Reconciling robustness with evolvability
November 19, 2010 § 5 Comments
Marc Kirschner pointed me to this interesting recent paper about the evolvability of proteins (Philips et al. 2010. Robustness and evolvability in the functional anatomy of a PER-ARNT-SIM (PAS) domain, PNAS PMID: 20889915). What’s evolvability? It’s a term used to indicate the qualities of a molecule or organism that allow it to evolve effectively; the ability to mutate, acquire new functions, and be acted on by natural selection. Implicit in this term is the idea that it’s possible for a species — or a molecule, or an ecology — to be better at evolving than is its competitor.
So, you may ask, what does it take to be good at evolving? There are probably many and varied routes to improved evolvability (many of which are discussed in the two books Marc has co-authored on the topic), but there is one that seems basic: you should not be extremely fragile to mutations. If most of the possible single mutations kill you, then nearly every improved function that requires two simultaneous mutations is out of your reach. But if single changes don’t bother you, then the population you are part of can accumulate many changes, most of which are silent. A second (or third, or nineteenth) change on top of one of these silent changes could lead to an important new function, and then you’re off to the natural selection races. The ability to absorb mutations without losing viability is called robustness; this, like evolvability, is something of a vague term (robust to what? Evolvable under what circumstances?) but the concept is none the less important.
Protein function does indeed seem to show robustness, in the sense that only about 15% of point mutations — mutations that change just one amino acid residue in a protein sequence — cause the protein to lose its function. This has led to the notion that relatively few amino acids significantly affect a protein’s function; for example, in an enzyme the critical residues might be at or near the active site, while the rest of the protein is only needed to fold up into the proper shape. And this may seem all very fine and sensible, but there’s a problem. The more robust you are to individual mutations, the harder it is to imagine evolving radically new functions. Robustness is essential for evolvability, but it also seems to be a barrier to evolvability. It’s a puzzle.
But in the context of protein function, this apparent paradox rests on the idea that if only 15% of mutations cause loss of function, the other 85% are not doing anything much. Since the main way this idea arose was from in vivo testing using outputs that integrate many factors (such as plaque formation in response to bacteriophage T4, or growth in the presence of antibiotics) this may not be a safe assumption, and Philips et al. set out to test it. They chose a small protein, photoactive yellow protein (PYP), which is used by the bacterium Halorhodospira halophila to detect light, which the bacterium then swims away from. PYP has several useful features. It undergoes a conformational change as a result of absorbing light in its central chromophore, and so there are several properties related to its function, such as light absorbance and the lifetime of the excited chromophore, that are easy to measure accurately. Its structure has also been determined at high resolution, making mutations easier to interpret structurally; and it is part of the large family of PAS-domain-containing signaling proteins.
Philips et al. used a high-throughput approach to mutate each of the 125 amino acids in the protein individually to alanine, express and purify the mutant protein, and quantify five different measures of protein function, plus protein expression levels. Alanine-scanning mutagenesis, as this approach is known, is a general way of probing how important a particular side-chain is. It was invented by Jim Wells‘ lab when he was at Genentech, and was first used to ask which residues on the human growth hormone are important for its binding to a receptor. I think of it as one of the first of the “I don’t know where to look so I’ll look everywhere” techniques that have become so prevalent and powerful in the world of omics. It’s not perfect as a simulation of what might actually happen in evolution; it basically just chops off the whole side chain except for one CH3 group, whereas real mutations would produce varied side chain changes. But if a particular side chain has a specific role in protein folding or protein function, this is a pretty good way to spot it.
When the authors examined their 125 single-mutant PYP proteins, two things jumped out at them. First, most of the mutations change some aspect of the protein’s functional behavior; for example, the maximum wavelength of light absorbed varies in these mutants from 439 to 468 nm, and the lifetime of the activated state varies over a factor of 3,000-fold. And second, a large fraction of the mutations (40%) change how well the protein is expressed, in some cases dramatically. Overall, about 60% of the alanine mutants show a functional difference from the wild-type protein that you would imagine could be acted on by natural selection. But, only one mutation led to a complete loss of protein function. So PYP is still robust — it’s hard to break it completely — but it is also evolvable, because it’s easy to make changes that can lead to altered function. And most of the mutations that cause changes are far from the active site.
The fact that about 40% of mutants alter protein expression is a big surprise. We’re used to the idea that changes in the regulatory regions of genes can easily change expression levels. But the observation that it is so common for sequence changes in the coding region to change expression level is startling. Many of the low-expressing mutants turn out to be insoluble; others are proteolytically degraded. Looking at the structure, Philips et al. find that the 39 amino acids that give the strongest effect are found in the core of the protein. PYP has only about 23 highly conserved residues — it turns out that most of these are among the 39 that show strong reductions in protein expression when mutated to alanine. So maintaining high levels of production seems to be a major selective pressure in the evolution of protein structure.
The cool thing about this paper is that it shows that proteins carrying just one change in their amino acid sequence could be a much greater force in evolution than we thought. It challenges the idea that most of evolution takes place as a kind of random walk in sequence space — the neutral theory of evolution — by showing that the majority of changes have consequences, and can therefore undergo selection. Which may in turn help us understand how evolution has come so far, so fast.
Philip AF, Kumauchi M, & Hoff WD (2010). Robustness and evolvability in the functional anatomy of a PER-ARNT-SIM (PAS) domain. Proceedings of the National Academy of Sciences of the United States of America, 107 (42), 17986-91 PMID: 20889915