The Extended Phenotype

by Richard Dawkins

  Where parasite genes exert shared power with host genes over the same host phenotypic characteristic, the confluence of the two powers may occur at any stage in the chain just described. Snail genes, and the genes of the fluke that parasitizes the snail, exert their power separately from each other at the cellular and even the tissue level. They influence the cytoplasmic chemistry of their respective cells separately, because they do not share cells. They influence tissue formation separately, because snail tissues are not intimately infiltrated by fluke tissues in the way that, say, the algal and fungal tissues of a lichen are intimate. Snail genes and fluke genes influence the development of organ systems, indeed of whole organisms, separately, because all the fluke cells are gathered together in one mass rather than being interspersed among snail cells. If fluke genes influence snail shell thickness, they do so by first collaborating with other fluke genes to make a whole fluke.

  Other parasites and symbionts more intimately infiltrate the systems of the host. At the extreme are the plasmids and other fragments of DNA which, as we saw in Chapter 9, literally insert themselves in the host chromosomes. It is impossible to imagine a more intimate parasite. ‘Selfish DNA’ itself is not more intimate, and indeed we may never know how many of our genes, whether junk’ or ‘useful’, originated as inserted plasmids. It seems to follow from the thesis of this book that there is no important distinction between our ‘own’ genes and parasitic or symbiotic insertion sequences. Whether they conflict or cooperate will depend not on their historical origins but on the circumstances from which they stand to gain now.

  Viruses have their own protein jacket, but they insert their DNA into the host’s cell. They are therefore in a position to influence the cellular chemistry of the host at an intimate level, if not quite such an intimate level as an insertion sequence in the host chromosome. Intracellular parasites in the cytoplasm, too, may be presumed to be in a position to exert considerable power over host phenotypes.

  Some parasites do not infiltrate the host at the cellular level, but at the tissue level. Examples are Sacculina, and many fungal and plant parasites, where parasite cells and host cells are distinct, but where the parasite invades the host’s tissues by means of an intricate and finely divided root system. The separate cells of parasitic bacteria and protozoa may infiltrate the host tissues with similarly comprehensive intimacy. To a slightly lesser extent than a cell parasite, such a ‘tissue parasite’ is in a strong position to influence organ development and gross phenotypic form and behaviour. Other internal parasites, such as the flukes we have been discussing, do not mix their tissues with those of the host, but keep their tissues to themselves and exert power only at the level of the whole organism.

  But we have not yet reached the end of our continuum of proximity. Not all parasites live physically inside their hosts. They may even seldom come into contact with their hosts. A cuckoo is a parasite in very much the same way as a fluke. Both are whole-organism parasites rather than tissue parasites or cell parasites. If fluke genes can be said to have phenotypic expression in a snail’s body, there is no sensible reason why cuckoo genes should not be said to have phenotypic expression in a reed warbler’s body. The difference is a practical one, and a rather smaller one than the difference between, say, a cellular parasite and a tissue parasite. The practical difference is that the cuckoo does not live inside the reed warbler’s body, so has less opportunity for manipulating the host’s internal biochemistry. It has to rely on other media for its manipulation, for instance sound waves and light waves. As discussed in Chapter 4, it uses a supernormally bright gape to inject its control into the reed warbler’s nervous system via the eyes. It uses an especially loud begging cry to control the reed warbler’s nervous system via the ears. Cuckoo genes, in exerting their developmental power over host phenotypes, have to rely on action at a distance.

  The concept of genetic action at a distance pushes our idea of the extended phenotype out to its logical culmination. That is where we must go in the next chapter.

