The Extended Phenotype

by Richard Dawkins

  Steele’s interpretation begins with the Burnet theory. Somatic mutation generates genetic diversity in the population of immune cells. Clonal selection favours those genetic varieties of cell that satisfactorily destroy the antigen, and they become very numerous. There is more than one solution to any antigenic problem, and the end result of the selection process is different in every rabbit. Now Temin’s proviruses step in. They transcribe a random sample of the genes in the immune cells. Because cells carrying successful antibody genes outnumber the others, these successful genes are statistically most likely to be transcribed. The proviruses cart these genes off to the germ cells, burrow into the germ-line chromosomes and leave them there, presumably snipping out the incumbent occupants of the locus as they do so. The next generation of rabbits is thereby able to benefit directly from the immunological experience of its parents, without having to experience the relevant antigens themselves, and without the painfully slow and wasteful intervention of selective organism death.

  The really impressive evidence only became available after Steele’s theory was cut and dried and published, a striking and rather surprising instance of science proceeding in the way philosophers think it proceeds. Gorczynski and Steele (1980) investigated the inheritance, via the father, of immune tolerance in mice. Using an extremely high-dosage version of the classic Medawar method, they exposed baby mice to cells from another strain, thereby rendering them tolerant as adults to subsequent grafts from the same donor strain. They then bred from these tolerant males, and concluded that their tolerance was inherited by about half their children, who were not exposed as infants to the foreign antigens. Furthermore, the effect seemed to carry over to the grandchild generation.

  Subject to confirmation, we have here a prima-facie case for the inheritance of acquired characteristics. Gorczynski and Steele’s brief discussion of their experiment, and of extended experiments reported more recently (Gorczynski & Steele 1981), resembles Steele’s interpretation of the rabbit work, paraphrased above. The main differences between the two cases are firstly that the rabbits could have inherited something in maternal cytoplasm while the mice could not; and secondly that the rabbits were alleged to have inherited an acquired immunity, while the mice are supposed to have inherited an acquired tolerance. These differences are probably important (Ridley 1980b; Brent et al. 1981), but I shall not make much of them, since I am not attempting to evaluate the experimental results themselves. I shall concentrate on the question of whether, in any case, Steele is really offering ‘a Lamarckian challenge to Darwinism’.

  There are some historical points to get out of the way first. The inheritance of acquired characteristics is not the aspect of his theory that Lamarck himself emphasized and, contra Steele (1979, p. 6), it is not true that the idea originated with him: he simply took over the conventional wisdom of his time and grafted to it other principles like ‘striving’ and ‘use and disuse’. Steele’s viruses seem more reminiscent of Darwin’s own pangenetic ‘gemmules’ than of anything postulated by Lamarck. But I mention history just to get it out of the way. We give the name Darwinism to the theory that undirected variation in an insulated germ-line is acted upon by selection of its phenotypic consequences. We give the name Lamarckism to the theory that the germ-line is not insulated, and that environmentally imprinted improvements may directly mould it. In this sense, is Steele’s theory Lamarckian and anti-Darwinian?

  By inheriting their parents’ acquired idiotypes, rabbits would undoubtedly benefit. They would begin life with a head start in the immunological battle against plagues that their parents met, and that they themselves are likely to meet. This is a directed, adaptive change, then. But is it really imprinted by the environment? If antibody formation worked according to some kind of ‘instructive’ theory, the answer would be yes. The environment, in the shape of antigenic protein molecules, would then directly mould antibody molecules in parent rabbits. If the offspring of those rabbits turned out to inherit a predilection to make the same antibodies, we would have full-blooded Lamarckism. But the conformations of the antibody proteins would, on this theory, somehow have to be reverse-translated into nucleotide code. Steele (p. 36) is adamant that there is no suggestion of such reverse translation, only reverse transcription from RNA to DNA. He is not proposing any violation of Crick’s central dogma, although of course others are at liberty to do so (I shall return to this point in a more general context later).

