The Epigenetics Revolution

Chapter 8. The Battle of the Sexes

The laboratory stick insect Carausius morosus is a very popular pet. As long as it has a few privet leaves to munch on it will be perfectly content, and after a few months it will begin to lay eggs. In due course, these will hatch into perfect little baby stick insects, looking just like miniature versions of the adults. If one of these baby stick insects is removed as soon as it is born, and kept in a tank on its own, then it too will lay eggs which will hatch into little stick insects in their turn. This is despite the fact that it has never mated.

Stick insects frequently reproduce this way. They are using a mechanism known as parthenogenesis, from the Greek for ‘virgin birth’. Females lay fertile eggs without ever mating with a male, and perfectly healthy little stick insects emerge from these eggs. These insects have evolved with special mechanisms to ensure that the offspring have the correct number of chromosomes. But these chromosomes all came from the mother.

This is very different from mice and humans, as we saw in the last chapter. For us and our rodent relatives, the only way to generate live young is by having DNA from both a mother and a father. It’s tempting to speculate that stick insects are highly unusual but they’re not. We mammals are the exceptions. Insects, fish, amphibians, reptiles and even birds all have a few species that can reproduce parthenogenetically. It’s we mammals who can’t. It’s our class in the animal kingdom which is the odd one out, so it makes sense to ask why this is the case. We can begin by looking at the features which are found only in mammals. Well, we have hair, and we have three bones in our middle ear. Neither of these characteristics is found in the other classes, but it seems unlikely these are the key features that have led us to abandon virgin birth. For this issue there is a much more important characteristic.

The most primitive examples of mammals are the small number of creatures like the duck-billed platypus and the echidna, which lay eggs. After them, in terms of reproductive complexity, are the marsupials such as the kangaroo and the Tasmanian devil, which give birth to very under-developed young. The young of these species go through most of their developmental stages outside the mother’s body, in her pouch. The pouch is a glorified pocket on the outside of the body.

By far the greatest numbers of our class are called placental (or eutherian) mammals. Humans, tigers, mice, blue whales – we all nourish our young in the same way. Our offspring undergo a really long developmental phase inside the mother, in the uterus. During this developmental stage, the young get their nourishment via the placenta. This large, pancake-shaped structure acts as an interface between the blood system of the foetus and the blood system of the mother. Blood doesn’t actually flow from one to the other. Instead the two blood systems pass so closely to one another that nutrients such as sugars, vitamins, minerals and amino acids can pass from the mother to the foetus. Oxygen also passes from the mother’s blood to the foetal blood supply. In exchange, the foetus gets rid of waste gases and other potentially harmful toxins by passing them back into the mother’s circulation.

It’s a very impressive system, and allows mammals to nurture their young for long periods during early development. A new placenta is created in each pregnancy and the code for its production isn’t carried by the mother. It’s all coded for by the foetus. Think back yet again to our model of the early blastocyst in Chapter 2. All the cells of the blastocyst are descendants of the fertilised single-cell zygote. The cells that will ultimately become the placenta are the tennis ball cells on the outside of the blastocyst. In fact, one of the earliest decisions that cells make as they begin to roll down Waddington’s epigenetic landscape is whether they are turning into future placental cells, or future body cells.

While the placenta is a great method for nourishing a foetus, the system has ‘issues’. To use business or political speech, there’s a conflict of interest, because in evolutionary terms, our bodies are faced with a dilemma.

This is the evolutionary imperative for the male mammal, phrased anthropomorphically:

This pregnant female is carrying my genes in the form of this foetus. I may never mate with her again. I want my foetus to get as big as possible so that it has the greatest chance of passing on my genes.

For the female mammal, the evolutionary imperative is rather different:

I want this foetus to survive and pass on my genes. But I don’t want it to be at the cost of draining me so much that I never reproduce again. I want more than this one chance to pass on my genes.

This battle of the sexes in mammals has reached an evolutionary Mexican stand-off. A series of checks and balances ensures that neither the maternal nor the paternal genome gets the upper hand. We can get a better understanding of how this works if we look once again at the experiments of Azim Surani, Davor Sobel and Bruce Cattanach. These are the scientists who created the mouse zygotes that contained only paternal DNA or only maternal DNA.

