The Extended Phenotype

by Richard Dawkins

  There is no suggestion here that the two subgroups of bees were genetically different, though they may have been. What matters for the point I am making is that each individual follows local behavioural rules, the combined effect of which normally gives rise to coordinated swarm behaviour. These evidently include rules for resolving ‘disputes’ in favour of the majority. Disagreements over the preferred location for the outer wall of a termite mound might be just as serious for colony survival as disagreement over nesting sites among Lindauer’s bees (colony survival matters, because of its effects on the survival of the genes causing individuals to resolve disputes). As a working hypothesis we might expect that disputes resulting from genetic heterogeneity in termites would be resolved by similar rules. In this way the extended phenotype could take up a discrete and regular shape, despite being built by genetically heterogeneous workers.

  The analysis of artefacts given in this chapter seems, at first sight, vulnerable to reductio ad absurdum. Isn’t there a sense, it may be asked, in which every effect that an animal has upon the world is an extended phenotype? What about the footprints left in the mud by an oystercatcher, the paths worn through the grass by sheep, the luxuriant tussock that marks the site of a last year’s cowpat? A pigeon’s nest is an artefact without a doubt, but in gathering the sticks the bird also changes the appearance of the ground where they had lain. If the nest is called extended phenotype, why shouldn’t we so call the bare patch of ground where the sticks used to lie?

  To answer this we must recall the fundamental reason why we are interested in the phenotypic expression of genes in the first place. Of all the many possible reasons, the one which concerns us in this book is as follows. We are fundamentally interested in natural selection, therefore in the differential survival of replicating entities such as genes. Genes are favoured or disfavoured relative to their alleles as a consequence of their phenotypic effects upon the world. Some of these phenotypic effects may be incidental consequences of others, and have no bearing on the survival chances, one way or the other, of the genes concerned. A genetic mutation that changes the shape of an oystercatcher’s foot will doubtless thereby influence the oystercatcher’s success in propagating it. It may, for instance, slightly reduce the bird’s risk of sinking into mud, while at the same time it slightly slows him down when he is running on firm ground. Such effects are likely to be of direct relevance to natural selection. But the mutation will also have an effect on the shape of the footprints left behind in the mud—arguably an extended phenotypic effect. If, as is perfectly likely, this has no influence on the success of the gene concerned (Williams 1966, pp. 12–13), it is of no interest to the student of natural selection, and there is no point in bothering to discuss it under the heading of the extended phenotype, though it would be formally correct to do so. If, on the other hand, the changed footprint did influence survival of the oystercatcher, say by making it harder for predators to track the bird, I would want to regard it as part of the extended phenotype of the gene. Phenotypic effects of genes, whether at the level of intracellular biochemistry, gross bodily morphology or extended phenotype, are potentially devices by which genes lever themselves into the next generation, or barriers to their doing so. Incidental side-effects are not always effective as tools or barriers, and we do not bother to regard them as phenotypic expressions of genes, either at the conventional or the extended phenotype level.

  It is unfortunate that this chapter has had to be rather hypothetical. There have been only a few studies of the genetics of building behaviour in any animal (e.g. Dilger 1962), but there is no reason to think that ‘artefact genetics’ will be any different, in principle, from behaviour genetics generally (Hansell 1984). The idea of the extended phenotype is still sufficiently unfamiliar that it might not immediately occur to a geneticist to study termite mounds as a phenotype, even if it were practically easy to do so—and it wouldn’t be easy. Yet we must acknowledge at least the theoretical validity of such a branch of genetics if we are to countenance the Darwinian evolution of beaver dams and termite mounds. And who can doubt that, if termite mounds fossilized plentifully, we would see graded evolutionary series with trends as smooth (or as punctuated!) as any that we find in vertebrate skeletal palaeontology (Schmidt 1955; Hansell 1984)?

  Permit me one further speculation to lead us into the next chapter. I have spoken as if the genes inside a termite mound were all enclosed in the nuclei of cells of termite bodies. The ‘embryological’ forces bearing on the extended phenotype have been assumed to originate from the genes of individual termites. Yet the chapter on arms races and manipulation should have alerted us to another way of looking at it. If all the DNA could be distilled out of a termite mound, perhaps as much as one-quarter of it would not have originated from termite nuclei at all. Some such proportion of the body weight of each individual termite is typically made up of symbiotic cellulose-digesting microorganisms in the gut—flagellates or bacteria. The symbionts are obligately dependent on the termites, and the termites on them. The proximal phenotypic power of the symbiont genes is exerted via protein synthesis in symbiont cytoplasm. But just as termite genes reach out beyond the cells that enclose them and manipulate the development of whole termite bodies and hence of the mound, is it not almost inevitable that the symbiont genes will have been selected to exert phenotypic power on their surroundings? And will this not include exerting phenotypic power on termite cells and hence bodies, on termite behaviour and even termite mounds? Along these lines, could the evolution of eusociality in the Isoptera be explained as an adaptation of the microscopic symbionts rather than of the termites themselves?