  The very essence of Steele’s hypothesis is that the adaptive improvement comes about through selection of initially random variation. It is about as Darwinian a theory as it is possible to be, provided we think of the replicator and not the organism as the unit of selection. Nor is it just vaguely analogous to Darwinism, in the way of, say, the ‘meme’ theory, or Pringle’s (1951) theory that learning results from selection among a pool of oscillation frequencies in a population of neuronal coupled oscillators. Steele’s replicators are DNA molecules in cell nuclei. They are not just analogous to the replicators of Darwinism. They are the very same replicators. The scheme for natural selection which I outlined in Chapter 5 needs no modification to plug straight into the Steele theory. Steele’s kind of Lamarckism only seems like the imprinting of environmental features on the germ-line if we think at the level of the individual organism. It is true that he is claiming that characteristics acquired by the organism are inherited. But if we look at the lower level of genetic replicators, it is clear that the adaptation comes about through selection not ‘instruction’ (see below). It just happens to be selection within the organism. Steele (1979, p. 43) would not disagree: ‘… it depends very much on essential Darwinian principles of natural selection’.

  Despite Steele’s avowed indebtedness to Arthur Koestler, there is naught here for the comfort of those, usually non-biologists, whose antipathy to Darwinism is fundamentally provoked by the bogy of ‘blind chance’. Or, for that matter, the twin bogy of a ruthlessly indifferent grim reaper, mocking us as the sole First Cause of our exalted persons, modifying ‘all things by blindly starving and murdering everything that is not lucky enough to survive in the universal struggle for hogwash’ (Shaw 1921). If Steele proves right, we should hear no triumphant chuckles from the shade of Bernard Shaw! Shaw’s live spirit rebelled passionately against the Darwinian ‘chapter of accidents’. ‘… it seems simple, because you do not at first realize all that it involves. But when its whole significance dawns on you, your heart sinks into a heap of sand within you. There is a hideous fatalism about it, a ghastly and damnable reduction of beauty and intelligence, of strength and purpose, of honor and aspiration …’ If we must place emotion before truth, I have always found natural selection to have an inspiring, if grim and austere, poetry of its own—a ‘grandeur in this view of life’ (Darwin 1859). All I am saying here is that if you are squeamish about ‘blind chance’, do not look to Steele’s theory for an escape. But perhaps it is not too much to hope that a proper understanding of Steele’s theory may help to show that ‘blind chance’ is not the adequate epitome of Darwinism that Shaw, Cannon (1959), Koestler (1967) and others think it is.

  Steele’s theory, then, is a version of Darwinism. The cells which are selected, according to the Burnet theory, are vehicles for active replicators, namely the somatically mutated genes within them. They are active, but are they germ-line replicators? The essence of what I am saying is that the answer is an emphatic yes, if Steele’s addition to the Burnet theory is true. They do not belong to what we have conventionally thought of as the germ-line, but it is a logical implication of the theory that we have simply been mistaken as to what the germ-line truly is. Any gene in a ‘somatic’ cell which is a candidate for proviral conveyance into a germ cell is, by definition, a germ-line replicator. Steele’s book might be retitled The Extended Germ-line! Far from being uncomfortable for neo-Weismannists, it turns out to be deeply congenial to us.

  Perhaps, then, it is not really all that ironic that, apparently unknown to Steele, something
bearing more than a passing resemblance to his theory was adopted by, of all people, Weismann himself, in 1894. The following account is taken from Ridley (1982; Maynard Smith, 1980, also noted the precedent). Weismann developed an idea from Roux, which he called ‘intra-selection’. I quote from Ridley: ‘Roux had argued that there is a struggle for food between the parts of an organism just like the struggle for existence between organisms … Roux’s theory was that the struggle of the parts, together with the inheritance of acquired characters, was sufficient to explain adaptation.’ Substitute ‘clones’ for ‘parts’, and you have Steele’s theory. But, as might be expected, Weismann did not go all the way with Roux in postulating the literal inheritance of acquired characteristics. Instead, in his theory of ‘germinal selection’, he invoked the pseudo-Lamarckian principle which later became known as the ‘Baldwin Effect’ (Weismann was not the only one to discover the idea before Baldwin). Weismann’s use of the theory of intra-selection in explaining coadaptation will be dealt with below, for it closely parallels one of Steele’s own preoccupations.