After they had created these test tube zygotes, the scientists implanted them into the uterus of mice. None of the labs ever generated living mice from these zygotes. However, the zygotes did develop for a while in the womb, but very abnormally. The abnormal development was quite different, depending on whether all the chromosomes had come from the mother or the father.

In both cases the few embryos that did form were small and retarded in growth. Where all the chromosomes had come from the mother, the placental tissues were very underdeveloped[62]. If all the chromosomes came from the father, the embryo was even more retarded but there was much better production of the placental tissues[63]. Scientists created embryos from a mix of these cells – cells which had only maternally inherited or paternally inherited chromosomes. These embryos still couldn’t develop all the way to birth. When examined, the researchers found that all the tissues in the embryo were from the maternal-only cells whereas the cells of the placental tissues were the paternal-only type[64].

All these data suggested that something in the male chromosomes pushes the developmental programme in favour of the placenta, whereas a maternally-derived genome has less of a drive towards the placenta, and more towards the embryo itself. How is this consistent with the conflict or evolutionary imperative laid out earlier in this chapter? Well, the placenta is the portal for taking nutrients out of the mother and transferring them into the foetus. The paternally-derived chromosomes promote placental development, and thereby create mechanisms for diverting as much nutrition as possible from the mother’s bloodstream. The maternal chromosomes act in the opposite way, and a finely poised stalemate develops in normal pregnancies.

One obvious question is whether all the chromosomes are important for these effects. Bruce Cattanach used complex genetic experiments on mice to investigate this. The mice contained chromosomes that had been rearranged in different ways. The simplest way to explain this is that each mouse had the right amount of chromosomes, but they’d been ‘stuck together’ in unusual ways. He was able to create mice which had precise abnormalities in the inheritance of their chromosomes. For example, he could create mice which inherited both copies of a specific chromosome from just one parent.

The first experiments he reported were using mouse chromosome 11. For all the other pairs of chromosomes, the mice inherited one of each pair maternally, and one paternally. But for chromosome 11, Bruce Cattanach created mice that had inherited two copies from their mother and none from their father, or vice versa. Figure 8.1 represents the results[65].

^{Figure 8.1 Bruce Cattanach created genetically modified mice, in which he could control how they inherited a particular region of chromosome 11. The middle mouse inherited one copy from each parent. Mice which inherited both copies from their mother were smaller than this normal mouse. In contrast, mice which inherited both copies from their father were larger than normal.}

Once again this is consistent with the idea that there are factors in the paternal chromosomes that push towards development of larger offspring. Factors in the maternal chromosomes either act in the ‘opposite direction’ or are broadly neutral.

As we explored in the last chapter, these factors are epigenetic, not genetic. In the example above, let’s assume that the parents came from the same inbred mouse strain, so were genetically identical. If you sequenced both copies of chromosome 11 in any of the three types of offspring, they would be exactly the same. They would contain the same millions of A, C, G and T base-pairs, in the same order. But the two copies of chromosome 11 do clearly behave differently at a functional level, as shown by the different sizes of the different types of mice. Therefore there must be epigenetic differences between the maternal and paternal copies of chromosome 11.

Because the two copies of the chromosome behave differently depending on their parent-of-origin, chromosome 11 is known as an imprinted chromosome. It has been imprinted with information about its origins. As our understanding of genetics has improved we’ve realised that only certain stretches of chromosome 11 are imprinted. There are large regions where it doesn’t matter at all which parent donated which chromosome, and the regions from the two parents are functionally equivalent. There are also entire chromosomes that are not imprinted.

So far, we’ve described imprinting in mainly phenomenological terms. Imprinted regions are stretches of the genome where we can detect parent-of-origin effects in offspring. But how do these regions carry this effect? In imprinted regions, certain genes are switched on or switched off, depending on how they were inherited. In the chromosome 11 example above, genes associated with placental growth are switched on and are very active in the copy of the chromosome inherited from the father. This carries risks of nutrient depletion for the mother who is carrying the foetus, and a compensatory mechanism has evolved. The copies of these same genes on the maternal chromosome tend to be switched off, and this limits the placental growth. Alternatively, there may be other genes that counterbalance the effects of the paternal genes, and these counter-balancing genes may be expressed mainly from the maternal chromosome.

