The Epigenetics Revolution

Chapter 5. Why Aren’t Identical Twins Actually Identical?

Identical twins have been a source of fascination in human cultures for millennia, and this fascination continues right into the present day. Just taking Western European literature as one source, we can find the identical twins Menaechmus and Sosicles in a work of Plautus from around 200 B.C.; the re-working of the same story by Shakespeare in The Comedy of Errors, written around 1590; Tweedledum and Tweedledee in Lewis Carroll’s Through the Looking-Glass, and What Alice Found There written in 1871; right up to the Weasley twins in the Harry Potter novels of J. K. Rowling. There is something inherently intriguing about two people who seem exactly the same as one another.

But there is something that interests all of us even more than the extraordinary similarities of identical twins, and that is when we can see their differences. It’s a device that’s been repeatedly used in the arts, from Frederic and Hugo in Jean Anhouil’s Ring around the Moon to Beverley and Elliott Mantle in David Cronenberg’s Dead Ringers. Taking this to its extreme you could even cite Dr Jekyll and his alter ego Mr Hyde, the ultimate ‘evil twin’. The differences between identical twins have certainly captured the imaginations of creative people from all branches of the arts, but they have also completely captivated the world of science.

The scientific term for identical twins is monozygotic (MZ) twins. They were both derived from the same single-cell zygote formed from the fusion of one egg and one sperm. In the case of MZ twins the inner cell mass of the blastocyst split into two during the early cell divisions, like slicing a doughnut in half, and gave rise to two embryos. And these embryos are genetically identical.

This splitting of the inner cell mass to form two separate embryos is generally considered a random event. This is consistent with the frequency of MZ twins being pretty much the same throughout all human populations, and with the fact that identical twins don’t run in families. We tend to think of MZ twins as being very rare but this isn’t really the case. About one in every 250 full-term pregnancies results in the birth of a pair of MZ twins, and there are around ten million pairs of identical twins around the world today.

MZ twins are particularly fascinating because they help us to determine the degree to which genetics is the driving force for life events such as particular illnesses. They basically allow us to explore mathematically the link between the sequences of our genes (genotype) and what we are like (phenotype), be this in terms of height, health, freckles or anything else we would like to measure. This is done by calculating how often both twins in a pair present with the same disease. The technical term for this is the concordance rate.

Achondroplasia, a relatively common form of short-limbed dwarfism, is an example of a condition in which MZ twins are almost invariably affected in the same way. If one twin has achondroplasia, so does the other one. The disease is said to show 100 per cent concordance. This isn’t surprising as achondroplasia is caused by a specific genetic mutation. Assuming that the mutation was present in either the egg or the sperm that fused to form the zygote, all the daughter cells that form the inner cell mass and ultimately the two embryos will also carry the mutation.

However, relatively few conditions show 100 per cent concordance, as the majority of illnesses are not caused by one overwhelming mutation in a key gene. This creates the problem of how to determine if genetics plays a role, and if so, how great this role is. This is where twin studies have become so valuable. If we study large groups of MZ twins we can determine what percentage of them is concordant or discordant for a particular condition. If one twin has a disease, does the other twin also tend to develop it as well?

Figure 5.1 is a graph showing concordance rates for schizophrenia. This shows that the more closely related we are to someone with this disease, the more likely we are to develop it ourselves. The most important parts of the graph to look at are the two bars at the bottom, which deal with twins. From this we can compare the concordance rates for identical and non-identical (fraternal) twins. Non-identical twins share the same developmental environment (the uterus) but genetically are no more similar than any other pair of siblings, as they arose from two separate zygotes as a consequence of the fertilisation of two eggs. The comparison between the two types of twins is important because generally speaking, the twins in a pair (whether identical or non-identical) are likely to have shared pretty similar environments. If schizophrenia was caused mainly by environmental factors, we would expect the concordance rates for the disease to be fairly similar between identical and non-identical twins. Instead, what we see is that in non-identical twins, if one twin develops schizophrenia, the other twin has a 17 per cent chance of doing the same. But in MZ twins this risk jumps to nearly 50 per cent. The almost three-fold higher risk for identical versus non-identical twins tells us that there is a major genetic component to schizophrenia.

^{Figure 5.1 The concordance rates for schizophrenia. The more genetically related two individuals are, the more likely it is that if one individual has the disease, their relative will also develop the disorder. However, even in genetically identical monozygotic twins, the concordance rate for schizophrenia does not reach 100 per cent. Data taken from The Surgeon General’s Report on Mental Health http://www.surgeongeneral.gov/library/mentalhealth/chapter4/sec4_1.html#etiology}

Similar studies have shown that there is also a substantial genetic component to a significant number of other human disorders, including multiple sclerosis, bipolar disorder, systemic lupus erythematosus and asthma. This has been really useful in understanding the importance of genetic susceptibility to complex diseases.

But in many ways, it’s the other side of the question that is more interesting. It’s not the MZ twins who both develop a specific disease who are most interesting. It’s the MZ twins who end up with very different outcomes – one a paranoid schizophrenic, one mentally very healthy, for example – who create the most intriguing scientific problem. Why do two genetically identical individuals, who in many cases have experienced very similar environments, have such variable phenotypes? Similarly, why is it quite rare for both MZ twins in a pair to develop type 1 diabetes? What is it, in addition to the genetic code, that governs these health outcomes?

One possible explanation would be that quite randomly the twin with schizophrenia had spontaneously developed mutations in genes in certain cells, for example in the brain. This could happen if the DNA replication machinery had malfunctioned at some point during brain development. These changes might increase his or her susceptibility to a disorder. This is theoretically possible, but scientists have failed to find much data to support this theory.

Of course, the standard answer has always been that discordancy between the twins is due to differences in their environments. Sometimes this is clearly true. If we were monitoring longevity, for example, one twin getting knocked over and killed by a number 47 bus would certainly represent an environmental difference. But this is an extreme scenario. Many twins share a fairly similar environment, especially in early development. Even so, it is certainly possible that there are multiple subtle environmental differences that may be hard to monitor appropriately.

But if we invoke the environment as the other important factor in development of disease, this raises another problem. It still leaves the question of how the environment does this. Somehow the environmental stimuli – be these compounds in our food, chemicals in cigarette smoke, UV rays in sunlight, pollutants from car exhausts or any of the thousands of molecules and radiation sources that we’re exposed to every day – must impact on our genes and cause a change in expression.

The majority of non-infectious diseases that afflict most people take a long time to develop, and then remain as a problem for many years if there is no cure available. The stimuli from the environment could theoretically be acting on the genes all the time in the cells that are acting abnormally, leading to disease. But this seems unlikely, especially because most of the chronic diseases probably involve the interaction of multiple stimuli with multiple genes. It’s hard to imagine that all these stimuli would be present for decades at a time. The alternative is that there is a mechanism that keeps the disease-associated cells in an abnormal state, i.e. expressing genes inappropriately.

In the absence of any substantial evidence for a role for somatic mutation, epigenetics seems like a strong candidate for this mechanism. This would allow the genes in one twin to stay mis-regulated, ultimately leading to a disease. We’re only at the beginning of the investigation but some evidence has started accumulating that suggests this may indeed be the case.

One of the most straightforward experiments conceptually, is to analyse if chromatin modification patterns (the epigenome) change as MZ twins get older. In the simplest case, we wouldn’t even need to investigate this in the context of disease. We could start by testing a much simpler hypothesis – that genetically identical individuals become epigenetically non-identical as they age. If this hypothesis is correct, this would support the idea that MZ twins can vary from each other at the epigenetic level. This in turn would strengthen our confidence in moving forwards to examining the role of epigenetic changes in disease.

In 2005, a large collaborative group headed by Professor Manel Esteller, then at the Spanish National Cancer Centre in Madrid, published a paper in which they examined this issue[31]. They made some interesting discoveries. If they examined chromatin from infant MZ twin pairs, they couldn’t see much difference in the levels of DNA methylation or of histone acetylation between the two twins. When they looked at pairs of MZ twins who were much older, such as in their fifties, there was a lot of variation within the pair for the amount of DNA methylation or histone acetylation. This seemed to be particularly true of twins that had lived apart for a long time.

The results from this study were consistent with a model where genetically identical twins start out epigenetically very similar, and then diverge as they get older. The older MZ twins who had led separate and different lives for the longest would be expected to be the ones who had encountered the greatest differences in their environments. The finding that these were precisely the twin pairs who were most different epigenetically was consistent with the idea that the epigenome (the overall pattern of epigenetic modifications on the genome) reflects environmental differences.

Children who eat breakfast are statistically more likely to do well at school than children who skip breakfast. This doesn’t necessarily mean that learning can be improved by a bowl of cornflakes. It may simply be that children who eat breakfast are more likely to be children whose parents make an effort to get them to school every day, on time, and help them with their studies. Similarly, Professor Esteller’s data are correlative. They show there is a relationship between the ages of twins and how different they are epigenetically, but they don’t prove that age has caused the change in the epigenome. But at least the hypothesis can remain in play.

A team led by Dr Jeffrey Craig in 2010 at the Royal Children’s Hospital in Melbourne also examined DNA methylation in identical and fraternal twin pairs[32]. They investigated a few relatively small regions of the genome in greater detail than in Manel Esteller’s earlier paper. Using samples just from newborn twin pairs, they showed that there was a substantial amount of difference between the DNA methylation patterns of fraternal twins. This isn’t unexpected, since fraternal twins are genetically non-identical and we expect different individuals to have different epigenomes. Interestingly, though, they also found that even the MZ twins differed in their DNA methylation patterns, suggesting identical twins begin to diverge epigenetically during development in the uterus. Combining the information from the two papers, and from additional studies, we can conclude that even genetically identical individuals are epigenetically distinct by the time of birth, and these epigenetic differences become more pronounced with age and exposure to different environments.

These data are consistent with a model where epigenetic changes could account for at least some of the reasons why MZ twins aren’t phenotypically identical, but there’s still a lot of supposition involved. That’s because for many purposes humans are a quite hopeless experimental system. If we want to be able to assess the role of epigenetics in the problem of why genetically identical individuals are phenotypically different from one another, we would like to be able to do the following:

Analyse hundreds of identical individuals, not just pairs of them;

Manipulate their environments, in completely controlled ways;

Transfer embryos or babies from one mother to another, to investigate the effects of early nurture;

Take all sorts of samples from the different tissues of the body, at lots of different time points;

Control who mates with whom;

Carry out studies on four or five generations of genetically identical individuals.

Needless to say, this isn’t feasible for humans.

This is why experimental animals have been so useful in epigenetics. They allow scientists to address really complex questions, whilst controlling the environment as much as possible. The data that are generated in these animal studies produce insights from which we can then try to infer things about humans.

The match may not be perfect, but we can unravel a surprising amount of fundamental biology this way. Various comparative studies have shown that many systems have stayed broadly the same in different organisms over almost inconceivably long periods. The epigenetic machinery of yeast and humans, for example, share more similarities than differences and yet the common ancestor for the two species lies about one billion years in the past[33]. So, epigenetic processes are clearly fairly fundamental things, and using model systems can at least point us in a helpful direction for understanding the human condition.

In terms of the specific question we’ve been looking at in this chapter – why genetically identical twins often don’t seem to be identical – the animal that has been most useful is our close mammalian relative, the mouse. The mouse and human lineages separated a mere 75 million or so years ago[34]. 99 per cent of the genes found in mice can also be detected in humans, although they aren’t generally absolutely identical between the two species.

Scientists have been able to create strains of mice in which all the individuals are genetically identical to each other. These have been incredibly useful for investigating the roles of non-genetic factors in creating variation between individuals. Instead of just two genetically identical individuals, it’s possible to create hundreds, or thousands. The way this is done would have made even the Ptolemy dynasty of ancient Egypt blush. Scientists mate a pair of mice who are brother and sister. Then they mate a brother and sister from the resulting litter. They then mate a brother and sister from their litter and so on. When this is repeated for over twenty generations of brother-sister matings, all the genetic variation gets bred out, throughout the genome. All mice of the same sex from the strain are genetically identical. In a refinement of this, scientists can take these genetically identical mice and introduce just one change into their DNA. They may use such genetic engineering to create mice which are identical except for just one region of DNA that the experimenters are most interested in.

The most useful mouse model for exploring how epigenetic changes can lead to phenotypic differences between genetically identical individuals is called the agouti mouse. Normal mice have hair which is banded in colour. The hair is black at the tip, yellow in the middle and black again at the base. A gene called agouti is essential for creating the yellow bit in the middle, and is switched on as part of a normal cyclical mechanism in mice.

There is a mutated version of the agouti gene (called a) which never switches on. Mice that only have the a, mutant version of agouti have hair which is completely black. There is also a particular mutant mouse strain called Avy, which stands for agouti viable yellow. In Avy mice, the agouti gene is switched on permanently and the hair is yellow through its entire length. Mice have two copies of the agouti gene, one inherited from the mother and one from the father. The Avy version of the gene is dominant to the a version, which means that if one copy of the gene is Avy and one is a, the Avy will ‘overrule’ a and the hairs will be yellow throughout their length. This is all summarised in Figure 5.2.

^{Figure 5.2 Hair colour in mice is affected by the expression of the agouti gene. In normal mice, the agouti protein is expressed cyclically, leading to the characteristic brindled pattern of mouse fur. Disruption of this cyclical pattern of expression can lead to hairs which are either yellow or black throughout their length.}

Scientists created a strain of mice that contained one copy of Avy and one copy of a in every cell. The nomenclature for this is Avy/a. Since Avy is dominant to a, you would predict that the mice would have completely yellow hair. Since all the mice in the strain are genetically identical, you would expect that they would all look the same. But they don’t. Some have the very yellow fur, some the classic mouse appearance caused by the banded fur, and some are all shades in-between, as shown in Figure 5.3.

^{Figure 5.3 Genetically identical mice showing the extent to which fur colour can vary, depending on expression of the agouti protein. Photo reproduced with the kind permission of Professor Emma Whitelaw.}

This is really odd, since the mice are all genetically exactly the same. All the mice have the same DNA code. We could argue that perhaps the differences in coat colour are due to environment, but laboratory conditions are so standardised that this seems unlikely. It’s also unlikely because these differences can be seen in mice from the same litter. We would expect mice from a single litter to have very similar environments indeed.

Of course, the beauty of working with mice, and especially with highly inbred strains, is that it’s relatively easy to perform detailed genetic and epigenetic studies, especially when we already have a reasonable idea of where to look. In this case, the region to examine was the agouti gene.

Mouse geneticists knew how the yellow phenotype was caused in Avy yellow mice. A piece of DNA had been inserted in the mouse chromosome just before the agouti gene. This piece of DNA is called a retrotransposon, and it’s one of those DNA sequences that doesn’t code for a protein. Instead, it codes for an abnormal piece of RNA. Expression of this RNA messes up the usual control of the downstream agouti gene and keeps the gene switched on continuously. This is why the hairs on the Avy mice are yellow rather than banded.

That still doesn’t answer the question of why genetically identical Avy/a mice had variable coat colour. The answer to this has been shown to be due to epigenetics. In some Avy/a mice the CpG sequences in the retrotransposon DNA have become very heavily methylated. As we saw in the previous chapter, DNA methylation of this kind switches off gene expression. The retrotransposon no longer expressed the abnormal RNA that messed up transcription from the agouti gene. These mice were the ones with fairly normal banded mouse coat colour. On other genetically identical Avy mice, the retrotransposon was unmethylated. It produced its troublesome RNA which messed up the transcription from the agouti gene so that it was switched on continuously and the mice were yellow. Mice with in-between levels of retrotransposon methylation had in-between levels of yellow fur. This model is shown in Figure 5.4.

^{Figure 5.4 Variations in DNA methylation (represented by black circles) influence expression of a retrotransposon. The variation in expression of the retrotransposon in turn affects expression of the agouti gene, leading to coat colour variability between genetically identical animals.}

Here, DNA methylation is effectively working like a dimmer switch. When the retrotransposon is unmethylated, it shines to its fullest extent, producing lots of the abnormal RNA. The more the retrotranposon is methylated, the more its expression gets turned down.

The agouti mouse has provided a quite clear-cut example of how epigenetic modification, in this case DNA methylation, can make genetically identical individuals look phenotypically different. However, there is always the fear that agouti is a special case, and maybe this is a very uncommon mechanism. This is particularly of concern because it’s proved very difficult to find an agouti gene in humans – it seems to be in that 1 per cent of genes we don’t share with our mouse neighbours.

There is another interesting condition found in mice, in which the tail is kinked. This is called Axin-fused and it also demonstrates extreme variability between genetically identical individuals. This has been shown to be another example where the variability is caused by differing levels of DNA methylation in a retrotransposon in different animals, just like the agouti mouse.

This is encouraging as it suggests this mechanism isn’t a one off, but kinked tails still don’t really represent a phenotype that is of much concern to the average human. But there’s something we can all get on board with: body weight. Genetically identical mice don’t all have the same body weight.

No matter how tightly scientists control the environment for the mice, and especially their access to food, identical mice from inbred mouse strains don’t all have exactly the same body weight. Experiments carried out over many years have shown that only about 20–30 per cent of the variation in body weights can be attributed to the post-natal environment. This leaves the question of what causes the other 70–80 per cent of variation in body weight[35]. Since it isn’t being caused by genetics (all the mice are identical) or by the environment, there has to be another source for the variation.

In 2010, Professor Emma Whitelaw, the terrifically enthusiastic and intensely rigorous mouse geneticist working at the Queensland Institute of Medical Research, published a fascinating paper. She used an inbred strain of mice and then used genetic engineering to create subsets of animals which were genetically identical to the starting stock, except that they only expressed half of the normal levels of a particular epigenetic protein. She performed the genetic engineering independently in a number of mice, so that she could create separate groups of animals, each of which was mutated in a different gene coding for epigenetic proteins.

When Professor Whitelaw analysed the body weights of large numbers of the normal or mutated mice, an interesting effect appeared. In a group of normal inbred mice, most of the animals had relatively similar body weights, within the ranges found in many other studies. In the mice with low levels of a certain epigenetic protein, there was a lot more variability in the body weights within the group. Further experiments published in the same paper assessed the effects of the decreased expression of these epigenetic proteins. Their decreased expression was linked to changes in expression levels of selected genes involved in metabolism[36], and increased variability in that expression. In other words, the epigenetic proteins were exerting some control over the expression of other genes, just as we might expect.

Emma Whitelaw tested a number of epigenetic proteins in her system, and found that only a few of them caused the increased variation in body weight. One of the proteins that had this effect was Dnmt3a. This is one of the enzymes that transfers methyl groups to DNA, to switch genes off. The other epigenetic protein that caused increased variability in body weight was called Trim28. Trim28 forms a complex with a number of other epigenetic proteins which together add specific modifications to histones. These modifications down-regulate expression of genes near the modified histones and are known as repressive histone modifications or marks. Regions of the genome that have lots of repressive marks on their histones tend to become methylated on their DNA, so the Trim28 may be important for creating the right environment for DNA methylation.

These experiments suggested that certain epigenetic proteins act as a kind of dampening field. ‘Naked’ DNA is rather prone to being switched on somewhat randomly, and the overall effect is like having a lot of background chatter in our cells. This is called transcriptional noise. The epigenetic proteins act to turn down the volume of this random chat. They do this by covering the histones with modifications that reduce the genes’ expression. It’s likely that different epigenetic proteins are important for suppressing different genes in some tissues rather than in others.

It’s clear that this suppression isn’t total. If it were, then all inbred mice would be identical in every aspect of their phenotype and we know this isn’t the case. There is variation in body weight even in the inbred strains, it’s just that there’s even more variation in the mice with the depressed levels of the epigenetic proteins.

This sophisticated balancing act, in which epigenetic proteins dampen down transcriptional noise but don’t entirely repress gene expression, is a cellular compromise. It leaves cells with enough flexibility of gene expression to be able to respond to new signals – be these hormones or nutrients, pollutants or sunlight – but without the genes being constantly ready to fire up just for the heck of it. Epigenetics allows cells to perform the difficult compromise between becoming (and remaining) different cell types with a variety of functions, and not being so locked into a single pattern of gene expression that they become incapable of responding to changes in their environment.

Something that is becoming increasingly clear is that early development is a key period when this control of transcriptional noise first becomes established. After all, very little of the variation in body weight in the original inbred strains could be attributed to the post-natal environment (just 20–30 per cent). Interest is increasing all the time in the role of a phenomenon called developmental programming, whereby events during foetal development can impact on the whole of adult life, and it is increasingly recognised that epigenetic mechanisms are what underlie a major proportion of this programming.

Such a model is entirely consistent with Emma Whitelaw’s work on the effects of decreased levels of Dnmt3a or Trim28 in her mouse studies. The body weight effects were apparent when the mice were just three weeks old. This model is also consistent with the fact that decreased levels of Dnmt3a resulted in the increased variability in body weight, but decreased levels of the related enzyme Dnmt1 had no effect in Emma Whitelaw’s experiments. Dnmt3a can add methyl groups to totally unmethylated DNA regions, which means it is responsible for establishing the correct DNA methylation patterns in cells. Dnmt1 is the protein that maintains pre-established methylation patterns on DNA. It seems that the most important feature for dampening down gene expression variability (at least as far as body weight is concerned) is establishing the correct DNA methylation patterns in the first place.

Scientists and policy-makers have recognised for many years the importance of good maternal health and nutrition during pregnancy, to increase the chances that babies will be born at a healthy weight and so be more likely to thrive physically. In more recent years, it’s become increasingly clear that if a mother is malnourished during pregnancy, her child may be at increased risk of ill-health, not just during the immediate post-birth infancy, but for decades. We’ve only recently begun to realise that this is at least in part due to molecular epigenetic effects, which result in impaired developmental programming and life-long defects in gene expression and cellular function.

As already highlighted, there are extremely powerful ethical and logistical reasons why humans are a difficult species to use experimentally. Tragically, historical events, terrible at the time, conspire to create human scientific study groups by accident. One of the most famous examples of this is the Dutch Hunger Winter, which was mentioned in the Introduction.

This was a period of terrible hardship and near-starvation during the Nazi fuel and food blockade of the Netherlands in the last winter of the Second World War. Twenty-two thousand people died and the desperate population ate anything they could find, from tulip bulbs to animal blood. The dreadful privations of the population created a remarkable scientific study population. The Dutch survivors were a well-defined group of individuals all of whom suffered just one period of malnutrition, all of them at exactly the same time.

One of the first aspects to be studied was the effect of the famine on the birthweights of children who had been in the womb during the famine. If a mother was well-fed around the time of conception and malnourished only for the last few months of the pregnancy, her baby was likely to be born small. If, on the other hand, the mother suffered malnutrition for the first three months of the pregnancy only (because the baby was conceived towards the end of this terrible episode), but then was well-fed, she was likely to have a baby with normal body weight. The foetus ‘caught up’ in body weight, because foetuses do most of their growing in the last few months of pregnancy.

But here’s the thing – epidemiologists were able to study these groups of babies for decades and what they found was really surprising. The babies who were born small stayed small all their lives, with lower obesity rates than the general population. Even more unexpectedly, the adults whose mothers had been malnourished only early in their pregnancy had higher obesity rates than normal. Recent reports have shown a greater incidence of other health problems as well, including certain aspects of mental health. If mothers suffered severe malnutrition during the early stages of pregnancy, their children were more likely than usual to develop schizophrenia. This has been found not just in the Dutch Hunger Winter cohort but also in the survivors of the monstrous Great Chinese Famine of 1958 to 1961, in which millions starved to death as a result of Mao Tse Tung’s policies.

Even though these individuals had seemed perfectly healthy at birth, something that had happened during their development in the womb affected them for decades afterwards. And it wasn’t just the fact that something had happened that mattered, it was when it happened. Events that take place in the first three months of development, a stage when the foetus is really very small, can affect an individual for the rest of their life.

This is completely consistent with the model of developmental programming, and the epigenetic basis to this. In the early stages of pregnancy, where different cell types are developing, epigenetic proteins are probably vital for stabilising gene expression patterns. But remember that our cells contain thousands of genes, spread over billions of base-pairs, and we have hundreds of epigenetic proteins. Even in normal development there are likely to be slight variations in the expression of some of these proteins, and the precise effects that they have at specific chromosomal regions. A little bit more DNA methylation here, a little bit less there.

The epigenetic machinery reinforces and then maintains particular patterns of modifications, thus creating the levels of gene expression. Consequently, these initial small fluctuations in histone and DNA modifications may eventually become ‘set’ and get transmitted to daughter cells, or be maintained in long-lived cells such as neurons, that can last for decades. Because the epigenome gets ‘stuck’, so too may the patterns of gene expression in certain chromosomal regions. In the short term the consequences of this may be relatively minor. But over decades all these mild abnormalities in gene expression, resulting from a slightly inappropriate set of chromatin modifications, may lead to a gradually increasing functional impairment. Clinically, we don’t recognise this until it passes some invisible threshold and the patient begins to show symptoms.

The epigenetic variation that occurs in developmental programming is at heart a predominantly random process, normally referred to as ‘stochastic’. This stochastic process may account for a significant amount of the variability that develops between the MZ twins who opened this chapter. Random fluctuations in epigenetic modifications during early development lead to non-identical patterns of gene expression. These become epigenetically set and exaggerated over the years, until eventually the genetically identical twins become phenotypically different, sometimes in the most dramatic of ways. Such a random process, caused by individually minor fluctuations in the expression of epigenetic genes during early development also provides a very good model for understanding how genetically identical Avy/a mice can end up with different coat colours. This can be caused by randomly varying levels of DNA methylation of the Avy retrotransposon.

Such stochastic changes in the epigenome are the likely reason why even in a totally inbred mouse strain, kept under completely standardised conditions, there is variation in body weight. But once a big environmental stimulus is introduced in addition to this stochastic variation, the variability can become even more pronounced.

A major metabolic disturbance during early pregnancy, such as the dramatically decreased availability of food during the Dutch Hunger Winter, would significantly alter the epigenetic processes occurring in the foetal cells. The cells would change metabolically, in an attempt to keep the foetus growing as healthily as possible despite the decreased nutrient supply. The cells would change their gene expression to compensate for the poor nutrition, and the patterns of expression would be set for the future because of epigenetic modifications to the genes. It’s probably no surprise that it was the children whose mothers had been malnourished during the very early stages of pregnancy, when developmental programming is at its peak, who went on to be at higher risk of adult obesity. Their cells had become epigenetically programmed to make the most of limited food supply. This programming remained in place even when the environmental condition that had prompted it – famine – was long over.

Recent studies examining DNA methylation patterns in the Dutch Hunger Winter survivors have shown changes at key genes involved in metabolism. Although a correlation like this doesn’t prove cause-and-effect, the data are consistent with under-nutrition during the early developmental period changing the epigenomic profile of key metabolic genes[37].

It’s important to recognise that even in the Dutch Hunger Winter cohort, the effects that we see are not all-or-nothing. Not every individual whose mother had been malnourished early in pregnancy became obese. When scientists studied the population they found an increased likelihood of adult obesity. This is again consistent with a model where random epigenetic variability, the genotypes of the individuals and early environmental events, and the responses of the genes and cells to the environment combine in one great big complicated – and as yet not easily decipherable – equation.

Severe malnutrition is not the only factor that has effects on a foetus that can last a lifetime. Excessive alcohol consumption during pregnancy is a leading preventable cause of birth defects and mental retardation (foetal alcohol syndrome) in the Western world[38]. Emma Whitelaw used the agouti mouse to investigate if alcohol can alter the epigenetic modifications in a mouse model of foetal alcohol syndrome. As we have seen, expression of the Avy gene is epigenetically controlled via DNA methylation of a retrotransposon. Any stimulus that alters DNA methylation of the retrotransposon would change expression of the Avy gene. This would affect the colour of the fur. In this model, fur colour becomes a ‘read-out’ that indicates changes in epigenetic modifications.

Pregnant mice were given free access to alcohol. The coat colour in the pups from the alcohol-drinking mothers was compared with the coat colour of the pups from pregnant mice that didn’t have access to booze. The distribution of coat colours was different between the two groups. So were the levels of DNA methylation of the retrotransposon, as predicted. This showed that the alcohol had led to a change in the epigenetic modifications in the mice. Disruption of epigenetic developmental programming may lead to at least some of the debilitating and lifelong symptoms of foetal alcohol syndrome in children of mothers who over-use alcohol during pregnancy.

Bisphenol A is a compound used in the manufacture of polycarbonate plastics. Feeding bisphenol A to agouti mice results in a change in the distribution of coat colour, suggesting this chemical has effects on developmental programming through epigenetic mechanisms. In 2011 the European Union outlawed bisphenol A in drinking bottles for babies.

Early programming may also be one of the reasons that it’s been very difficult to identify the environmental effects that lead to some chronic human conditions. If we study pairs of MZ twins who are discordant for a specific phenotype, for example multiple sclerosis, it may be well nigh impossible to identify an environmental cause. It may simply be that one of the pair was exceptionally unlucky in the random epigenetic fluctuations that established certain key patterns of gene expression early in life. Scientists are now mapping the distribution of epigenetic changes in concordant and discordant MZ twins for a number of disorders, to try to identify histone or DNA modifications that correlate with the presence or absence of disease.

Children conceived during famines and mice with yellow coats have each clearly taught us remarkable things about early development, and the importance of epigenetics in this process. Oddly enough, these two disparate groups have one other thing to teach us. At the very beginning of the 19th century, Jean-Baptiste Lamarck published his most famous work, Philosophie Zoologique. He hypothesised that acquired characteristics can be transmitted from one generation to the next, and that this drives evolution. As an example, a short-necked giraffe-like animal that elongated its neck by constant stretching would pass on a longer neck to its offspring. This theory has been generally dismissed and in most cases it is simply wrong. But the Dutch Hunger Winter cohort and the yellow mice have shown us that startlingly, the heretical Lamarckian model of inheritance can, just sometimes, be right on the money, as we are about to see.