  Steele does not venture far from his own field of immunology, but he would like a version of his theory to apply elsewhere, and particularly to the nervous system and the adaptive improvement mechanism known as learning. ‘If [the hypothesis] is to have any general applicability to the evolutionary adaptation process, it must account for the adaptive potential of the neuronal networks of the brain and central nervous system’ (Steele 1979, p. 49, his rather surprising emphasis). He seems a little vague as to exactly what might be selected within the brain, and, in case he can do something with it, I offer him a free gift of my own theory of ‘selective neurone death as a possible memory mechanism’ (Dawkins 1971).

  But is the theory of clonal selection really likely to apply outside the domain of the immune system? Is it limited by the very special circumstances of the immune system, or might it be linked up with the old Lamarckian principle of use and disuse? Could clonal selection be embraced by the blacksmith’s arms? Could the adaptive changes brought about by muscular exercise be inherited? I doubt it very much: the conditions are not right for natural selection to work within the blacksmith’s arms in favour of, say, cells that flourish in an aerobic environment over those preferring anaerobic biochemistry, the successful genes being reverse-transcribed into just the right chromosomal locus in the germ-line. But even if this kind of thing were conceivable for some example outside the immune system, there is a major theoretical difficulty.

  The problem is this. The qualities that make for success in clonal selection would necessarily be those that give cells an advantage over rival cells in the same body. These qualities need have no connection with what is good for the body as a whole, and our discussion of outlaws suggests that they might well actively conflict with what is good for the body as a whole. Indeed, to me a slightly unsatisfactory aspect of the Burnet theory itself is that the selection process at its heart is contrived ad hoc. It is assumed that those cells whose antibodies neutralize invading antigens will propagate at the expense of other cells. But this propagation is not due to any intrinsic cellular advantage: on the contrary, cells that did not risk their lives smothering antigens but selfishly left the task to their colleagues should, on the face of it, have a built-in advantage. The theory has to introduce an arbitrary and unparsimonious selection rule, imposed, as it were, from above, so that those cells which benefit the body as a whole become more numerous. It is as though a human dog-breeder deliberately selected for altruistic heroism in the face of danger. He could probably achieve it, but natural selection could not. Unadulterated clonal selection should favour selfish cells whose behaviour conflicts with the best interests of the body as a whole.

  In the terms of Chapter 6, what I am saying is that vehicle selection at the cellular level is likely, under the Burnet theory, to come into conflict with vehicle selection at the organism level. This, of course, does not worry me, since I carry no brief for the organism as pre-eminent vehicle; I simply add one more entry to the list of ‘outlaws I have known’; one more ingenious byway of replicator propagation, along with jumping genes and selfish DNA. But it should worry anybody, such as Steele, who sees clonal selection as a supplementary means by which bodily adaptations come about.

  The problem goes deeper than that. It is not just that clonally selected genes would tend to be outlaws as far as the rest of the body is concerned. Steele looks to clonal selection to speed up evolution. Conventional Darwinism proceeds by differential individual success, and its speed, other things being equal, will be limited by individual generation time. Clonal selection would be limited by cellular generation time, which is perhaps two orders of magnitude shorter. This is why it might be thought to speed up evolution but, to anticipate the argument of my final chapter, it raises a deep difficulty. The success of a complex, multicellular organ, like an eye, cannot be judged in advance of the eye’s starting to work. Cellular selection could not improve the design of an eye, because the selective events all take place in the prefunctional eye of an embryo. The embryo’s eye is closed, and never sees an image until after cellular selection, if it existed, would be complete. The general point is that cellular selection cannot achieve the speeding up of evolution attributed to it, if the adaptation of interest has to develop on the slow scale of multicellular cooperation.