Major strides have been made in understanding the molecular biology of these effects. For example, later researchers worked on a region on chromosome 7 in mice. There is a gene in this region called insulin-like growth factor 2 (Igf2). The Igf2 protein promotes embryonic growth, and is normally expressed only from the paternally-derived copy of chromosome 7. Experimenters introduced a mutation into this gene, which stopped the gene coding for a functional Igf2 protein. They studied the effects of the mutation on offspring. When the mutation was passed on from the mother, the young mice looked the same as any other mice. This is because the Igf2 gene is normally switched off on the maternal chromosome anyway, and so it didn’t matter that the maternal gene was mutated. But when the mutant Igf2 gene was passed down from father to offspring, the mice in the litter were much smaller than usual. This was because the one copy of the Igf2 gene that they ‘relied on’ for strong foetal growth had been switched off by the mutation[66].

There is a gene on mouse chromosome 17 called Igf2r. The protein encoded by this gene ‘mops up’ Igf2 protein and stops it acting as a growth promoter. The Igf2r gene is also imprinted. Because Igf2r protein has the ‘opposite’ effect to Igf2 in terms of foetal growth, it probably isn’t surprising to learn that the Igf2r gene is usually expressed from the maternal copy of chromosome 17[67].

Scientists have detected about 100 imprinted genes in mice, and about half this number in humans. It’s not clear if there are genuinely fewer imprinted genes in humans than in mice, or if it’s just more difficult to detect them experimentally. Imprinting evolved about 150 million years ago[68], and it really only occurs to a great extent in placental mammals. It isn’t found in those classes that can reproduce parthenogenetically.

Imprinting is a complicated system, and like all complex machinery, it can break down. We now know that there are disorders in humans that are caused by problems with the imprinting mechanism.

Prader-Willi syndrome (PWS) is named after two of the authors of the first description of the condition[69]. PWS affects about one in 20,000 live births. The babies have a low birth weight and their muscles are really floppy. In early infancy, it can be difficult to feed these babies and initially they fail to thrive. This is dramatically reversed by early childhood. The children are constantly hungry, so over-eat to an incredible degree and can become dangerously obese. Along with other characteristic features such as small feet and hands, delayed language development and infertility, the individuals with PWS are often mildly or moderately mentally retarded. They may also have behavioural disturbances, including inappropriate temper outbursts[70].

There’s another disorder in humans that affects about the same number of people as PWS. This is called Angelman syndrome (AS), and like PWS it is named after the person who first described the condition[71]. Children with AS suffer from severe mental retardation, small brain size and very little speech. Patients with AS will often laugh spontaneously for no obvious reason, which led to the spectacularly insensitive clinical description of these children as ‘happy puppets’[72].

In both PWS and AS, the parents of the affected children are normally perfectly healthy. Research suggested that the basic problem in each of these conditions was likely to be caused by an underlying defect in the chromosomes. Because the parents were unaffected, the defect probably arose during the production of the eggs or the sperm.

In the 1980s, researchers working on PWS used a variety of standard techniques to find the underlying cause of this condition. They looked for regions of the genome that were different between healthy children and those with the disorder. Scientists interested in AS were doing something similar. By the mid-1980s it was becoming clear that both groups were looking at the same part of the genome, a specific stretch on chromosome 15. In both PWS and AS, patients had lost a small, identical section of this chromosome.

But these two disorders are very unlike each other in their clinical presentation. Nobody would ever confuse a patient with PWS with one who was suffering from Angelman’s syndrome. How could the same genetic problem – the loss of a key region of chromosome 15 – result in such different symptoms?

In 1989 a group from The Children’s Hospital, Boston, showed that the important feature was not just the deletion, but how the deletion was inherited. It’s summarised in Figure 8.2. When the abnormal chromosome was inherited from the father, the child had PWS. When the same chromosome abnormality was inherited from the mother, the child had AS[73].

^{Figure 8.2 Two children may each have the same deletion on chromosome 15, shown schematically by the absence of the horizontally striped box. The phenotype of the two children will be different, depending on how they inherited the abnormal chromosome. If the abnormal chromosome was inherited from their father, the child will develop Prader-Willi syndrome. If the abnormal chromosome was inherited from their mother, the child will develop Angelman syndrome, which is a very different disorder from Prader-Willi.}

This is a clear case of epigenetic inheritance of a disorder. Children with PWS and AS had exactly the same problem genetically – they were missing a specific region of chromosome 15. The only difference was how they inherited the abnormal chromosome. This is another example of a parent-of-origin effect.

There’s another way in which patients can inherit PWS or AS. Some patients with these disorders have two totally normal copies of chromosome 15. There are no deletions, and no other mutations of any type, and yet the children develop the conditions. To understand how this can be, it’s helpful to think back to the mice who inherited both copies of chromosome 11 from one parent. Some of the same researchers who unravelled the story of the PWS deletion showed that in certain examples of this condition, the children have two normal copies of chromosome 15. The trouble is, they’ve inherited both from their mother, and none from their father. This is known as uniparental disomy – one parent contributing two chromosomes[74]. In 1991, a team from the Institute of Child Health in London showed that some cases of AS were caused by the opposite form of uniparental disomy to PWS. The children had two normal copies of chromosome 15, but had inherited both from their father[75].

This reinforced the notion that PWS and AS are each examples of epigenetic diseases. The children with uniparental disomy of chromosome 15 had inherited exactly the right amount of DNA, they just hadn’t inherited it from each parent. Their cells contained all the correct genes, in all the correct amounts, and yet still they suffered from these severe disorders.

It’s important that we inherit this fairly small region of chromosome 15 in the right way because this region is normally imprinted. There are genes in this region that are only expressed from either the maternal or the paternal chromosome. One of these genes is called UBE3A. This gene is important for normal functioning in the brain, but it’s only expressed from the maternally inherited gene in this tissue. But what if a child doesn’t inherit a copy of UBE3A from its mother? This could happen if both copies of UBE3A came from the father, because of uniparental disomy of chromosome 15. Alternatively, the child might inherit a copy of chromosome 15 from its mother which lacked the UBE3A gene, because part of the chromosome had been lost. In these cases, the child can’t express UBE3A protein in its brain, and this leads to the development of the symptoms of Angelman syndrome.

Conversely, there are genes that are normally only expressed from the paternal version of this stretch of chromosome 15. This includes a gene called SNORD116, but others may also be important. The same scenario applies as for UBE3A, but replace the word maternal with paternal. If a child doesn’t inherit this region of chromosome 15 from its father, it develops Prader-Willi syndrome.

There are other examples of imprinting disorders in humans. The most famous is called Beckwith-Wiedemann syndrome, again named after the people who first described it in the medical literature[76][77]. This disorder is characterised by over-growth of tissues, so that the babies are born with over-developed muscles including the tongue, and a range of other symptoms[78]. This condition has a slightly different mechanism to the ones described above. When imprinting goes wrong in Beckwith-Wiedemann syndrome, both the maternal and paternal copies of a gene on chromosome 11 get switched on, when normally only the paternally-derived version should be expressed. The key gene seems to be IGF2, which codes for the growth factor protein that we met earlier, on mouse chromosome 7. By expressing two copies of this gene, rather than just one, twice as much IGF2 protein as normal is produced and the foetus grows too much.

The opposite phenotype to Beckwith-Wiedemann syndrome is a condition called Silver-Russell syndrome[79][80]. Children with this disorder are characterised by retarded growth before and after birth and other symptoms associated with late development[81]. Most cases of this condition are also caused by problems in the same region of chromosome 11 as in Beckwith-Wiedemann syndrome, but in Silver-Russell syndrome IGF2 protein expression is depressed, and the growth of the foetus is dampened down.

So, imprinting refers to a situation where there is expression of only one member of a pair of genes, and the expression may be either maternal or paternal. What controls which gene is switched on? It probably isn’t surprising to learn that DNA methylation plays a really big role in this. DNA methylation switches genes off. Therefore, if a paternally-inherited region of a chromosome is methylated, the paternally-derived genes in this region will be repressed.

