the answer to this question is one way that creationists cast doubt on evolution’s ability to navigate trajectories like the one that underlies the change in human colour vision.

The answer to the question is, in a word, epistasis. Epistasis is a term from genetics, defined as a situation in which one gene’s influence depends on what is happening at another gene. Epistasis is common, indeed pervasive, probably because life is complex. Its pervasiveness is one reason why it is perilous to define a specific mutation as “bad” or “good,” since we know that there are cases in which a specific mutation in a particular gene A is “bad” in the presence of one form of a second gene B, and “good” in the presence of a different form. That extreme situation is called “sign epistasis.”

Here, the authors are using the term “epistasis” to refer (correctly) to different mutations in the same gene. The concept is the same—sometimes a mutation A is bad alone but can be quite helpful in the presence of a second mutation, B, in the same gene. This is what Yokoyama and colleagues found in the case of the blue-detecting pigment. Specifically, one mutation would kill the protein if it occurred alone, which means that none of the evolutionary trajectories from the UV-sensing pigment to the blue-sensing pigment could have started with that mutation. However, remarkably, that mutation is not only tolerated but helpful later on. It’s a classic example of epistasis—a mutation that has one effect (devastation) in one context and another effect (functional improvement) in another.

After testing all the different combinations of mutations for their ability to function as light sensors, the authors identified 355 possible trajectories by which the old pigment could have given rise to ours. All of these are feasible (as far as we know), but further analysis points to 8 trajectories that seem to be “best,” based on phylogenetic analysis of a bunch of other species and their likely transitions. (The image below is Figure 5, showing the 8 “best” trajectories with all the others in the background.) They conclude that there is a clear, smooth, gradual, stepwise transition between the UV sensor and the blue sensor, and that epistasis was central to the whole process. Without considering epistasis, the evolution of the blue sensor seems nearly impossible. By doing the hard work of analysing all of the potential transition points, Yokoyama and colleagues show us that the transition was a gradual evolutionary transition, indeed one that could have taken hundreds of different forms.

Yokoyama Fig 5

Evolutionary transitions can seem almost magical to the uninformed and especially to those prone to see magic in the world. In my opinion, the way things really work is a lot cooler than magic. I reserve my reverence for hard-working scientists like Yokoyama and colleagues, who have given us a complete map of a fascinating and important evolutionary transition, through ingenuity and creativity. Bravo!


Yokoyama, S., Xing, J., Liu, Y., Faggionato, D., Altun, A., & Starmer, W. (2014). Epistatic Adaptive Evolution of Human Color Vision PLoS Genetics, 10 (12) DOI: 10.1371/journal.pgen.1004884



%d bloggers like this: