Matt Ridley

produce hormones which raise the mother's blood pressure and blood sugar.

The mother responds by raising her insulin levels to combat this invasion. Yet, if for some reason the foetal hormone is missing, the mother does not need to raise her insulin levels and a normal pregnancy ensues. In other words, although mother and foetus have a common purpose, they argue fiercely about the details of how much of the mother's resources the foetus may have — exactly as they later will during weaning.

But the foetus is built parly with maternal genes, so it would not be surprising if these genes found themselves with, as it were, a conflict of interest. The father's genes in the foetus have no such worries. They do not have the mother's interest at heart, except insofar as she provides a home for them. To turn briefly anthropomorphic, the father's genes do not trust the mother's genes to make a sufficiently invasive placenta; so they do the job themselves. Hence the paternal imprinting of placental genes as discovered by the two-fathered embryos.

Haig's hypothesis made some predictions, many of which were soon borne out. In particular, it predicted that imprinting would not occur in animals that lay eggs, because a cell inside an egg has no means of influencing the investment made by the mother in yolk size: it is outside the body before it can manipulate her. Likewise, even marsupials such as kangaroos, with pouches in place of placentas, would not, on Haig's hypothesis, have imprinted genes. So far, it appears, Haig is right. Imprinting is a feature of placental mammals 2 1 0 G E N O M E

and of plants whose seeds gain sustenance from the parent plant.4

Moreover, Haig was soon triumphantly noting that a newly discovered pair of imprinted genes in mice had turned up exactly where he expected them: in the control of embryonic growth. I G F 2 is a miniature protein, made by a single gene, that resembles insulin. It is common in the developing foetus and switched off in the adult.

I G F 2 R is a protein to which I G F 2 attaches itself for a purpose that remains unclear. It is possible that I G F 2 R is there simply to get rid of I G F 2 . Lo and behold, both the IGF2 and the IGF2R

genes are imprinted: the first being expressed only from the paternal chromosome, the second from the maternal one. It looks very much like a little contest between the paternal genes trying to encourage the growth of the embryo and the maternal ones trying to moderate it.5

Haig's theory predicts that imprinted genes will generally be found in such antagonistic pairs. In some cases, even in human beings, this does seem to be the case. The human IGF2 gene on chromosome 11

is paternally imprinted and when, by accident, somebody inherits two paternal copies, they suffer from Beckwith-Wiedemann syndrome, in which the heart and liver grow too large, and tumours of embryonic tissues are common. Although in human beings IGF2R

is not imprinted, there does seem to be a maternally imprinted gene, H19, that opposes IGF2.

If imprinted genes exist only to combat each other, then you should be able to switch both off and it will have no effect at all on the development of the embryo. You can. Elimination of all imprinting leads to normal mice. We are back in the familiar territory of chromosome 8, where genes are selfish and do things for the benefit of themselves, not for the good of the whole organism. There is almost certainly nothing intrinsically purposeful about imprinting (though many scientists have speculated otherwise); it is another illustration of the theory of the selfish gene and of sexual antagonism in particular.

Once you start thinking in selfish-gene terms, some truly devious ideas pop into your head. Try this one. Embryos under the influence of paternal genes might behave differently if they share the womb S E X 2 1 1

with full siblings or if they share the womb with embryos that have different fathers. They might have more selfish paternal genes in the latter case. Having thought the thought, it was comparatively easy to do the deed and test this prediction with a natural experiment.

Not all mice are equal. In some species of mice, for example Peromyscus maniculatus, the females are promiscuous, and each litter generally contains babies fathered by several different males. In other species, for example Peromyscus polionatus, the females are strictly monogamous and each litter contains full siblings who share both father and mother.

So what happens when you cross a P. maniculatus mouse with a P. polionatus mouse? It depends on which species is the father and which is the mother. If the promiscuous P. maniculatus is the father, the babies are born giant-sized. If the monogamous P. polionatus is the father, the babies are born tiny. Do you see what is happening?

Paternal maniculatus genes, expecting to find themselves in a womb with competitors that are not even related, have been selected to fight for their share of the mother's resources at the expense of their co-foetuses. Maternal maniculatus genes, expecting to find embryos in their wombs that fight hard for her resources, have been selected to fight back. In the more neutral environment of polionatus wombs, the aggressive maniculatus genes from the father encounter only token opposition, so they win their particular battle: the baby is big if fathered by the promiscuous father and small if mothered by the promiscuous mother. It is a very neat demonstration of the imprinting theory.6

Neat as this tale is, it cannot be told without a caveat. Like many of the most appealing theories it may be too good to be true. In particular, it makes a prediction that is not borne out: that imprinted genes will be relatively rapidly evolving ones. This is because sexual antagonism would drive a molecular arms race in which each benefited from temporarily gaining the upper hand. A species-by-species comparison of imprinted genes does not bear this out. Rather, imprinted genes seem to evolve quite slowly. It looks increasingly as if the Haig theory explains some, but not all, cases of imprinting.

2 1 2 G E N O M E

Imprinting has a curious consequence. In a man, the maternal copy of chromosome 15 carries a mark that identifies it as coming from his mother, but when he passes it on to his son or daughter, it must somehow have acquired a mark that identifies it as coming from him: the father. It must switch from maternal to paternal and vice versa in the mother. That this switch does happen we know, because in a small proportion of people with Angelman syndrome there is nothing unusual about either chromosome except that both behave as if they were paternal. These are cases in which the switch failed to occur. They can be traced back to mutations in the previous generation, mutations that affect something called the imprinting centre, a small stretch of D N A close to both relevant genes, which somehow places the parental mark on the chromosome. The mark consists of one gene's methylation, of the kind encountered in chromosome 8.8

Methylation of the 'letter' C, you will recall, is the means by which genes are silenced, and it serves to keep selfish D N A under house arrest. But methylation is removed during the early development of the embryo - the creation of the so-called blastocyst — and then reimposed during the next stage of development, called gastrulation.