Let’s take the example of the UBE3A gene which we encountered in the discussion of Prader-Willi and Angelman syndromes. Normally, the copy inherited from the father contains methylated DNA and the gene is switched off. The copy inherited from the mother doesn’t have this methylation mark, and the gene is switched on. Something similar happens with Igf2r in mice. The paternal version of this is usually methylated, and the gene is inactive. The maternal version is non-methylated and the gene is expressed.

While a role for DNA methylation may not have come as a shock, it may be surprising to learn that it is often not the gene body that is methylated. The part of the gene that codes for protein is epigenetically broadly the same when we compare the maternal and paternal copies of the chromosome. It’s the region of the chromosome that controls the expression of the gene that is differently methylated between the two genomes.

Imagine a night-time summer party in a friend’s garden, beautifully lit by candles scattered between the plants. Unfortunately, this lovely ambience is constantly ruined because the movement of the guests keeps triggering a motion detector on a security system and turning on a floodlight. The floodlight is too high on the wall to be able to cover it, but finally it dawns on the guests that they don’t need to cover the light. They need to cover the sensor that is triggering the light’s activity. This is very much what happens in imprinting.

The methylation, or lack of it, is on regions known as imprinting control regions (ICRs). In some cases, imprinting control is very straightforward to understand. The promoter region of a gene is methylated on the gene inherited from one parent, and not on the one from the other. This methylation keeps a gene switched off. This works when there is a single gene in a chromosome region that is imprinted. But many imprinted genes are arranged in clusters, all very close to one another in a single stretch on one chromosome. Some of the genes in the cluster will be expressed from the maternally-derived chromosome, others from the paternally-derived one. DNA methylation is still the key feature, but other factors help it to carry out its function.

The imprinting control region may operate over long distances, and certain stretches may bind large proteins. These proteins act like roadblocks in a city, insulating different stretches on a chromosome from one another. This gives the imprinting process an additional level of sophistication, by inserting diversions between different genes. Because of this, an imprinting control region may operate over many thousands of base-pairs, but it doesn’t mean that every single gene in those thousands of base-pairs is affected the same way. Different genes in a particular imprinted stretch of chromatin may loop out from their chromosome to form physical associations with each other, so that repressed genes huddle together in a sort of chromatin knot. Activated genes from the same stretch of chromosome may cling together in a different bundle[82].

The impact of imprinting varies from tissue to tissue. The placenta is particularly rich in expression of imprinted genes. This is what we would expect from our model of imprinting as a means of balancing out the demand on maternal resources. The brain also appears to be very susceptible to imprinting effects. It’s not so clear why this should be the case. It’s harder to reconcile parent-of-origin control of gene expression in the brain with the battle for nutrients we’ve been considering so far. Professor Gudrun Moore of University College London has made an intriguing suggestion. She has proposed that the high levels of imprinting in the brain represent a post-natal continuation of the war of the sexes. She has speculated that some brain imprints are an attempt by the paternal genome to promote behaviour in young offspring that will stimulate the mother to continue to drain her own resources, for example by prolonged breast-feeding[83].

The number of imprinted genes is quite low, rather less than 1 per cent of all protein-coding genes. Even this small percentage won’t be imprinted in all tissues. In many cells the expression from the maternally and paternally-derived copies will be the same. This is not because the methylation pattern is different between the tissues but because cells vary in the ways that they ‘read’ this methylation.

The DNA methylation patterns on the imprinting control regions are present in all the cells of the body, and show which parent transmitted which copy of a chromosome. This tells us something very revealing about imprinted regions. They must evade the reprogramming that takes place after the sperm and egg fuse to form the zygote. Otherwise, the methylation modifications would be stripped off and there would be no way for the cell to work out which parent had donated which chromosome. Just as the IAP retrotransposons stay methylated during zygotic reprogramming, mechanisms have evolved to protect imprinted regions from this broad-brush removal of methylation. It’s not really very clear how this happens, but it’s essential for normal development and health.