  13 Action at a Distance

  Snail shells coil either to the right or to the left. Usually all individuals in one species coil the same way, but a few polymorphic species are to be found. In the Pacific island land snail Partula suturalis some local populations are right-handed, others are left-handed, and others are mixed in various proportions. It is therefore possible to study the genetics of directionality of coiling (Murray & Clarke 1966). When snails from right-handed populations were crossed with snails from left-handed populations, every offspring coiled the same way as its ‘mother’ (the parent that provided the egg: the snails are hermaphrodites). This might be thought to indicate a non-genetic maternal influence. But when Murray and Clarke crossed F1 snails with each other they obtained a curious result. All the progeny were left-handed, regardless of the direction of coiling of either parent. Their interpretation of the results is that coiling is genetically determined, with left-handedness dominant to right-handedness, but that an animal’s phenotype is controlled not by its own genotype but by its mother’s genotype. Thus the F1 individuals displayed the phenotypes dictated by their mothers’ genotypes, although all contained the same heterozygous genotypes since they were produced by mating two pure strains. Similarly, the F2 progeny of F1 matings all displayed the phenotype appropriate to an F1 genotype—left-handed since that is dominant and the F1 genotype was heterozygous. The underlying genotypes of the F2 generation presumably segregated in classic 3:1 Mendelian fashion, but this did not show itself in their phenotypes. It would have shown itself in the phenotypes of their progeny.

  Note that it is the mother’s genotype, not her phenotype, which controls her offspring’s phenotype. The F1 individuals themselves were left-handed or right-handed in equal proportion, yet all had the same heterozygous genotype, and all therefore produced left-handed offspring. A similar effect had been obtained earlier in the freshwater snail Limnaea peregra, though in that case right-handedness was dominant. Other such ‘maternal effects’ have long been known to geneticists. As Ford (1975) put it, ‘We have here simple Mendelian inheritance the expression of which is constantly delayed one generation.’ The phenomenon perhaps arises when the embryological event determining the phenotypic trait occurs so early in development as to be influenced by maternal messenger RNA from the egg cytoplasm, before the zygote has begun to manufacture its own messenger RNA. The direction of coiling in snails is determined by the initial direction of spiral cleavage, which occurs before the embryo’s own DNA has begun to work (Cohen 1977).

  This kind of effect provides a special opportunity for the kind of maternal manipulation of offspring that we discussed in Chapter 4. More generally, it is a special example of genetic ‘action at a distance’. It illustrates, in a particularly clear and simple manner, that the power of a gene may extend beyond the boundaries of the body in whose cells it sits (Haldane 1932b). We cannot hope that all genetic action at a distance will reveal itself in so elegant a Mendelian manner as in the case of the snails. Just as, in conventional genetics, the Mendelian major genes of the schoolroom are the tip of the iceberg of reality, so we may make conjectures about a polygenic ‘extended genetics’, a genetics in which action at a distance is rife but in which the effects of the genes are complex and interacting, and therefore difficult to sort out. Again as in conventional genetics, we do not necessarily have to do genetic experiments in order to infer the presence of genetic influence on variation. Once we have satisfied ourselves that a given characteristic is a Darwinian adaptation this, in itself, is tantamount to satisfying ourselves that variation in that character must at one time have had a genetic basis. If it had not, selection could not have preserved the advantageous adaptation in the population.

  One phenomenon that looks like an adaptation and which, in some sense, involves action at a distance, is the ‘Bruce Effect’. A female mouse who has just been inseminated by one male has her pregnancy blocked by exposure to chemical influence from a second male. T
he effect seems also to occur in a variety of species of mice and voles in nature. Schwagmeyer (1980) considers three main hypotheses of the adaptive significance of the Bruce Effect, but for the sake of argument I shall not, here, advocate the hypothesis that Schwagmeyer attributes to me—that the Bruce Effect represents a kind of female adaptation. Instead I shall look at it from the male point of view, and simply assume that the second male benefits himself by preventing the female’s pregnancy, thereby eliminating the offspring of a male rival, while at the same time bringing the female quickly into oestrus so that he can mate with her himself.

  I have expressed the hypothesis in the language of Chapter 4, the language of individual manipulation. But it can equally well be expressed in the language of the extended phenotype and genetic action at a distance. Genes in male mice have phenotypic expression in female bodies, in just the same sense as genes in mother snails have phenotypic expression in the bodies of their children. In the snail case, the medium of the action at a distance was assumed to be maternal messenger RNA. In the mouse case it is apparently a male pheromone. My thesis is that the difference between the two cases is not a fundamental one.

