The Extended Phenotype

by Richard Dawkins

  Why … should a plant make the gall such a perfect domicile for an insect that is its enemy? Actually we are dealing here with two selection pressures. On the one hand, selection works on a population of gall insects and favors those whose gall-inducing chemicals stimulate the production of galls giving maximum protection to the young larva. This, obviously, is a matter of life or death for the gall insect and thus constitutes a very high selection pressure. The opposing selection pressure on the plant is in most cases quite small because having a few galls will depress viability of the plant host only very slightly. The ‘compromise’ in this case is all in favor of the gall insect. Too high a density of the gall insect is usually prevented by density-dependent factors not related to the plant host.

  Mayr is here invoking the equivalent of the ‘life/dinner principle’ to explain why the plant does not fight back against the remarkable manipulation by the insect. It is necessary for me to add only this. If Mayr is right that the gall is an adaptation for the benefit of the insect and not the plant, it can have evolved only through the natural selection of genes in the insect gene-pool. Logically, we have to regard these as genes with phenotypic expression in plant tissue, in the same sense as some other gene of the insect, say one for eye colour, can be said to have phenotypic expression in insect tissue.

  Colleagues with whom I discuss the doctrine of the extended phenotype repeatedly come up with the same entertaining speculations. Is it just an accident that we sneeze when getting a cold, or could it be a result of manipulation by viruses to increase their chances of infecting another host? Do any venereal diseases increase the libido, even if only by inducing an itch, like extract of Spanish fly? Do the behavioural symptoms of rabies infection increase the chance of the virus being passed on (Bacon & Macdonald 1980)? ‘When a dog gets rabies, its temper changes quickly. It is often more affectionate for a day or two, and given to licking its human contacts, a dangerous practice, for the virus is already in its saliva. Soon it grows restless and wanders off, ready to bite anyone that gets in its way’ (Encyclopaedia Britannica 1977). Even non-carnivorous animals are driven by the rabies virus to vicious biting, and there are recorded cases of humans contracting the disease from the bites of normally harmless fruit-eating bats. Apart from the obvious power of biting to spread a saliva-borne virus, ‘restless wandering’ might very well serve to spread the virus more effectively (Hamilton & May 1977). That the widespread availability of cheap air travel has had a dramatic impact on the spread of human disease is obvious: dare we wonder whether the phrase ‘travel bug’ might have a more than metaphorical significance?

  The reader will probably, like me, find such speculations far-fetched. They are intended only as light-hearted illustrations of the kind of thing that might go on (see also Ewald, 1980, who draws attention to the medical significance of this kind of thinking). All I really need to establish is that in some examples host symptoms can properly be regarded as parasitic adaptation; say, for instance, the Peter Pan syndrome in Tribolium induced by protozoan-synthesized juvenile hormone. Given such an admitted parasite adaptation, the conclusion I wish to draw is not really disputable. If host behaviour or physiology is a parasite adaptation, there must be (have been) parasite genes ‘for’ modifying the host, and the host modifications are therefore part of the phenotypic expression of those parasite genes. The extended phenotype reaches out of the body in whose cells the genes lie, reaches out to the living tissues of other organisms.

  The relationship of Sacculina gene to crab body is not in principle different from the relationship of caddis gene to stone, nor indeed different from the relationship of human gene to human skin. This is the first of the points that I intended to establish in this chapter. It has the corollary, which I have already emphasized in other terms in Chapter 4, that the behaviour of an individual may not always be interpretable as designed to maximize its own genetic welfare: it may be maximizing somebody else’s genetic welfare, in this case that of a parasite inside it. In the next chapter we shall go further, and see that some of the attributes of an individual may be regarded as phenotypic expression of genes in other individuals who need not necessarily be parasites inside.

