Evolution 2nd edition

Textbook by Mark Ridley, Blackwell Science Inc, November 1996 (Amazon.com)

Excerpt from Chapter 12 - "The Units of Selection"

12.3 Another sense of "unit of selection" is the entity whose frequency is adjusted directly by natural selection

Natural selection over the generations adjusts the frequencies of entities at all levels. We have implicitly seen this adjustment in the example of the lion hunt. If the lions of one pride become more efficient at hunting, perhaps because of some new behavioral trick, natural selection will favor them. If the trick is inherited, that type of lion will increase in frequency relative to other types of lion. All things associated with the trick will increase in frequency as well. The type of lion, its type of neurons, of proteins, and their encoding genes would all increase in frequency relative to their alternatives. When the hunting success of lions as a whole increases, the frequency of lions in the ecosystem will likewise increase too. Over geological time, lions might come to replace other competing predators on the plains. The question in this section focuses on whether natural selection directly adjusts the frequency of any of these units--nucleotides, genes, neurons, individual lions, lion pride, or lion species?

The answer was most clearly given by Williams, in Adaptation and Natural Selection (1966), and Dawkins, in The Selfish Gene (1976). It is at least implicit in all theoretical population genetics (and in the previous section of this chapter). For natural selection to adjust the frequency of something over the generations, the entity must have a sufficient degree of permanence. You cannot adjust the frequency of an entity between times t1 and t2 if between the two times the entity has ceased to exist. For a character to increase in frequency under natural selection, therefore, it must be inherited.

We can work through this argument using the example of an improvement in lion hunting skill. (We will express the improvement in terms of selection on a mutation; the same arguments apply when gene frequencies are being adjusted at a polymorphic locus.) When the improvement first appeared, it was a single genetic mutation. At a physiological level, the mutation would produce its effect by making some minor change in the lion's developmental program. After the mutation has appeared, a "pool" of two types exists--the new mutation, and the rest of the population (i.e., all alleles of the mutation, and the behavior patterns they produce). Genetic variation will, of course, develop at loci other than the one where the mutation arose, but that variation can be ignored because it will be randomly distributed among the mutant and non-mutant types. The lions with the mutation will survive better and produce more offspring. Natural selection is starting to work. Now we can ask whose frequency natural selection is adjusting. Is it lions? Lion genomes? Or the mutation?

Williams' and Dawkins' answer is the gene--more specifically, the mutation that produces improved hunting. Natural selection cannot work on whole lions because lions die; they are not permanent. Nor can it work on the genome. The mutant lion's offspring inherit only genetic fragments, not a copy of a whole genome, from their parents. Meiotic recombination breaks the genome. In Williams' expression, "meiosis and recombination destroy genotypes [i.e., genomes] as surely as death." What matters, in the process of natural selection, is that some of the lion's offspring inherit the mutation. These offspring, in turn, produce more offspring, and the gene increases in frequency. The gene can increase in frequency because it is not fragmented by meiosis (like the genome) or returned to dust by death (like the phenotype). The gene, in the form of copies of itself, is potentially immortal, and is at least Permanent enough to allow its frequency to be altered in successive generations.

Objections may be raised that recombination breaks genes as well as genomes. Recombination strikes at almost random intervals in the DNA and, therefore, could strike within the mutation with which we are concerned. A little reflection, however, shows that issue to be irrelevant. The information of the gene, not its physical continuity, is what matters. Consider the length of chromosome containing the gene and its mutant form. A number of polymorphic loci usually surround around the mutant locus (Figure 12.3a). Now consider what happens when recombination strikes in either a neighboring gene or the gene itself. Nearby recombination breaks the information in the chromosome--which simply means that recombination destroys the genome (Figure 12.3b). Recombination within the gene does not usually alter the outcome (Figure 12.3c). If the locus was homozygous before the mutation, all of the genes except for the mutant base pair will be identical in the original and mutant forms. Thus, intragenic recombination produces exactly the same result within the gene as no recombination, as it merely alters the combinations of genes.

Intragenic recombination can destroy the heritable information in a gene in one special circumstance. If the locus was heterozygous before the mutation, and recombination occurs between the mutant site and the site that differs between the two strands, the products of recombination will differ from the initial strands (Figure 12.4). When such an event occurs, the length of DNA whose information is inherited is shorter than a gene. For this reason, if we take a long enough view, the only finally permanent units in the genome are nucleotide bases, because recombination does not alter them. However, this long view holds Little interest for us, as we are concerned with the time scale of natural selection. It takes a thousand or so generations for a mutation's frequency to be significantly altered (section 5.6, p. 100) and, over this time, genes, but not genomes or phenotypes, will be practically unaltered. Genes will then act as units of selection-they will he permanent enough to have their frequency altered by natural selection.

Williams defined the gene to make it almost true by definition that the gene is the unit of selection. He defined the gene as "that which segregates and recombines with appreciable frequency." According to this definition, the gene need not be the same as a cistron (i.e., the length of DNA encoding one protein, or polypeptide). Rather, it is the length of chromosome that has sufficient permanence for natural selection to adjust its frequency. Longer lengths are broken by recombination and shorter lengths have no more permanence that the gene (for the reason shown in Figure 12.3). The gene on Williams definition is what Dawkins calls the replicator. In practice, the replicator (or Williams' gene) does not consistently correspond to any particular length of DNA.

