The Extended Phenotype

by Richard Dawkins

  There is a sense in which a ‘vehicle’ is worthy of the name in inverse proportion to the number of outlaw replicators that it contains. The idea of a discrete vehicle maximizing a unitary quantity—fitness—depends on the assumption that the replicators that it serves all stand to gain from the same properties and behaviour of their shared vehicle. If some replicators would benefit from the vehicle’s doing act X, while other replicators would benefit from its doing act Y, that vehicle is correspondingly less likely to behave as a coherent unit. It will have the attributes of a human organization that is governed by a quarrelsome committee—pulled this way and that, and unable to show decisiveness and consistency of purpose.

  There is a superficial analogy here with group selection. One of the problems of the theory that groups of organisms function as effective gene vehicles is that outlaws (from the group’s point of view) are very likely to arise and be favoured by selection. If we are hypothesizing the evolution of individual restraint through group selection, a gene that makes individuals behave selfishly in an otherwise altruistic group is analogous to an outlaw. It is the near inevitability of such ‘outlaws’ arising that has dashed the hopes of many a group selection modeller.

  The individual body is a much more persuasive gene vehicle than the group because, among other reasons, outlaw replicators within the body are not very likely to be strongly favoured over their alleles. The fundamental reason for this is the drilled formality of the mechanisms of individual reproduction, the ‘gavotte of chromosomes’ as Hamilton (1975b) called it. If all replicators ‘know’ that their only hope of getting into the next generation is via the orthodox bottleneck of individual reproduction, all will have the same ‘interests at heart’; survival of the shared body to reproductive age, successful courtship and reproduction of the shared body, and a successful outcome to the parental enterprise of the shared body. Enlightened self-interest discourages outlawry when all replicators have an equal stake in the normal reproduction of the same shared body.

  Where reproduction is asexual the stake is equal and total, for all replicators have the same 100 per cent chance of finding themselves in every child produced by their joint efforts. Where reproduction is sexual the corresponding chance for each replicator is only half as great, but the ritualized courtliness of meiosis, Hamilton’s ‘gavotte’, largely succeeds in guaranteeing to each allele an equal chance of reaping the benefits of the success of the joint reproductive enterprise. It is, of course, another question why the gavotte of the chromosomes is so courtly. It is an immensely important question which I shall evade on a simple plea of cowardice. It is one of a set of questions about the evolution of genetic systems with which better minds than mine have wrestled more or less unsuccessfully (Williams 1975, 1980; Maynard Smith 1978a), a set of questions which moved Williams to remark that ‘there is a kind of crisis at hand in evolutionary biology’. I don’t understand why meiosis is the way it is, but given that it is much follows. In particular the organized fair-dealing of meiosis helps to account for the coherence and harmony which unites the parts of an individual organism. If, at the level of the group of individuals as potential vehicle, the privilege of reproduction was granted with the same scrupulous probity in a similarly well-disciplined ‘gavotte of the organisms’, group selection might become a more plausible theory of evolution. But, with the possible exception of the very special case of the social insects, group ‘reproduction’ is anarchical and favourable to individual outlawry. Even social insect colonies will never seem fully harmonious again after Trivers and Hare’s ingenious analysis of sex ratio conflicts (see Chapter 4).

  This consideration tells us where we should look first if we want to discover outlaws within the individual body as vehicle. Any replicator that managed to subvert the rules of meiosis so that it enjoyed more than the ordained 50 per cent chance of ending up in a gamete would, other things being equal, tend to be favoured over its alleles in natural selection. Such genes are known to geneticists under the name of meiotic drive genes or segregation distorters. I have already used them as an example to illustrate my definition of an outlaw.

  ‘Genes that beat the system’

  The account of segregation distorters that I shall mainly follow is that of Crow (1979), who uses language congenial to the spirit of this book. His paper is called ‘Genes that violate Mendel’s rules’, and it ends as follows: ‘The Mendelian system works with maximum efficiency only if it is scrupulously fair to all genes. It is in constant danger, however, of being upset by genes that subvert the meiotic process to their own advantage … There are many refinements of meiosis and sperm formation whose purpose is apparently to render such cheating unlikely. And yet some genes have managed to beat the system.’