  The second point of this present chapter is that the genes that bear upon any given extended phenotypic trait may be in conflict rather than in concert with one another. I could talk in terms of any of the examples given above, but I shall stick to one, the case of the snail shell thickened by the influence of a fluke. To recapitulate that story in slightly different terms, a student of snail genetics and a student of fluke genetics might each look at the same phenotypic variation, variation in snail shell thickness. The snail geneticist would partition the variance between a genetic and an environmental component, by correlating the thickness of shells in parent snails and their offspring. The fluke geneticist would independently partition the same observed variance into a genetic and an environmental component, in his case by correlating the shell thickness of snails containing particular flukes with the shell thickness of snails containing the offspring of those same flukes. As far as the snail geneticist is concerned, the fluke contribution is part of what he calls the ‘environmental’ variation. Reciprocally, for the fluke geneticist, variation due to snail genes is ‘environmental’ variation.

  An ‘extended geneticist’ would acknowledge both sources of genetic variation. He would have to worry about the form of their interaction—is it additive, multiplicative, ‘epistatic’, etc.?—but in principle such worries are already familiar to both the snail geneticist and the fluke geneticist. Within any organism, different genes influence the same phenotypic traits, and the form of the interaction is a problem with respect to genes within one normal genome just as much as it is for genes in an ‘extended’ genome. Interactions between the effects of snail gene and fluke gene are, in principle, no different from interactions between the effects of one snail gene and another snail gene.

  And yet, it may be said, is there not a rather important difference? Snail gene may interact with snail gene in additive, multiplicative, or any other ways, but do they not both have the same interests at heart? Both have been selected in the past because they worked for the same end, the survival and reproduction of the snails that bore them. Fluke gene and fluke gene, too, are working for the same end, the reproductive success of the fluke. But snail gene and fluke gene do not have the same interests at heart. One is selected to promote snail reproduction, the other to promote fluke reproduction.

  There is truth in the protestations of the last paragraph, but it is important to be clear about exactly where that truth lies. It is not that there is some obvious trade-union spirit uniting fluke genes against a rival union of snail genes. To persist with this harmless anthropomorphism, each gene is fighting only its alleles at the same locus, and it will ‘unite’ with genes at other loci only insofar as doing so assists it in its selfish war against its own alleles. A fluke gene may ‘unite’ with other fluke genes in this way but, equally, if it was convenient to do so, it might unite with particular snail genes. And if it remains true that snail genes are in practice selected to work together with each other and against an opposing gang of fluke genes, the reason is only that snail genes tend to gain from the same events in the world as do other snail genes. Fluke genes stand to gain from other events. And the real reason why snail genes stand to gain from the same events as each other, while fluke genes stand to gain from a different set of events, is simply this: all snail genes share the same route into the next generation—snail gametes. All fluke genes, on the other hand, must use a different route, fluke cercariae, to get into the next generation. It is this fact alone that ‘unites’ snail genes against fluke genes and vice versa. If it were the case that the parasite genes passed out of the host’s body inside the host’s gametes, things might be very different. The interests of host genes and parasite genes might not be quite identical, but they would then be very much closer than in the case of fluke and snail.

&
nbsp; It follows from the extended phenotype view of life, then, that crucial importance attaches to the means by which parasites propagate their genes out of a given host into a new host. If the parasite’s means of genetic exit from the host’s body is the same as the host’s, namely the host’s gametes or spores, there will be relatively little conflict between the ‘interests’ of parasite and host genes. For example, both would ‘agree’ about the optimum thickness of the host’s shell. Both will be selected to work not only for host survival but for host reproduction as well, with all that that entails. This might include host success in courtship and even, if the parasites aspire to be ‘inherited’ by the host’s own offspring, host success in parental care. Under such circumstances the interests of parasite and host would be likely to coincide to such an extent that it might become difficult for us to detect that a separate parasite existed at all. It is clearly of great interest for parasitologists and ‘symbiologists’ to study such very intimate parasites or symbionts, symbionts with an interest in the success of their host’s gametes as well as in the survival of their host’s body. Some lichens might be promising examples, and so might bacterial endosymbionts of insects which are transmitted transovarially and in some cases seem to influence the host sex ratio (Peleg & Norris 1972).