  Steele has a point to make about coadaptation. As Ridley (1982) comprehensively documents, multidimensional coadaptation was one of the bugbears of the early Darwinians. For instance, to take the eye again, J. J. Murphy said ‘It is probably no exaggeration that, in order to improve such an organ as the eye at all, it must be improved in ten different ways at once’ (1866, quoted by Ridley). It may be remembered that, speaking of the evolution of whales, I used a similar premise for a different purpose in Chapter 6. Fundamentalist orators still find the eye one of their most telling standbys. Incidentally, The Sunday Times (13 July 1980) and The Guardian (21 November 1978) both raise the debating point of the eye as though it were a new one, the latter paper reassuring us that an eminent philosopher (!) was rumoured to be giving the problem his best attention. Steele seems to have been initially attracted by Lamarckism because of his unease about coadaptation, and he believes that his clonal selection theory could in principle alleviate the difficulty, if difficulty there be.

  Let us make use of that other chestnut of the schoolroom, the giraffe’s neck, discussing it in conventional Darwinian terms first. A mutation to elongate the ancestral neck might work on, say, the vertebrae, but it seems to a naive observer too much to hope that the same mutation will simultaneously elongate the arteries, veins, nerves, etc. Actually, whether it is too much to hope depends on details of embryology which we should learn to be more aware of: a mutation that acts sufficiently early in development could easily have all those parallel effects simultaneously. However, let us play along with the argument. The next step is to say that it is difficult to imagine a mutant giraffe with elongated vertebrae being able to exploit its treetop browsing advantage, because its nerves, blood vessels, etc. are too short for its neck. Conventional Darwinian selection, naively understood, has to wait for an individual that is fortunate enough to combine all the necessary coadapted mutations simultaneously. This is where clonal selection could come to the rescue. One major mutation, say the elongation of the vertebrae, sets up conditions in the neck which select for clones of cells that can flourish in that environment. Maybe the elongated vertebrae provide an overstretched, highly strung, tense environment in the neck, where only elongated cells flourish. If there is genetic variation among the cells, genes ‘for’ elongated cells survive and are passed on to the giraffe’s children. I have put the point facetiously, but a more sophisticated version of it could, I suppose, follow from the clonal selection theory.

  I said I would return to Weismann at this point, for he too saw the utility of within-body selection as a solution to the coadaptation problem. Weismann thought that ‘intra-selection’�
�the selective struggle among parts within the body—‘would ensure that all the parts inside the organism were of the best mutual proportions’ (Ridley 1982). ‘If I am not mistaken, the phenomenon which Darwin called correlation, and justly regarded as an important factor in evolution, is for the most part an effect of intra-selection’ (Weismann, quoted in Ridley). As already stated, Weismann, unlike Roux, did not go on to invoke the direct inheritance of intra-selected varieties. Rather, ‘… in each separate individual the necessary adaptation will be temporarily accomplished by intra-selection … Time would thus be gained till, in the course of generations, by constant selection of those germs the primary constituents of which are best suited to one another, the greatest possible degree of harmony may be reached.’ I think I find Weismann’s ‘Baldwin Effect’ version of the theory more plausible than Steele’s Lamarckian version, and just as satisfying an explanation of coadaptation.

  I used the word ‘scare’ in the heading of this section, and went so far as to say that a true revival of Lamarckism would devastate my world-view. Yet the reader may now feel that the statement was a hollow one, like that of a man who dramatically threatens to eat his hat, while knowing full well that his hat is made of pleasantly flavoured ricepaper. A Lamarckian zealot may complain that the last resort of a Darwinian, having failed to discredit awkward experimental results, is to claim them for his own; to make his own theory so resilient that there is no experimental result that could falsify it. I am sensitive to this criticism, and must reply to it. I must show that the hat that I threatened to eat really is tough and distasteful. So, if Steele’s kind of Lamarckism is really Darwinism in disguise, what kind of Lamarckism would not be?