  Crow suggests that segregation distorters may be much more common than we ordinarily realize, for the methods of geneticists are not well geared to detecting them, especially if they produce only slight, quantitative effects. The SD genes in Drosophila are particularly well studied, and here there is some indication as to the actual mechanism of distortion. ‘While the homologous chromosomes are still paired up during meiosis, the SD chromosome might do something to its normal partner (and rival) that later causes a dysfunction of the sperm receiving the normal chromosome … SD might actually break the other chromosome’ (Crow 1979, my macabre emphasis). There is evidence that, in SD-heterozygous individuals, sperms not containing the SD chromosome have abnormal, and presumably faulty, tails. It might be thought, then, that the faulty tail results from some sabotage to the non-SD chromosome in the sperms containing it. This cannot be the whole story, as Crow points out, because sperms have been shown to be capable of developing normal tails without any chromosomes at all. Indeed, the whole sperm phenotype seems usually to be under the control of the diploid genotype of the father, not its own haploid genotype (Beatty & Gluecksohn-Waelsch 1972; see below). ‘The effect of the SD chromosome on its homologue cannot, then, be simply to inactivate some function, because no function is required. SD must somehow induce its partner to commit a positive act of sabotage.’

  Segregation distorters prosper when rare, because the chances are then good that their victims are alleles, not copies. When common, the distorter tends to occur homozygously, and therefore sabotages copies of itself, making the organism virtually sterile. The story is more complicated than this, but computer simulations described by Crow suggest that a stable proportion of segregation distorter genes will be maintained at a frequency somewhat greater than would be accounted for by recurrent mutation alone. There is some evidence that this is so in real life.

  In order to qualify as an outlaw, a segregation distorter must do harm to most of the rest of the genome, not just to its alleles. Segregation distorters could have this effect by reducing the total gamete count of the individual. Even if they did not do this, there is a more general reason for expecting there to be selection at other loci in favour of suppressing them (Crow 1979). The argument needs to be developed stage by stage. Firstly many genes, when compared with their alleles, have several pleiotropic effects. Lewontin (1974) goes so far as to speak of ‘… the undoubted truth that every gene affects every character …’. While to call this an ‘undoubted truth’ may be, to put it mildly, an enthusiastic exaggeration, for my purposes I need only assume that most new mutations have several pleiotropic effects.

  Now it is reasonable to expect that most such pleiotropic effects will be deleterious—mutational effects usually are. If a gene is favoured by selection resulting from one beneficial effect, this will be because the advantages of its beneficial effect quantitatively outweigh the disadvantages of its other effects. Normally, by ‘beneficial’ and ‘deleterious’, we mean beneficial and deleterious to the whole organism. In the case of a segregation distorter, however, the beneficial effect we are talking about is beneficial to the gene alone. Any pleiotropic effects that it may have on the body are pretty likely to be deleterious to the whole body’s survival and reproduction. S
egregation distorters are therefore, on the whole, likely to be outlaws: we expect that selection will favour genes at other loci whose phenotypic effect is to reduce the segregation distortion. This brings us to the topic of modifiers.

  Modifiers

  The classic proving ground for the theory of modifier genes was R. A. Fisher’s explanation of the evolution of dominance. Fisher (1930a, but see Charlesworth 1979) suggested that the beneficial effects of a given gene tend to become dominant through the selection of modifiers, while its deleterious effects tend to become recessive. He noted that dominance and recessiveness are not properties of genes themselves, but properties of their phenotypic effects. Indeed a given gene can be dominant in one of its pleiotropic effects and recessive in another. A phenotypic effect of a gene is the joint product of itself and its environment, an environment which includes the rest of the genome. This interactive view of gene action, which Fisher had to argue for at length in 1930, had become so well accepted by 1958 that he was able to take it for granted in the second edition of his great book. It follows from it that dominance or recessiveness, like any other phenotypic effect, may itself evolve, through the selection of other genes elsewhere in the genome, and this was the basis of Fisher’s theory of dominance. Although these other genes are known as modifiers, it is now realized that there is no separate category of modifier genes as distinct from major genes. Rather, any gene may serve as modifier of the phenotypic effects of any other gene. Indeed, any given gene’s phenotypic effects are susceptible to modification by many other genes in the genome, genes which themselves may have many other major and minor effects (Mayr 1963). Modifiers have been invoked for various other theoretical purposes, for instance in the Medawar/Williams/Hamilton progression of theories of the evolution of senescence (Kirkwood & Holliday 1979).