  Mitochondria, chloroplasts, and other cell organelles with their own replicating DNA might also be good candidates for study in this connection. A fascinating account of cell organelles and microorganisms seen as semi-autonomous symbionts inhabiting a cellular ecology is given in the proceedings of a symposium entitled The Cell as a Habitat, edited by Richmond and Smith (1979). The closing words of Smith’s introductory chapter are particularly memorable and apposite: ‘In non-living habitats, an organism either exists or it does not. In the cell habitat, an invading organism can progressively lose pieces of itself, slowly blending into the general background, its former existence betrayed only by some relic. Indeed, one is reminded of Alice in Wonderland’s encounter with the Cheshire Cat. As she watched it, “it vanished quite slowly, beginning with the tail, and ending with the grin, which remained some time after the rest of it had gone”’ (Smith 1979). Margulis (1976) gives an interesting survey of all the degrees of vanishing of the grin.

  Richmond’s (1979) chapter, too, is very congenial to the present thesis: ‘It is conventional to regard cells as units of biological function. Another view, particularly apposite to this symposium, is that the cell is the minimal unit capable of replicating DNA … Such a concept places DNA at the centre of biology. Thus DNA is not regarded simply as a hereditary means of ensuring the long term survival of the organisms of which it forms part. Rather, it stresses that the primary role of cells is to maximize the amount and diversity of DNA in the biosphere …’ This last remark, incidentally, is unfortunate. Maximizing the amount and diversity of DNA in the biosphere is the concern of nobody and nothing. Rather, each small piece of DNA is selected for its power to maximize its own survival and replication. Richmond goes on: ‘If the cell is considered as a unit for the replication of DNA, it follows that DNA additional to that required for the duplication of the cell may also be carried; molecular parasitism, symbiosis and mutualism may occur at the DNA level, as it does at higher organizational levels in biology.’ We have arrived back at the concept of ‘selfish DNA’ which was a subject of Chapter 9.

  It is interesting to speculate on whether mitochondria, chloroplasts, and other DNA-bearing organelles originated from parasitic prokaryotes (Margulis 1970, 1981). But, important as that question is as a matter of history, it does not bear, one way or the other, on my present concern. I am here interested in whether mitochondrial DNA is likely to work for the same phenotypic ends as nuclear DNA, or whether it is likely to be in conflict with it. This should depend not on the historic origins of mitochondria but on their present method of propagating their DNA. Mitochondrial genes are passed out of one metazoan body into a body of the next generation in egg cytoplasm. An optimal female phenotype from the point of view of the female’s own nuclear genes is likely to be very much the same as an optimal female phenotype from the point of view of her mitochondrial DNA. Both have an interest in her successfully surviving, reproducing, and rearing offspring. At least, that is true as far as female offspring are concerned. Mitochondria presumably have no ‘wish’ for their bodies to have sons: a male body represents the end of the line as far as mitochondrial descent is concerned. All mitochondria in existence have spent the vast majority of their ancestral careers in female bodies, and they might tend to have what it takes to persist in inhabiting female bodies. In birds the interests of mitochondrial DNA will be closely similar to those of Y-chromosomal DNA, and slightly divergent from those of autosomal and X-chromosomal DNA. And if mitochondrial DNA could exert phenotypic power in the egg of a mammal, it is perhaps not too fanciful to imagine it frantically fighting off the kiss of death of Y-bearing sperms (Eberhard 1980; Cosmides & Tooby 1981). But in any case, if the interests of mitochondrial DNA and nuclear DNA are not always identical, they are very close, certainly much closer than the interests of fluke DNA and snail DNA.

  The message of the present section is this. The fact that snail genes conflict with fluke genes more than they conflict with other snail genes at different loci is not the obvious foregone conclusion it might appear to be. It results simply from the fact that any two genes in a snail nucleus are obliged to use the same exit route from the present body into the future. Both have the same stake in the success of the present snail in manufacturing gametes, getting them fertilized, and securing the survival and reproduction of the offspring so formed. Fluke genes conflict with snail genes in their influence on the shared phenotype, simply because their destiny is shared for only a short part of the future: their common cause is limited to the life of the present host body, and does not carry over into the gametes and offspring of the present host.