Yet this presents us with a bit of a problem. If imprinted DNA methylation marks are so stable, how do they change as they are transmitted from parent to offspring? We know that they do, because of Azim Surani’s experiments with mice that we encountered in the previous chapter. These showed how methylation of a sequence monitored for experimental purposes changed as it was passed down the generations. This was the experiment that was described using the mice with ‘black’ and ‘white’ DNA in the previous chapter.

In fact, once scientists recognised that parent-of-origin effects exist, they predicted that there must be a way to reset the epigenetic marks, even before they knew what these marks were. Let’s consider chromosome 15, for example. I inherited one copy from my mother and one from my father. The UBE3A imprinting control region from my mother was unmethylated, whereas the same region on the chromosome from my father was methylated. This ensured appropriate expression patterns of UBE3A protein in my brain.

When my ovaries produce eggs, each egg inherits just one copy of chromosome 15, which I will pass on to a child. Because I’m a woman, each copy of chromosome 15 must carry a maternal mark on UBE3A. But one of my copies of chromosome 15 has been carrying the paternally-derived mark I inherited from my father. The only way I can make sure that I pass on chromosome 15 with the correct maternal mark to my children is if my cells have a way of removing the paternal mark and replacing it with a maternal one.

A very similar process would have to take place when males produce sperm. All maternally-derived modifications would need to be stripped off the imprinted genes, and paternally derived ones put on in their place. This is indeed exactly what happens. It’s a very restricted process which only takes place in the cells that give rise to the germ line.

The general principle is shown diagrammatically in Figure 8.3.

Following fusion of the egg and sperm the blastocyst forms, and most regions of the genome become reprogrammed. The cells begin to differentiate, forming the precursors to the placenta and also the various cell types of the body. So, at this point the cells that had been part of the ICM are all marching to the developmental drumbeat, heading down the various troughs in Waddington’s epigenetic landscape. But a very small number (less than 100) begin to march to a different beat. In these cells a gene called Blimp1 switches on. Blimp1 protein sets up a new cascade in signalling, which stops the cells heading towards their somatic dead-ends. These cells start travelling back up Waddington’s trenches[84]. They also lose the imprinted marks which told the cell which parent donated which of a pair of chromosomes.

^{Figure 8.3 Diagram showing how the somatic cells arising from a fertilised zygote all carry the same DNA methylation patterns as each other at imprinted genes, but the imprinting methylation is removed and then re-established in the germ cells. This ensures that females only pass on maternal marks to their offspring, and males only pass on paternal ones.}

The tiny population of cells that carry out this process are know as the primordial germ cells. It’s these cells that will ultimately settle in the developing gonads (testicles or ovaries) and act as the stem cells that produce all the gametes (sperm or eggs respectively). In the stage described in the previous paragraph, the primordial germ cells are reverting to a state more like that of the cells of the inner cell mass (ICM). Essentially, they are becoming pluripotent, and potentially able to code for most of the tissue types in the body. This phase is fleeting. The primordial germ cells quickly get diverted into a new developmental pathway where they differentiate to form stem cells that will give rise to eggs or sperm. To do so, they gain a new set of epigenetic modifications. Some of these modifications are ones that define cellular identity, i.e. switch on the genes that make an egg an egg. But a small number are the ones that serve as parent-of-origin marks, so that in the next generation the imprinted regions of the genome can be recognised with respect to their parent-of-origin.

This seems horribly complicated. If we follow the path from the sperm that fertilised the egg to a new sperm being formed in male offspring, the sequence goes like this:

The sperm that enters the egg has epigenetic modifications on it;

The epigenetic modifications get taken off, except at the imprinted regions (in the immediate post-fertilisation zygote);

Epigenetic modifications get put on (as the cells of the ICM begin to specialise);

The epigenetic modifications get taken off, including at the imprinted regions (as the primordial germ cells break away from the somatic differentiation pathway);

Epigenetic modifications get put on (as the sperm develops).

This could seem like an unnecessarily complicated way to get back to where we started from, but it’s essential.

The modifications that make a sperm a sperm, or an egg an egg, have to come off at stage 2 or the zygote wouldn’t be totipotent. Instead it would have a genome that was half-programmed to be an egg and half-programmed to be a sperm. Development wouldn’t be possible if the inherited modifications stayed on. But to create primordial germ cells, some of the cells from the differentiating ICM have to lose their epigenetic modifications. This is so they can become temporarily more pluripotent, lose their imprinting marks and transfer across into the germ cell lineage.