  This chapter has explored the idea of the extended phenotype, first of genes in a single individual, then of genes from different but closely related individuals, members of a kin-group. The logic of the argument now seems to compel us to contemplate the possibility of an extended phenotype’s being jointly manipulated, not necessarily cooperatively, by genes from distantly related individuals, individuals of different species, even different kingdoms. This is the direction in which our next outward step must take us.

  12 Host Phenotypes of Parasite Genes

  Let us briefly take stock of where we have reached in our outward march. The phenotypic expression of a gene can extend outside the cell in which the genes exert their immediate biochemical influence, to affect gross features of a whole multicellular body. This is commonplace, and we are conventionally used to the idea of a gene’s phenotypic expression being extended this far.

  In the previous chapter we took the small further step of extending the phenotype to artefacts, built by individual behaviour which is subject to genetic variation, for instance caddis houses. Next we saw that an extended phenotype can be built under the joint influence of genes in more than one individual body. Beaver dams and termite mounds are collectively built by the behavioural efforts of more than one individual. A genetic mutation in one individual beaver could show itself in phenotypic change in the shared artefact. If the phenotypic change in the artefact had an influence on the success of replication of the new gene, natural selection would act, positively or negatively, to change the probability of similar artefacts existing in the future. The gene’s extended phenotypic effect, say an increase in the height of the dam, affects its chances of survival in precisely the same sense as in the case of a gene with a normal phenotypic effect, such as an increase in the length of the tail. The fact that the dam is the shared product of the building behaviour of several beavers does not alter the principle: genes that tend to make beavers build high dams will themselves, on average, tend to reap the benefits (or costs) of high dams, even though every dam may be jointly built by several beavers. If two beavers working on the same dam have different genes for dam height, the resulting extended phenotype will reflect the interaction between the genes, in the same way as bodies reflect gene interactions. There could be extended genetic analogues of epistasis, of modifier genes, even of dominance and recess
iveness.

  Finally, at the end of the chapter, we saw that genes ‘sharing’ a given extended phenotypic trait might come from different species, even different phyla and different kingdoms. This chapter will develop two further ideas. One is that phenotypes that extend outside the body do not have to be inanimate artefacts: they can themselves be built of living tissue. The other idea is that wherever there are ‘shared’ genetic influences on an extended phenotype, the shared influences may be in conflict with each other rather than cooperative. The relationships we shall be concerned with are those of parasites and their hosts. I shall show that it is logically sensible to regard parasite genes as having phenotypic expression in host bodies and behaviour.

  The caddis larva rides inside the stone house that it built. It therefore seems appropriate to regard the house as the outer wall of the gene vehicle, the casing of the survival machine. It is even easier to regard the shell of a snail as part of the phenotypic expression of snail genes since, although the shell is inorganic and ‘dead’, its chemical substance was directly secreted by snail cells. Variations in, say, shell thickness would be called genetic if genes in snail cells affected shell thickness. Otherwise they would be called ‘environmental’. But there are reports of snails with trematode parasites having thicker shells than unparasitized snails (Cheng 1973). From the point of view of snail genetics, this aspect of shell variation is under ‘environmental’ control—the fluke is part of the environment of the snail—but from the point of view of fluke genetics it might well be under genetic control: it might, indeed, be an evolved adaptation of the fluke. It is admittedly also possible that the thickened shell is a pathological response of the snail, a dull byproduct of infection. But let me explore the possibility that it is a fluke adaptation, because it is an interesting idea to use in further discussion.

  If we consider snail shell variation as, in part, phenotypic expression of snail genes, we might recognize an optimum shell thickness in the following sense. Selection presumably penalizes snail genes that make shells too thick, as well as those that make shells too thin. Thin shells provide inadequate protection. Genes for too-thin shells therefore endanger their germ-line copies, which are thus not favoured by natural selection. Shells that are too thick presumably protect their snails (and the enclosed germ-line genes for extra thickness) superlatively, but the extra cost of making a thick shell detracts from the snail’s success in some other way. In the economy of the body, resources that are consumed in making extra-thick shells, and in carrying the extra weight around, might better have been diverted into making, say, larger gonads. Continuing with the hypothetical example, therefore, genes for extra-thick shells will tend to induce in their bodies some compensating disadvantage such as relatively small gonads, and they will therefore not be passed on to the next generation so effectively. Even if there is, in fact, no trade-off between shell thickness and gonad size, there is bound to be some kind of analogous trade-off, and a compromise will be reached at an intermediate thickness. Genes that tend to make snail shells either too thick or too thin will not prosper in the snail gene pool.

  But this whole argument presupposes that the only genes that have power over variation in shell thickness are snail genes. What if some of the causal factors that are, by definition, environmental from the snail’s point of view, turn out to be genetic from some other point of view, say that of the fluke? Suppose we adopt the suggestion made above that some fluke genes are capable, through an influence on snail physiology, of exerting an effect on snail shell thickness. If shell thickness influences the replication success of such fluke genes, natural selection is bound to work on their frequencies relative to their alleles in the fluke gene-pool. Changes in snail shell thickness may be regarded, then, at least in part, as potential adaptations for the benefit of fluke genes.