  Consider how an ‘extended geneticist’ might talk about the genetical evolution of the Bruce Effect. A mutant gene arose which, when present in the body of a male mouse, had phenotypic expression in the bodies of female mice with whom he came in contact. The route of action of the gene on its final phenotype was long and complex, but not noticeably more so than routes of genetic action within bodies customarily are. In conventional within-body genetics, the chain of causation leading from gene to observed phenotype may have many links. The first link is always RNA, the second is protein. A biochemist may detect the phenotype that interests him at this second link stage. Physiologists or anatomists will not pick up the phenotype that interests them until more stages have been passed. They will not concern themselves with the details of these earlier links in the chain, but will take them for granted. Whole-organism geneticists find it sufficient to do breeding experiments looking only at what, for them, is the final link in the chain, eye colour, crinkliness of hair, or whatever it is. The behaviour geneticist looks at an even more distant link—waltzing in mice, creeping-through mania in sticklebacks, hygiene in honeybees, etc. He arbitrarily chooses to regard a behaviour pattern as the end link in the chain, but he knows that the abnormal behaviour of a mutant is caused by, say, abnormal neuroanatomy, or abnormal endocrine physiology. He knows that he could have looked with a microscope at the nervous system in order to detect his mutants, but he preferred to look at behaviour instead (Brenner 1974). He made an arbitrary decision to regard observed behaviour as the end link in the chain of causation.

  Whichever link in the chain a geneticist chooses to regard as the ‘phenotype’ of interest, he knows that the decision was an arbitrary one. He might have chosen an earlier stage, and he might have chosen a later one. So, a student of the genetics of the Bruce Effect could assay male pheromones biochemically in order to detect the variation upon which to base his genetic study. Or he could look further back in the chain, ultimately to the immediate polypeptide products of the genes concerned. Or he could look later in the chain.

  What is the next later link in the chain after the male pheromone? It is outside the male body. The chain of causation extends across a gap into the female body. It goes through a number of stages in the female body, and once again our geneticist does not have to bother himself with the details. He chooses, for convenience, to end his conceptual chain at the point where the gene causes pregnancy blockage in females. That is the phenotypic gene-product which he finds most easy to assay, and it is the phenotype which is of direct interest to him as a student of adaptation in nature. Abortion in female mice, according to this hypothesis, is a phenotypic effect of a gene in male mice.

  How, then, would the ‘extended geneticist’ visualize the evolution of the Bruce Effect? The mutant gene which, when present in males, has the phenotypic effect in female bodies of causing them to abort, is favoured by natural selection over its alleles. It is favoured because it tends to be carried in the bodies of the offspring which the female bears after blocking her previous pregnancy. But, following the habit of Chapter 4, we now guess that females would be unlikely to submit to such manipulation without resistance, and that a kind of arms race might develop. In the language of individual advantage, selection would favour mutant females that resisted the pheromonal manipulation of the males. How would the ‘extended geneticist’ think about this resistance? By invoking the concept of the modifier gene.

  Once again, we turn first to conventional within-body genetics to remind ourselves of a principle, then carry that principle over into the realm of extended genetics. In within-body genetics we are quite used to the idea of more than one gene affecting variation in any given phenotypic character. Sometimes it is convenient to designate one locus as having the ‘major’ effect on the character, the others having ‘modifying’ effects. At other times no one locus predominates over the others sufficiently to be called major. All the genes may be thought of as modifying the effects of each other. In the chapter on ‘Outlaws and Modifiers’, we saw that two loci bearing on the same phenotypic character may be subject to conflicting selection pressures. The end result may be stalemate, compromise, or outright victory for one side or the other. The point is that conventional within-body genetics is already accustomed to thinking of the natural selection of genes at different loci bearing upon the same phenotypic character but in opposite directions.