  The relevance of modifiers to the subject of outlaw genes has already been alluded to. Since any gene’s phenotypic effects may be subject to modification by genes at other loci, and since outlaws, by definition, work to the detriment of the rest of the genome, we should expect selection to favour genes that happen to have the effect of neutralizing the outlaw’s deleterious effects on the body as a whole. Such modifiers would be favoured over alleles that did not influence the outlaw’s effects. Hickey and Craig (1966), studying a sex-ratio distorting gene (see below) in the yellow-fever mosquito Aedes aegypti, found evidence for evolutionary diminution of the distorting effect which could be interpreted as resulting from the selection of modifiers (though their own interpretation was slightly different). If outlaws do, in general, call forth the selection of suppressing modifiers, there will presumably be an arms race between each outlaw and its modifiers.

  As in any other arms race (Chapter 4), we now ask whether there is any general reason to expect one side to prevail over the other. Leigh (1971, 1977), Alexander and Borgia (1978), Kurland (1979, 1980), Hartung (in press), and others suggest that there is just such a general reason. Since, for any single outlaw, suppressing modifiers may arise anywhere in the genome, the outlaw will be outnumbered. As Leigh (1971) puts it, ‘It is as if we had to do with a parliament of genes: each acts in its own self-interest, but if its acts hurt the others, they will combine together to suppress it … However, at loci so closely linked to a distorter that the benefits of “riding its coattails” outweigh the damage of its disease, selection tends to enhance the distortion effect. Thus a species must have many chromosomes if, when a distorter arises, selection at most loci is to favor its suppression. Just as too small a parliament may be perverted by the Cabals of a few, a species with only one, slightly linked chromosome is an easy prey to distorters’ (Leigh 1971, p. 249). I am not sure what I think about Leigh’s point about chromosome numbers, but his more general point that there is some sense in which outlaws may be ‘outnumbered’ (Alexander & Borgia 1978, p. 458) by their modifiers seems to me to have promise.

  I suppose ‘outnumbering’ could, in practice, work in two main ways. Firstly, if different modifiers each cause a quantitative diminution of the outlaw’s effect, several modifiers might combine additively. Secondly, if any one of several modifiers would suffice to neutralize the outlaw, the chance of effective neutralization goes up with the number of available modifier loci. Alexander and Borgia’s metaphor of ‘outnumbering’, and Leigh’s metaphor of the power of the collective in a ‘parliament’ of the many, could be given meaning in either or both of these two ways. It is important for the argument that segregation distorters at different loci could not, in any obvious sense, ‘pool their efforts’. They are not working for some common end of ‘general segregation distortion’. Rather, each one is working to distort segregation in favour of itself, and this will hurt other segregation distorters just as much as it hurts non-distorters. Suppressors of segregation distorters, on the other hand, could, in a sense, pool their efforts.

  The parliament of genes is one of those metaphors which, if we are not careful, tricks us into thinking that it explains more than it does. Like all humans, but unlike genes, human Members of Parliament are highly sophisticated computers capable of using foresight and language to conspire and reach agreements. Outlaws may seem to be suppressed by agreement of a collective in the parliament of genes, but what is really going on is the selection of modifier genes in preference to non-modifying alleles at their respective loci. Needless to say, Leigh and the other advocates of the ‘parliament of genes’ hypothesis are well aware of this. I now want to extend the list of outlaws.

