Falls Climbing Mount Improbable:
Mutator Genes and Natural Selection

In Chapter 3 of his last book, "Climbing Mount Improbable", Richard Dawkins makes three questionable assessments. The first is asserting that "mutator genes (genes that promote an increased mutation rate; parenthesis mine) will always tend to disappear from the population". This statement is prompted by another with which I differ in opinion: namely, that the fact that a "gene’s long-term fate depends on its average effects: its effects averaged over all the different bodies in which it finds itself, over the long term", implies that if a gene has one genetic context in which it is favorable and many in which it is unfavorable or neutral, it will be overall disadvantageous and disappear. He then concludes -I believe erroneously- that "natural selection favors a mutation rate of zero".

Let us start with the second concept. Dawkins argues that if the effects of a gene on most bodies are deleterious, the gene will disappear, because the average effect of the gene is bad. The first fallacy in this argument is that the effects of the gene on the one or few bodies (genotypes) in which it is good can be so good as to make the average effect good despite the many bodies in which the effect is bad. In the limit case of the "good genotype + gene in question" combination being fundamental for survival, only the effects of the gene in those individuals matter. The second and most important fallacy is that this average effect is not constant over generations because favorable genotypes are more likely to survive, and thus are increasingly overrepresented as generations go by. Hence, the average effect of the gene will be better each generation, because of a gradual increase in the proportion of individuals carrying a genotype that is favorable in combination with the gene in question. Thus, a gene’s fate is mostly determined by its effects in the best genetic context it finds itself in. If this were not so, animals would never have developed genes which are only beneficial in the context of many other genes whose products interact with the product of the gene in question. The marvelous consequence of natural selection is that it does not matter that most of the bodies in which a gene is expressed are unfit, as long as there is one genotype in which that gene is beneficial that survives. Natural selection will then cause the genotype of that body to tend to be more represented in the next generation. A gene for a receptor does not disappear from a population because it is useless, or even deleterious, in all individuals except those possessing the gene for its agonist. If the combination is beneficial, the genotype containing both will tend to endure if it appears. The example in the next paragraph may help explain this concept.

Taking the above into account, let us tackle the first concept. Even if a mutator gene is bad for the average body, it will tend to be kept if it ever finds itself in one favorable genetic environment. A mutator gene may be regulated so that it is only expressed in response to an environmental crisis -this possibility is suggested by Dawkins himself, who calls the indicator of an environmental crisis stress. In this case, the mutator gene will not be expressed in an organism that underwent a mutation -as compared with its parents- and which is well-adapted to the cause of stress, provided we take the definition of stress to be an internal one: a condition of an animal which is negatively correlated with chance of survival. Such a stress measure could derive from blood oxygen or glucose contents, metabolic rate, a combination of the above, etc. Because the new, mutated, adapted organism would not be under stress, the mutator gene would not be expressed. Its descendants would not express the mutator gene unless another, different, environmental crisis occurred. The mutator gene would tend to be preserved, because its effect would be neutral in favorable conditions and good in unfavorable conditions. In other words, if there were two genotypes, one with the gene and one without it, the one with it would be more likely to survive a drastic change in environmental conditions. So even if the mutator gene’s average effect is bad, if it ever finds its way into one genetic regulatory network that ensures its expression only during internal stress, it is likely to endure. The question that arises with such a gene, whose effect is only present in occasional generations, is whether a copy of it would endure long periods of no selective pressure - because of no environmental crises - sufficiently unchanged, so that when an environmental crisis occurred, there would still be working copies of the gene in the population to be favored by natural selection. This is an empirical question, and depends on the frequency of such environmental crises. If the mutator gene is necessary enough during periods of environmental crisis, then we can say that only those species that did preserve working copies of the gene at the moments of the crises survived. So that even if that preservation is a rare case, existing species will be biased toward having the gene. Because a lot of species do disappear during extreme environmental changes, an advantage during those times of disaster may well have an important selective advantage. I call this a selective advantage, as opposed to a competitive advantage, because the populations to be selected amongst need not be in competition for a common ecological niche. The kind of selection at work here is not only between individuals competing for limited resources, but between populations, only those of which that can outlast such crises will survive. The disappearance of species is not necessarily a consequence of another species consuming away its resources; it can be the result of no member of a species having what it takes to survive a change in the environment. This implies that existing species will only include those which have always carried the equipment -or luck- it takes to survive such crises. The importance of catastrophic times in determining the genotypes that populate our planet will be made clearer in the next paragraph.

There is another, different, sense in which it is not the average "climate of companion genes", as Dawkins aptly puts it, that determines whether a gene survives in the long term. Just as the selection process on all coexisting members of a population is a parallel process, and requires just one of the parallel branches to succeed, the selection process on one of these branches through time is a serial one, and it is sufficient for the gene to do badly in one generation for it to be wiped out of history forever after. In this sense, it is the effect of a gene in the generation with the worst "climate" that determines its long-term fate. This fact is well illustrated by Dawkins’ statement that all of the planet’s inhabitants possess the remarkable family history of an uninterrupted succession of predecessors who all lived beyond reproductive age, found a willing mate and mated successfully.

The third concept is perhaps the most important. To say that natural selection favors a mutation rate of zero is tantamount to underscoring the vital importance of variation for the effective operation of natural selection. Natural selection operates by selectively amplifying the individuals with greatest fitness in a population. The amplification of this subpopulation, coupled with the generation of new variation in it, permits the next generation to produce some individuals which are even fitter than the best of the previous generation. This is what Dawkins refers to when he speaks of cumulative improvement. Without variation, natural selection has nothing to work with; there is no selecting to be done if all candidates are equal. Cumulative improvement cannot occur without variation. Thus, individuals with no variation programmed into their genomes, no matter how fit to one environment at a given time, will eventually be selected against because of the inability of their genes to adapt to a change in the environment. That is the reason why we do not find mutation rates of zero, not that the "genetic nirvana (of a mutation rate of zero) is fortunately never quite attained". Natural selection favors an intermediate rate of mutation, not so low that variation is abolished, and not so high that all progeny lose viability.

True, variation could occur due to sexual recombination rather than mutation. But there are only a relatively very limited set of genes that can be given birth to through the recombination of any given preexisting set of genes. There is no way a protein as well-adapted to its function as hemoglobin could have arisen exclusively through recombination of preexisting genes for myoglobin, for example; its present form is the result of gradual improvement through mutation. And, in any case, the preexisting genes must have arisen through mutation in the first place. A genome is more likely to be well adapted to a changing environment if it is endowed with the machinery for generating small changes in its genes through mutation.

There is also empirical evidence that the rate of mutation can be actively regulated by natural selection, as much to make it high as to make it low. For example, human B-lymphocytes that have been activated by binding to their target antigens will undergo a hypermutation process selectively in the sequences coding for the variable region of their antibodies, during which the mutation rate is about a million times the spontaneous mutation rate in other genes. This is an example of just the kind of regulated mutator gene described above, which would be deleterious if segregated from its regulatory elements, but has adaptive value in the genetic context in which it is found and has thus been selected for. The human immunodeficiency virus (HIV) provides another example: strains with a high mutation rate have been selected because their constant change allows them to elude their main enemy, the human immune system.

Alejandro Bäcker

Division of Biology, 139-74

California Institute of Technology

Pasadena, CA 91125

Originally published online sometime between 1995 and 1999; previous version at http://alexbacker.com/caltech/FallsClimbingMountImprobable.html, with the original version on Caltech's website until after I finished my Ph.D. and therefore had to give up my account.