Once the primordial germ cells have been diverted, epigenetic modifications again get attached to the genome. This is partly because pluripotent cells are potentially extremely dangerous as a multi-cellular organism develops. It might seem like a great idea to have cells in our body that can divide repeatedly and give rise to lots of other cell types, but it’s not. Those sorts of cells are the type that we find in cancer. Evolution has favoured a mechanism where the primordial germ cells can regain pluripotency for a period, but then this pluripotency is re-suppressed by epigenetic modifications. Coupled with this, the wiping out of the imprints means that chromosomes can be marked afresh with their parent-of-origin.

Occasionally this process of setting up the new imprints on the progenitors of egg or sperm can go wrong. There are cases of Angelman syndrome and Prader-Willi syndrome where the imprint has not been properly erased during the primordial germ cell stage[85]. For example, a woman may generate eggs where chromosome 15 still has the paternal mark on it that she inherited from her father, rather than the correct maternal mark. When this egg is fertilised by a sperm, both copies of chromosome 15 will function like paternal chromosomes, and create a phenotype just like uniparental disomy.

Research is ongoing into how all these processes are controlled. We don’t fully understand how imprints are protected from reprogramming following fusion of the egg and the sperm, nor how they lose this protection during the primordial germ cell stage. We’re also not entirely sure how imprints get put back on in the right place. The picture is still quite foggy, although details are starting to emerge from the haze.

Part of this may involve the small percentage of histones that are present in the sperm genome. Many of these are located at the imprinting control regions, and may protect these regions from reprogramming when the sperm and the egg fuse[86]. Histone modifications also play a role in establishing ‘new’ imprints during gamete production. It seems to be important that the imprinting control regions lose any histone modifications that are associated with switching genes on. Only then can the permanent DNA methylation be added[87]. It’s this permanent DNA methylation that marks a gene with a repressive imprint.

The reprogramming events in the zygote and in primordial germ cells impact on a surprising number of epigenetic phenomena. When somatic cells are reprogrammed in the laboratory using the Yamanaka factors, only a tiny percentage of them form iPS cells. Hardly any seem to be exactly the same as ES cells, the genuinely pluripotent cells from the inner cell mass of the blastocyst. A group in Boston, based at Massachusetts General Hospital and Harvard University, assessed genetically identical iPS and ES cells from mice. They looked for genes that varied in expression between the two types of cells. The only major differences in expression were in a chromosomal region known as Dlk1-Dio3[88]. A few iPS cells expressed the genes in this region in a way that was very similar to how the ES cell did this. These were the best iPS cells for forming all the different tissues of the body.

Dlk1-Dio3 is an imprinted region on chromosome 12 of the mouse. It’s perhaps not surprising that an imprinted region turned out to be so important. The Yamanaka technique triggers the reprogramming process that normally occurs when a sperm fuses with an egg. Imprinted regions of the genome are resistant to reprogramming in normal development. It is likely that they present too high a barrier to reprogramming in the very artificial environment of the Yamanaka method.

The Dlk1-Dio3 region has been of interest to researchers for quite some time. In humans, uniparental disomy in this region is associated with growth and developmental defects, amongst other symptoms[89]. This region has also been shown to be critical for the prevention of parthenogenesis, at least in mice. Researchers from Japan and South Korea genetically manipulated just this region of the genome in mice. They reconstructed a fertilised egg with two female pronuclei. The Dlk1-Dio3 region in one of the pronuclei had been altered so that it carried the equivalent of a paternal rather than maternal imprint. The live mice that were born were the first example of a placental mammal with two maternal genomes[90].

The reprogramming that occurs in the primordial germ cells isn’t completely comprehensive. It leaves the methylation on some IAP retrotransposons more or less intact. The DNA methylation level of the AxinFu retrotransposon in sperm is the same as it is in the body cells of this strain of mice. This shows that the DNA methylation was not removed when the PGCs were reprogrammed, even though most other areas of the genome did lose this modification. This resistance of the AxinFu retrotransposon to both rounds of epigenetic reprogramming (in the zygote and in the primordial germ cells) provides a mechanism for the transgenerational inheritance of the kinked tail trait that we met in earlier chapters.

