09 April 2008

Mutations, selection, and bacteria

Several weeks ago, a commenter (Donald) asked an interesting question about natural selection and genetic variation, and I promised to address it because I want the issue to be a theme on QoD in the coming months. Here's Donald:
The link below is to a NYT blog where it says that E coli studies have found that there are 100,000 harmful mutations for each single beneficial one. I'm no population geneticist, but this kind of thing does make you wonder how selection could work with that much noise to overcome.

That aside, I have read a little of Ronald Fisher and I recall his mathematical argument that for mutations of very small effect, there was a 50 percent chance that the net effect would be beneficial. This is in "The Genetical Theory of Natural Selection". The mutations with large effects, on the other hand, are almost certainly going to be deleterious.

So are these studies only detecting mutations with large effects, or was Fisher wrong?
The blog article that Donald is citing is at The Wild Side by Olivia Judson, and the figure of 100,000 deleterious mutants for every helpful one is widely referenced.

Donald raises two questions, which I'll rephrase somewhat.

1. How can natural selection lead to adaptation when there is so much interference from harmful mutations?

I think there are at least three misconceptions that are acting together to create this common misunderstanding. First, that widely-cited ratio of harmful to helpful mutations is apparently an overestimate, off by three orders of magnitude, or a factor of 1000. The study that reported this dramatic correction in our understanding of bacterial mutations was published in Science last August, and represents a wonderful case study of the difference between real scientific thinking and the thinking of most design advocates. (Subject of an upcoming post.)

Second, the existence of harmful mutations doesn't necessarily "interfere" with adaptation. Many deleterious mutations will just kill the organism, and that's that. Natural selection does that all the time, and it doesn't get in the way of life in general, so there's no special reason to worry that it will get in the way of adaptation.

But most importantly, I think Donald is a little confused about the material on which natural selection acts, and understandably so. (This error is the centerpiece of Michael Behe's ludicrous recent book The Edge of Evolution.) The mistake seems subtle, but it's gigantic, and I think it arises in part from a semantic shortcut that is often used when explaining selection and adaptation. To see the problem, consider these two alternative descriptions of the process of adaptation.
  • Adaptive evolution occurs when natural selection favors certain mutations which are beneficial as opposed to harmful. When new challenges arise, new adaptations arise as new beneficial mutations are generated and selection favors these mutations.
  • Adaptive evolution occurs when natural selection favors previously-existing genetic combinations that are more fit than others. When new challenges arise, new adaptations arise as selection favors individuals whose genetic endowments are best suited to the new challenges.
The first description probably sounds more familiar to you than the second one does, but they're quite different, and the second description is far more accurate than the first. The distinction between these two scenarios lies in the implication of the first scenario that new mutations must arise "on demand" or "just in time." Michael Behe's whole silly book is based on calculations that assume that new mutations must be generated, simultaneously, after the introduction of the new challenge. (His main example is the adaptation of the malaria parasite in the face of drugs intended for its destruction.) Those who promulgate this error (intentionally or not) tend to emphasize natural selection acting on mutations, and consequently it's easy to picture a species "mutating around" a challenge or obstacle. (Behe, for example, uses such language repeatedly.)

But that's a mistaken view of the process, and the way to avoid the trap is to picture selection acting on variation, specifically on variation that is always present in any population of organisms. (Populations without significant genetic variation, when confronted with serious challenges, are more likely to illustrate extinction than evolution.) Such variation is continuously generated and therefore continuously present. This is the lesson from studies of the effects of human selection on domesticated species of all kinds: when selection is applied, such populations typically reveal a remarkable propensity for rapid and dramatic change, because they harbor vast resources in the form of genetic diversity. If you carefully attend to this distinction, you will understand Darwinian evolution far better than any ID advocate.

2. Are most large-effect mutations harmful, and many small-effect mutations beneficial, as predicted by Fisher?

Well, first of all, kudos to Donald for reading Fisher. I've been browsing The Genetical Theory of Natural Selection, and it's demanding (but comprehensible). Michael Behe either hasn't read it, or didn't understand it, and in either case is therefore unqualified to write on evolutionary genetics.

Fisher was certainly right that large-effect mutations are almost never beneficial, but it is largely unknown whether very small-effect mutations are frequently beneficial, as he postulated. Theoretical and experimental work in this field has recently accelerated, and the current model is that effects of beneficial mutations are exponentially distributed, such that beneficial mutations are far more likely to be of very small effect than of large effect. This was Allen Orr's proposal, and it has been borne out in some very recent experimental analyses. The most recent, and significant, was the Science paper I mentioned above, in which the authors found that beneficial mutations in bacteria are far more common than previously estimated, but have relatively small effects (individually). Here's an excerpt from their last paragraph:
...our estimate of [the beneficial mutation rate] implies that 1 in 150 newly arising mutations is beneficial and that 1 in 10 fitness-affecting mutations increases the fitness of the individual carrying it. Hence, an enterobacterium has an enormous potential for adaptation and may help explain how antibiotic resistance and virulence evolve so quickly.
That's enough for now. Start with papers by Allen Orr when reading on the genetics of adaption; his historical overview in Nature Reviews Genetics in 2005 is particularly helpful.

3 comments:

Dennis Venema said...

another great post - thanks steve! I'll look forward to the rest of the series.

Anonymous said...

Thanks for replying.

I was always fascinated by Fisher's 50 percent argument--the logic is so clever. He works in phenotype space and assumes that some point in a multidimensional space represents the optimum phenotype, and the actual phenotype of some given organism is at another point. Then it's obvious that if you take a very small step from that second point(relative to its distance from the optimum), there's a 50 percent chance you'll come closer to the optimum. If you take a larger step the odds of coming closer drop and if you take a really large step you're certain to end up further away.

Of course, moving from the abstraction of geometry into the real world of biology makes me wonder if there's any way to test the argument. Apparently the papers you cite make some progress in that direction.

I glanced at the Orr paper on the exponential distribution of fitnesses among mutations, though I'd need to do much more than glance to really follow it (if then--I may not have enough of the relevant math background). He's working in terms of genotype rather than phenotype--Fisher is defining "small" mutations in terms of having "small" effects on the organism, whereas Orr is talking about single nucleotide changes and assumes that the wild type allele is already near the top of the fitness rank. Orr talks about phenotypes in another paper, I gather, but it's one in Nature Genetics that I can't read without paying 32 dollars. Sigh.

Anyway, I look forward to your future posts on this.

Donald

Stephen Matheson said...

Donald, email me and I might be able to solve your Nature Genetics problem.