Survival of the Abudant: Mutational Networks Constrain Evolution

There’s been a bit of talk around the blogs about Olivia Judson’s recent article entitled “Let’s Get Rid of Darwinism”.  I do have an opinion on it, but I don’t consider myself well-versed enough on the topic to blog about it specifically.  I do recommend other people’s posts on the subject, Coturnix, Evolving Thoughts, The Frontal Cortex, and Laelaps.

But it seems like a lovely coincidence that, just as Natural Selection celebrates its 150th birthday, (well it was July 1), there is a new paper come out in PLoS Computational Biology on evolution!  I can’t say that I’ll get it right, but I think it’s really cool, and if I don’t get it all right, read it yourself and let me know.  I really do some to this from pretty much a lay perspective, being both bad at math and not very well-versed in the actual ins and outs of evolution, mutations, and fitness (hangs head in shame). 

Cowperthwaite MC, Economo EP, Harcombe WR, Miller EL, Ancel Meyers L 2008 The Ascent of the Abundant: How Mutational Networks Constrain Evolution. PLoS Comput Biol 4(7): e1000110. doi:10.1371/journal.pcbi.1000110

Evolution has been crudely defined for years as being “survival of the fittest”.  But is it REALLY about who is the fittest?  And how do you know who’s fit in the first place?  Scientists have been trying to study this for years.  Eigen and Schuster (1994) came up with a model which involves RNA.  Basically, you can take a strip of RNA and determine its relative “fitness” by how easily it can fold into its predicted secondary structure.  RNA doesn’t just float around the cell in little lines, that leads it too open to tampering.  Instead it can fold into a secondary structure.  But folding requires energy.  The easiest or most “fit” secondary structure would be the one that requires the least energy to achieve. 

One of the things that has come out of this is the idea of “neutral networks”.  As we all know, evolution is the result of mutations taking place that are more or less “fit” and therefore more or less likely to succeed.  However, not all mutations are necessarily better or worse.  Some are just “neutral”.  For example, in some places in a protein, it really doesn’t matter whether you have a glycine or a valine in that particular spot, the protein will still fold the same way.  The same goes for RNA, there are certain areas where a different base is not going to make a difference in the way the RNA folds.  This is known as a “neutral” mutation, it doesn’t have any positive or negative benefit, so it’s neither more nor less likely to be passed on. 

What has been found over the last few years is that these neutral mutations occur in networks.  That means that there are little fleets of genotypes, all of the same “fitness”, that have overlapping series of neutral mutations.  Most of these fleets are small, but a few are larger, and its the larger fleets of genotypes that the researchers in this study focused on.  The large networks tend to be adjacent to a pretty large number of phenotypes.  So you have all these little neutral mutations, next to RNA with a wide variety of phenotypes.  Do these little neutral mutations influence evolution after all?

Since this was computational biology, it’s a lot of computer modeling.  The researchers predicted the structions of all RNA molecules that are between 12-18 nucleotides.   What they got was a large number of genotypes that gave rise to a relatively limited number of phenotypes.  This means that these large number of genotypes had a lot of overlapping neutral mutations, that didn’t affect the final phenotype of the RNA.  They also found that the number of phenotypes you get increases drastically as you increase the number of nucleotides.  This makes a lot of logical sense, because increasing the number of nucleotides increases the likelihood that you’re going to get a different folding pattern. 

These various phenotypes are connected to each other by their overlapping mutations.  You can then look at these mutational networks and see which mutations account for each phenotype.  The mutational networks contain a certain number of options for mutations, which ones can arise, and if they arise, whether or not they will prove beneficial or harmful.  The networks mutate from phenotype to phenotype, and the structure of the network will influence whether or not any given phenotype will arise (they call it accessability), and, if it arises, whether or not it will be worthwhile (which they call evolvability).  

With all this in mind, it makes logical sense that abundant phenotypes (where lots of genotypes form the same phenotype) are more evolvable than rare ones, because if you have lots of genotypes forming a single phenotype, you have a higher number of genotypes that can mutate, and so a greater possibility of producing a new phenotype (there’s math to explain this, but I’m horrid at understanding math.  If you want to see the math, read this paper or the other ones on this part:  Wagner, 2008, Wagner, 2005, and Reidys 1997.) 

However, just having a higher number of genotypes that can mutate doesn’t mean that any particular genotype is likely to produce a useful mutation.  The probability that a mutation will prove worthwhile usually declines as the size of the neutral network grows.  All these genotypes producing only one phenotype (a large neutral network) means that they all have certain things they retain in common or similarity that allow for the same phenotype. 

Interestingly, what they found was that random mutations were more likely to move genotype to a large neutral network than to a small one.  The mutations follow the in crowd, and this works particularly for phenotypes that are linked to other common phenotypes.  And once a genotypes gets into the large neutral network phenotype, it’s hard to escape, because the possibility of more mutations proving worthwhile becomes reduced.  So a large neutral network CONSTRAINS evolution! 

This also means that, because it’s easy to get to the common phenotype, and hard to get out, the most abundant phenotypes possible may not actually be the ones that are the most worthwhile, because the probability of the worthwhile mutation occurring decreases once a genotype is within the common phenotype.

“These results suggest the following hypothesis: the evolution of phenotypes, whether complex whole-organism phenotypes or RNA shapes, may be biased toward abundant phenotypes, even if those phenotypes are not optimal.”

So they set out to test the abundance of various RNA folding shapes using another computational model, to see if naturally occurring RNA shapes are biased toward abundantly occurring shapes.    Indeed they are, and the researchers say that this provides support for the “ascent of the abundant” hypothesis, the idea that survival is not necessarily of the fittest, but of the most common.  Abundant phenotypes are easier to mutate into and harder to mutate out of. 

Of course, this is all simulation, and involves just RNA molecules.  So more complex molecules or organisms could very easily work in different ways.  Also, they did not include natural selection for thermostability of the RNA molecules, so that could have an influence on whether or not genotypes are able to form large neutral networks of phenotypes, and how easily they can produce new phenotypes.  But the implications for how evolution works (at least in RNA molecules) appear to have gotten a little more complex.

Cowperthwaite, M.C., Economo, E.P., Harcombe, W.R., Miller, E.L., Meyers, L.A., Bourne, P.E. (2008). The Ascent of the Abundant: How Mutational Networks Constrain Evolution. PLoS Computational Biology, 4(7), e1000110. DOI: 10.1371/journal.pcbi.1000110