  Now, the optimum shell thickness from the point of view of the fluke genes is hardly likely to be the same as the optimum from the point of view of snail genes. For instance, snail genes will be selected for their beneficial effects on snail reproduction as well as snail survival, but (except under special circumstances which we shall come on to) fluke genes may value snail survival but they will not value snail reproduction at all. In the inevitable trade-off between the demands of snail survival and snail reproduction, therefore, snail genes will be selected to produce an optimal compromise, while fluke genes will be selected to devalue snail reproduction to the advantage of snail survival, and hence to thicken the shell. A thickening of the shell in parasitized snails is, it will be remembered, the observed phenomenon with which we began.

  It may be objected here that, although a fluke has no direct stake in the reproduction of its own snail host, it does have a stake in there being a new generation of snails at large. This is true, but we must be very careful before we use the fact to predict that selection would favour fluke adaptations to enhance snail reproduction. The question we have to ask is this. Given that the fluke gene-pool was dominated by genes that aided snail reproduction at the expense of snail survival, would selection favour a selfish fluke gene that sacrificed the reproduction of its particular snail host, even parasitically castrated the snail, in the interest of prolonging the life of that host, and hence of promoting its own survival and reproduction? Except under special circumstances the answer is surely yes; such a rare gene would invade the fluke gene-pool, since it could exploit the free supply of new snails encouraged by the public-spirited majority of the fluke population. In other words, favouring snail reproduction at the expense of snail survival would not be a fluke ESS. Fluke genes that manage to shift the snail’s investment of resources away from reproduction and into survival will tend to be favoured in the fluke gene-pool. It is entirely plausible, therefore, that the extra thickness of shells observed in parasitized snails is a fluke adaptation.

  On this hypothesis, the shell phenotype is a shared phenotype, influenced by fluke genes as well as by snail genes, just as the beaver dam is a phenotype shared by genes in more than one individual beaver. According to the hypothesis, there are two optimum thicknesses of snail shells: a relatively thick fluke optimum, and a somewhat thinner snail optimum. The observed thickness in parasitized snails will probably be somewhere between the two optima, since snail genes and fluke genes are both in a position to exert power, and they are exerting their power in opposite directions.

  As for parasite-free snails, it might be expected that their shells would have the snail-optimal thickness, since there are no fluke genes to exert power. This is too simple, however. If the population at large has a high incidence of fluke infestation, the gene-pool will probably contain genes that tend to compensate for the thickening effect of fluke genes. This would lead to uninfected snails having over-compensating phenotypes, shells that are thinner even than the snail optimum. I therefore predict that shell thicknesses in fluke-free areas should be intermediate between those of infected snails and uninfected snails in fluke-infested areas. I do not know of any evidence bearing on this prediction, but it would be interesting to look. Note that this prediction does not depend on any ad hoc assumption about snails ‘winning’ or flukes ‘winning’. It assumes that both snail genes and fluke genes exert some power over the snail phenotype. The prediction will follow regardless of the quantitative details of that power.

  Flukes live inside snail shells in a sense that is not too far removed from the sense in which snails live inside snail shells, and caddis larvae live inside their stone houses. Having accepted the idea that the form and colour of a caddis house might constitute phenotypic expression of caddis genes, it is not difficult to accept the idea of the form and colour of a snail shell being phenotypic expression of genes in a fluke inside the snail. If we could fancifully imagine a fluke gene and a snail gene intelligently discussing with a caddis gene the problems of making a hard outer wall for protection, I doubt if the conversation would make any reference to the fact that the fluke was a parasite while the caddis and the snail were not. Th
e rival merits of secreting calcium carbonate, recommended by the fluke and the snail genes, versus picking up stones, preferred by the caddis gene, would be discussed. There might be some reference to the fact that a convenient and economical way of secreting calcium carbonate involves the use of a snail. But, from the gene’s-eye viewpoint, I suspect that the concept of parasitism would be treated as irrelevant. All three genes might regard themselves as parasitic, or alternatively as using comparable levers of power to influence their respective worlds so as to survive. The living cells of the snail would be regarded by the snail gene and the fluke gene as useful objects to be manipulated in the outside world, in exactly the same sense as the stones on the bottom of the stream would be regarded by the caddis gene.

  By discussing inorganic snail shells, I have retained continuity with the caddis houses and other non-living artefacts of the previous chapter, thereby pursuing my policy of sustaining credulity by extending the concept of the phenotype gradually by insensible degrees. But now it is time to grasp the living snail firmly by the horns. Flukes of the genus Leucochloridium invade the horns of snails where they can be seen through the skin, conspicuously pulsating. This tends to make birds, who are the next host in the life cycle of the fluke, bite off the tentacles mistaking them, Wickler (1968) suggests, for insects. What is interesting here is that the flukes seem also to manipulate the behaviour of the snails. Whether it is because the snail’s eyes are at the ends of the horns, or whether through some more indirect physiological route, the fluke manages to change the snail’s behaviour with respect to light. The normal negative phototaxis is replaced in infected snails by positive light-seeking. This carries them up to open sites where they are presumably more likely to be eaten by birds, and this benefits the fluke.