  The role of mitochondria in the argument is to exemplify cases where parasite and host genes share the same gametic destiny, at least in part. If nuclear genes do not conflict with nuclear genes at other loci, it is only because meiosis is even-handed: meiosis does not normally favour some loci over others, nor some alleles over others, but scrupulously puts one gene at random from each diploid pair in every gamete. Of course there are instructive exceptions, and for my thesis they are sufficiently important to have dominated the two chapters on ‘outlaws’ and ‘selfish DNA’. There, as here, an important message is that replicating entities will tend to work against each other to the extent that they employ different methods of egress from vehicle to vehicle.

  Returning to the main subject of the present chapter, parasitic and symbiotic relationships can be classified in various ways for different purposes. The classifications developed by parasitologists and doctors are no doubt useful for their purposes, but I want to develop a particular classification based on the concept of gene power. It should be remembered that, from this point of view, the normal relationship between different genes in the same nucleus, even on the same chromosome, is just one extreme on the continuum of parasitic or symbiotic relationships.

  The first dimension of my classification has already been stressed. It concerns the degree of similarity or difference in the methods of egress from hosts, and propagation of host genes and parasite genes. At one extreme will be parasites that use the host’s propagules for their own reproduction. For such parasites an optimum host phenotype from the parasite’s point of view is likely to coincide with the optimum from the point of view of the host’s own genes. This is not to say that the host genes would not ‘prefer’ to be rid of the parasite altogether. But both have an interest in mass-producing the same propagules, and both have an interest in developing a phenotype that is good for mass-producing those same propagules: the right beak length, wing shape, courtship behaviour, clutch size, etc., down to minute details of all aspects of the phenotype.

  At the other extreme will be parasites whose genes are passed on not in the
host’s reproductive propagules but, say, in the host’s exhaled breath, or in the host’s dead body. In these cases the optimum host phenotype from the parasite genes’ point of view is likely to be very different from the optimum host phenotype from the host genes’ point of view. The phenotype which emerges will be a compromise. This, then, is one dimension of classification of host–parasite relationships. I shall call it the dimension of ‘propagule overlap’.

  A second dimension of classification concerns the time of action of parasite genes during host development. A gene, whether a host gene or a parasite gene, can exert a more fundamental influence on the final host phenotype if it acts early in the development of the host embryo than if it acts late. A radical change such as the development of two heads could be achieved by a single mutation (in host or parasite genome) provided the mutation acted sufficiently early in the embryonic development of the host. A late-acting mutant (again, in host or parasite genome)—a mutant that does not begin to act until the host body has reached adulthood—is likely to have only a small effect, since the general architecture of the body will, by then, have been laid down. Therefore a parasite that enters its host after the host has reached adulthood is less likely to have a radical effect on the host’s phenotype than a parasite that gets in early. There are notable exceptions, however, such as the parasitic castration of crustacea already mentioned.

  My third dimension of classification of host–parasite relations concerns the continuum from what may be called close intimacy to action at a distance. All genes exert power primarily by serving as templates for the synthesis of proteins. The locus of primary gene power is, therefore, the cell, in particular the cytoplasm surrounding the nucleus in which the gene sits. Messenger RNA streams through the nuclear membrane and mediates genetic control over cytoplasmic biochemistry. The phenotypic expression of a gene is then, in the first place, its influence on cytoplasmic biochemistry. In its turn, this influences the form and structure of the whole cell, and the nature of its chemical and physical interactions with neighbouring cells. This affects the build-up of multicellular tissues, and in turn the differentiation of a variety of tissues in the developing body. Finally emerge the attributes of the whole organism that gross anatomists and ethologists identify at their level as phenotypic expressions of genes.