  Apply the lesson in the extended genetics domain. The phenotypic trait of interest is abortion in female mice. The genes bearing upon it no doubt include a set of genes in the female’s own body, and also another set of genes in the male’s body. In the case of the male genes the links in the chain of causation include pheromonal action at a distance, and this may make the influence of the male genes seem very indirect. But the causal links in the case of the female genes are likely to be nearly as indirect, albeit they are confined inside her body. Probably they make use of various chemical secretions flowing in her bloodstream, whereas the male genes make use, in addition, of chemical secretions flowing in the air. The point is that both sets of genes, by long and indirect causal links, bear upon the same phenotypic character, abortion in the female, and either set of genes may be regarded as modifiers of the other set, just as some genes within each set may be regarded as modifiers of others within the same set.

  Male genes influence the female phenotype. Female genes influence the female phenotype, and also modify the influence of male genes. For all we know, female genes influence the male phenotype in counter-manipulation, in which case we expect the selection of modifiers among genes in males.

  This whole story could have been told in the language of Chapter 4, the language of individual manipulation. The language of extended genetics is not demonstrably more correct. It is a different way of saying the same thing. The Necker Cube has flipped. Readers must decide for themselves whether they like the new view better than the old. I suggest that the way the extended geneticist tells the story of the Bruce Effect is more elegant and parsimonious than the way the conventional geneticist would have told it. Both geneticists potentially have to contend with a formidably long and complex chain of causation, leading from gene to phenotype. Both admit that their choice of which link in the chain to designate as the phenotypic character of interest—earlier links being consigned to the embryologist—is arbitrary. The conventional geneticist makes the further arbitrary decision to cut off all chains at the point where they reach the outer wall of the body.

  Genes affect proteins, and proteins affect X which affects Y which affects Z which … affects the phenotypic character of interest. But the conventional geneticist defines ‘phenotypic effect’ in such a way that X, Y and Z must all be confined inside one individual body wall. The extended geneticist recognizes that this cut-off is arbitrary, and he is quite happy to allow h
is X, Y and Z to leap the gap between one individual body and another. The conventional geneticist takes in his stride the bridging of gaps between cells within bodies. Human red blood cells, for instance, have no nuclei, and must express the phenotypes of genes in other cells. So why should we not, when the occasion warrants it, conceive of the bridging of gaps between cells in different bodies? And when will the occasion warrant it? Whenever we find it convenient, and this will tend to be in any of those cases where, in conventional language, one organism appears to be manipulating another. The extended geneticist would, in fact, be quite happy to rewrite the whole of Chapter 4, fixing his gaze on the new face of the Necker Cube. I shall spare the reader any such rewriting, although it would be an interesting task to undertake. I shall not pile example on example of genetic action at a distance, but instead will discuss the concept, and problems that it raises, more generally.

  In the chapter on arms races and manipulation I said that an organism’s limbs might be adapted to work for the genes of another organism, and I added that this idea could not be made fully meaningful until later in the book. I meant that it could be made meaningful in terms of genetic action at a distance. So, what does it mean to say that a female’s muscles work for a male’s genes, or that a parent’s limbs work for its offspring’s genes, or that a reed warbler’s limbs work for a cuckoo’s genes? It will be remembered that the ‘central theorem’ of the selfish organism claims that an animal’s behaviour tends to maximize its own (inclusive) fitness. We saw that to talk of an individual behaving so as to maximize its inclusive fitness is equivalent to talking of the gene or genes ‘for’ that behaviour pattern maximizing their survival. We have now also seen that, in precisely the same sense as it is ever possible to talk of a gene ‘for’ a behaviour pattern, it is possible to talk of a gene, in one organism, ‘for’ a behaviour pattern (or other phenotypic characteristic) in another organism. Putting these three things together we arrive at our own ‘central theorem’ of the extended phenotype: An animal’s behaviour tends to maximize the survival of the genes ‘for’ that behaviour, whether or not those genes happen to be in the body of the particular animal performing it.