We know that not all transgenerational inheritance happens in the same way. In the agouti mouse the phenotype is transmitted via the mother, but not via the father. In this case, the DNA methylation on the IAP retrotransposon is removed in both males and females during normal primordial germ cell reprogramming. However, mothers whose retrotransposon originally carried DNA methylation pass on a specific histone mark to their offspring. This is a repressive histone modification and it acts as a signal to the DNA methylation machinery. This signal attracts the enzymes that put the repressive DNA methylation onto a specific region on a chromosome. The final outcome is the same – the DNA methylation in the mother is restored in the offspring. Male agouti mice don’t pass on either DNA methylation or repressive histone modifications on their retrotransposon, which is why transmission of the phenotype only occurs through the maternal line[91].

This is a slightly more indirect method of transmitting epigenetic information. Instead of direct carry-over of DNA methylation, an intermediate surrogate (a repressive histone modification) is used instead. This is probably why the maternal transmission of the agouti phenotype is a bit ‘fuzzy’. Not all offspring are exactly the same as the mother, because there is a bit of ‘wriggle-room’ in how DNA methylation gets re-established in the offspring.

In the summer of 2010, there were reports in the British press about cloned farm animals. Meat that had come from the offspring of a cloned cow had entered the human food chain[92]. Not the cloned cow itself, just its offspring, created by conventional animal breeding. Although there were a few alarmist stories about people unwittingly eating ‘Frankenfoods’, the coverage in the mainstream media was pretty balanced.

To some extent, this was probably because of a quite intriguing phenomenon, which has allayed certain fears originally held by scientists about the consequences of cloning. When cloned animals breed, the offspring tend to be healthier than the original clones. This is almost certainly because of primordial germ cell reprogramming. The initial clone was formed by transfer of a somatic nucleus into an egg. This nucleus only went through the first round of reprogramming, the one that normally happens when a sperm fertilises an egg. The likelihood is that this epigenetic reprogramming wasn’t entirely effective – it’s a big ask to get an egg to reprogram a ‘wrong’ nucleus. This is likely to be the reason why clones tend to be unhealthy.

When the cloned animals breed, they pass on either an egg or a sperm. Before the clone produced these gametes, its primordial cells underwent the second round of reprogramming, as part of the normal primordial germ cell pathway. This second reprogramming stage seems to reset the epigenome properly. The gametes lose the abnormal epigenetic modifications of their cloned parent. Epigenetics explains why cloned animals have health issues, but also explains why their offspring don’t. In fact, the offspring are essentially indistinguishable from animals produced naturally.

Assisted reproductive technologies in humans (such as in vitro fertilisation) share certain technical aspects with some of the methods used in cloning. In particular, pluripotent nuclei may be transferred between cells, and cells are cultured in the laboratory before being implanted in the uterus. There is a substantial amount of controversy in the scientific journals about the abnormality rates from these procedures[93]. Some authors claim there is an increased rate of imprinting disorders in pregnancies from assisted reproductive technologies. This would imply that procedures such as culturing fertilised eggs outside the body may disrupt the delicately poised pathways that control reprogramming, especially of imprinted regions. It’s important to note, however, that there is no consensus yet on whether this really is a clinically relevant issue.

All the reprogramming of the genome in early development has multiple effects. It allows two highly differentiated cell types to fuse and form one pluripotent cell. It balances out the competing demands of the maternal and paternal genomes, and ensures that this balancing act can be re-established in every generation. Reprogramming also prevents inappropriate epigenetic modifications being passed from parent to offspring. This means that even if cells have accumulated potentially dangerous epigenetic changes, these will be removed before they are passed on.

This is why we don’t normally inherit acquired characteristics. But there are certain regions of the genome, such as IAP retrotransposons, that are relatively resistant to reprogramming. If we want to work out how certain acquired characteristics – responses to vinclozolin or responses to paternal nutrition, for example – get transmitted from parent to offspring, these IAP retrotransposons might be a good place to start looking.