  Sex-linked outlaws

  If a segregation distorter occurs on a sex chromosome, it is not only an outlaw in conflict with the rest of the genome and therefore subject to suppression by modifiers: it also, incidentally, threatens the whole population with extinction. This is because, in addition to ordinary detrimental side effects, it also distorts the sex ratio, and may even eliminate one sex from the population altogether. In one of Hamilton’s (1967) computer simulations, a single mutant male with a ‘driving Y’ chromosome causing males to have only sons and no daughters, was introduced into a population of 1000 males and 1000 females. It took only fifteen generations to drive the model population extinct for lack of females. Something like this effect has been demonstrated in the laboratory (Lyttle 1977). The possibility of using a driving Y gene in the control of serious pests like the yellow-fever mosquito did not escape Hickey and Craig (1966). It is a method with sinister elegance because it is so cheap: all the work of spreading the pest control agent is done by the pests themselves together with natural selection. It is like ‘germ warfare’ except that the lethal ‘germ’ is not an extraneous virus but a gene in the species’s own gene-pool. Perhaps the distinction is not a fundamental one anyway (Chapter 9).

  X-linked drive is likely to have the same sort of detrimental effect on populations as Y-linked drive, but tends to take more generations to extinguish the population (Hamilton 1967). The driving gene on an X chromosome causes males to have daughters rather than sons (except in birds, Lepidoptera, etc.). As we saw in Chapter 4, if haploid male Hymenoptera could influence the amount of care devoted to their spouses’ offspring, they would favour daughters rather than sons, since males pass no genes on to sons. The mathematics of this situation are analogous to the case of X-linked drive, the whole genome of the male hymenopteran functioning like an X chromosome (Hamilton 1967, p. 481 and footnote 18).

  It is often the case that X chromosomes cross over with each other but not with Y chromosomes. It follows that all genes on X chromosomes could stand to gain from the presence in the gene-pool of a driving X gene which distorts gametogenesis in the heterogametic sex in favour of X gametes and against Y gametes. Genes on X chromosomes are, in a sense, united against Y genes, in a kind of ‘anti linkage group’, simply because they have no chance of finding themselves on a Y chromosome. Modifiers to suppress X-linked meiotic drive in the heterogametic sex might well not be favoured if they arose at other loci on X chromosomes. They would be favoured if they arose on autosomes. This
is different from the case of segregation distorters on autosomes: here there might well be selection in favour of suppressing them by modifiers at other loci even on the same chromosome. X-linked distorters affecting gamete production in the heterogametic sex are, then, outlaws from the point of view of the autosomal part of the gene-pool, but not from the point of view of the rest of the X-chromosomal part of the gene-pool. This potential ‘solidarity’ among the genes on sex chromosomes suggests that the concept of the outlaw gene may be too simple. It conveys the image of a single rebel standing out against the rest of the genome. At times we might, instead, do better to think in terms of wars between rival gangs of genes, for instance the X-chromosome genes against the rest. Cosmides and Tooby (1981) coin the useful term ‘coreplicon’ for such a gang of genes that replicate together and therefore tend to work for the same ends. In many cases neighbouring coreplicons will blur into each other.

  Ganging up by Y-chromosome genes is even more to be expected. As long as Y chromosomes do not cross over, it is clear that all genes on a Y chromosome stand to gain from the presence of a Y-linked distorter every bit as much as the distorter gene itself. Hamilton (1967) made the interesting suggestion that the reason for the well-known inertness of Y chromosomes (hairy ears seems to be the only conspicuous Y-linked trait in man) is that Y-suppressing modifiers have been positively selected elsewhere in the genome. It is not obvious how a modifier might go about suppressing the phenotypic activity of an entire chromosome, since the various phenotypic effects of a single chromosome are usually so heterogeneous. (Why would selection not suppress only the effects of the driving genes, leaving other Y-linked effects intact?) I suppose it might do it by physically deleting large chunks of Y chromosome, or by contriving to quarantine the Y chromosome from the cell’s transcription machinery.