Epistasis, neutrality and evolvability

Posted 27 January 2008 by PvM

Some may know that one of my 'hobby' aspects of evolution involves the evolution of evolution, aka evolvability and how neutrality is a necessary requirement for evolvability. Let's walk through an example which helps explain my position. I will use a recent paper by Andreas Wagner "Hypothesis: Robustness, evolvability, and neutrality" FEBS Letters 579 (2005) 1772–1778 Let's first look at a simple RNA molecule Note how some of the nucleotides are circled grey? These are locations which are neutral with respect to a mutation. In other words, a mutation to any of these does not change the folding of the RNA molecule. Now let's mutate one of these neutral nucleotides and see what happens. The nucleotide at position X is changed from a C into a G and as expected, the RNA molecule remains the same (after all X was a neutral location). However, some previously neutral sequences have now lost their neutrality (as indicated with a minus) and some previously non-neutral sites have become neutral (as indicated with a plus). Now let's mutate position Y from a G to an A and compare the effect on both, similar RNA molecules. The effect is dramatic, two different molecules despite the fact that the original RNA molecules were similar. This shows that the effect of a mutation at position Y depends on the mutation at position X, even though the mutation can be neutral in its effect. This is called epistasis. Both resulting structures evolved from the same shape RNA and yet, the one of the left has no mutation at X and the one on the right has X mutated to a G. However, this mutation was a neutral mutation before the Y mutation. This opens up some exciting new possibilities for evolution, and it should not come as a surprise that evolution indeed seems to have taken use of this opportunity.

12 Comments

Nigel D · 28 January 2008

I assume (not currently having access to FEBS Letters) that the relevance would be to any mRNA or snRNA, but less so for rRNA or tRNA?

Thus, I assume the altered shape of the RNA would have some impact in terms of up- or down- regulation of (potentially) both the transcription and the translation of the transcribed gene. Is it, or not really?

Lifewish · 28 January 2008

That is soooo cool... I think I just had a sciencegasm. Thanks for sharing this with us.

David Stanton · 28 January 2008

Nigel,

The same type of considerations would probably apply to both rRNA and tRNA sequences as well. The pattern of nucleotide conservation in these molecules shows that some substitutions are probably neutral (mostly loop regions and expansion segments in rRNA), some are deleterious (mostly stem positions in rRNA or stem and identifier positions in tRNA), and some are neurtal if compensatory mutations also occur (mostly stems in rRNAs). The same would probably apply to siRNA as well, although there exact matching to target sequences is sometimes required.

Nigel D · 28 January 2008

David, thanks for the clarifying detail. I must confess that rRNA and tRNA are things about which I know little more than the basics.

So, if I understand this correctly, in terms of the overall cell biology, this type of supposedly-neutral mutation could alter the transcriptional rate of a gene. Or, it could alter whether the transcript gets digested before significant translation has occurred. Or it could alter the translation of one or more genes in perhaps three different ways (different mRNA - just the transcript is affected; different rRNA - may affect all translation episodes involving a sub-population of ribosomes; different tRNAs - may affect all translation episodes involving a sub-population of tRNAs).

Nigel D · 29 January 2008

It's funny how posts on the actual science tend to draw fewer comments, isn't it?

I wonder if that's because the evidence kinda speaks for itself?

JGB · 29 January 2008

It's been a little while since I've stayed up to date on some of this. Outside of RNA are there any other well study molecular epistatic systems?

Pete Dunkelberg · 30 January 2008

Jargon alert.

Epistasis is used in more than one way. Often it refers to the effect of a whole gene being modified by one or a few other genes, which may be referred to as 'modifier genes'. These modifiers often 'mask the phenotypic effect' of the gene in question.

And then there is epigenetics, which is heritable changes in phenotype not due to a DNA sequence change. This is usually caused by methylation of DNA, which tends to reduce expression of a gene, or in other words mask it's phenotypic effect.
(The reason methylation does not count as a sequence change is that we humans define "the" sequence in terms of nucleotides only.)

Nigel D · 30 January 2008

Pete, I think there is more to it than that.

Methylation of DNA is a mechanism of identifying the "correct" strand. As DNA is being replicated, errors occasionally creep in. Sometimes a T is misincorporated in place of a C, or a G in place of an A, or whichever. Errors such as these result in a mismatch of base pairing, and there are enzymes that detect and correct these erors. However, there is no intrinsic difference between the template strand and the newly-synthesiesed strand, so 50% of error corrections would result in the incorporation of a mutation were it not for methylation. Methylation of DNA occurs distinctly from replication, so there is a time after replication when one strand is methylated and the other is not. Thus, the enzymes have a means to identify the correct strand and hence make a correct correction.

Of course, the extent of methylation also influences gene expression as you say. I am unsure about the mechanism for this, so I cannot comment.

tomh · 30 January 2008

Nigel D said: It’s funny how posts on the actual science tend to draw fewer comments, isn’t it?

Perhaps because one actually has to know something to comment on a post like this one, as opposed to arguments on religion, ID, atheism, etc. That's not to say that we uneducated types don't appreciate this type of post. This one was especially good, both brief and clear, which is not easy to do.

Pvm · 30 January 2008

Thanks. I am trying to slowly make a case as to why neutrality can be an important component of evolvability. At least for RNA there seems to be a clear case. However, things may not be similar for DNA, more on that later. What the above example shows is that the same phenotype can have different genotypes and that because of 'epistasis', phenotypes can still respond differently to the same mutation. In other words, a population of the same phenotype may still have significant variation in the underlying genotype and thus the impact of the same mutation can differ greatly.

tomh:
Nigel D said: It’s funny how posts on the actual science tend to draw fewer comments, isn’t it?
Perhaps because one actually has to know something to comment on a post like this one, as opposed to arguments on religion, ID, atheism, etc. That's not to say that we uneducated types don't appreciate this type of post. This one was especially good, both brief and clear, which is not easy to do.

Henry J · 30 January 2008

I am trying to slowly make a case as to why neutrality can be an important component of evolvability.

That certainly makes sense intuitively, since (as I understand it) selection needs varieties to be already present for it to work on them, which means that some variation has to be generated before selection can act on it, and without some neutral variation it would depend on mutations happening when needed, which would be a roadblock, so to speak. (Of course, an actual argument for that would need the relevant technical details, and those would probably be over my head.) Henry

Herbert M Sauro · 30 January 2008

That's a very nice example of epistasis, could be readily used in class, I thank you pointing it out.

Another way of thinking about epistasis, though certainly not as elegant as the RNA example, is to think of a function of two variables, z = f(x, y) which forms a surface. Take the partial derivatives, df/dx and df/dy. If I change x I will move across the surface to a new location where both slopes are likely to be different (Think of walking across a rolling countryside). If we were to observe df/dx and df/dy as we change x we would see one as a function of the other, that is an epistatic interaction.