When selection takes place at one locus, a cistron at a neighboring locus will to some extent (depending on the amount of recombination) have its frequency adjusted as a consequence. In a population genetic sense, this hitchhiking (section 8.9, p. 211) builds up linkage disequilibrium between genes. The same will be true of loci further down the DNA from the selected locus. Although the hitch-hiking effect is gradually reduced with distance by recombination, no clear cut-off has been distinguished. This situation poses no problem for Williams' definition of the gene. The neighboring allele that is hitchhiking with the selected mutation is, in Williams' definition, part of the gene that is having its frequency altered.

Williams' "gene" has a statistical reality, because shorter lengths of DNA are more permanent and longer lengths less permanent. The random hits of recombination will generate a frequency distribution of genome lengths lasting for different periods of evolutionary time. The average length that survives long enough to undergo the effects of natural selection has been defined by Williams and Dawkins as the gene. Population geneticists have scolded Williams and Dawkins from time to time for assuming a one-locus zero linkage disequilibrium view of evolution, but the dispute is a matter of definition, not substance. These critics identify "gene" with "cistron." It would be interesting to know whether the gene in Williams' sense is also a physical cistron, but this is a secondary question and it is unrelated to the fundamental logic of Williams' and Dawkins' argument.

We must discuss one other matter before considering the significance of the genic unit of selection. Critics, such as Gould, have objected that gene frequencies change between generations only in a passive, "bookkeeping" sense. The frequency changes provide a record of evolution, but are not its fundamental cause. True natural selection, the critics would say, happens at the level of organismic survival and reproduction. For instance, the actual selection in the lion example happens when a lion catches, or fails to catch, its prey. The differential hunting success drives the gene frequency changes, and it is a mistake to identify the gene frequency changes as causal. Williams and Dawkins, however, do not deny that the ecological processes causing differential organismic survival produce gene frequency changes within a generation. What they deny is that this ecological interaction of organisms means that natural selection directly adjusts the frequencies of organisms over the evolutionary time scale of many generations.

An easy philosophical method has been developed for deciding whether natural selection works on genes, or larger phenotypic units. We can consider a phenotypic change such as a new hunting skill, and ask whether natural selection can work on it if it is produced genically and if it is produced non-genically. In the lion's case, the skill is produced genically--the advantageous new hunting behavior was caused by a genetic mutation. Now suppose that the same advantageous phenotypic change was caused by a non-heritable phenotypic change, such as individual learning or some developmental accident in the lion's nervous system. The thought-based experiment provides a test case between the organismic, phenotypic, and genic accounts of evolution. In the genic case, we know that natural selection favors the improved hunting type and the gene for it increases in frequency. But what happens in the phenotypic case? The individual lion with improved hunting ability will survive and produce more offspring than an average lion, but no evolution, or natural selection in any interesting sense, will occur. The trait will not be passed on to the next generation. Natural selection cannot work directly on organisms.

The change in gene frequency over time, therefore, is not just a Passive "bookkeeping" record of evolution. Genes are crucial if natural selection is to take place. The need for inheritance, and the fact that acquired characters are not inherited, gives the gene a priority over the organism as a unit of selection. Whenever a gene is being selected, it produces a phenotypic change and the frequency of different organismal types will change concurrently with the gene frequency. The change in organism frequency is a consequence of the change in gene frequency, however. That is, natural selection actually works on the gene frequency. For this reason, Williams and Dawkins maintain that the gene represents the unit of selection.

The argument is more a matter of logic than a testable claim that could be refuted by facts or experiments. By no means do all biologists accept the argument. The main controversy has been concerned with the unit of selection in the first sense (see sections 12.1-12.2), but Williams and Dawkins' argument itself has also been criticized for the two reasons we have discussed. We believe that these criticisms have been misdirected. Confusion about the definition of "gene" has spurred arguments that hitchhiking and linkage disequilibrium mean that selection adjusts the frequency of larger units than the gene. In addition, the claim that natural selection "really" works on organisms rather than genes overlooks the importance of heredity.

Why does the argument matter? Its importance is to tell us for the benefit of which entities adaptations exist. Evolutionary biologists work on particular characters (like banding patterns in snails, and sex), trying to explain why the characters exist. The ultimate, abstract answer is that any adaptation exists because it increases the reproduction of the genes encoding it, relative to that of the alleles for alternative characters. The genes that exist in nature are the genes that in the past have out reproduced alternative alleles. Natural selection will always favor a character that increases the replication of the genes encoding it.

It is important to recognize the ultimate beneficiaries of adaptations. When we are attempting to explain the existence of particular characters, we need to know whether a proposed explanation is correct. The argument that genes are units of selection provides the fundamental logic behind relevant tests of these explanations. We imagine different genetic forms of the character, and the correct explanation must specify how the genes for the observed form of the character will out reproduce other genetic types. In practice, several possible hypotheses may be put forward, and they can be tested by the methods described in chapter 11. Before those methods are applied, however, we must ensure that the hypothesis make theoretical sense. We can rule out a hypothesis about adaptation before the practical testing stage if it contradicts the theory of gene selection.