The Epigenetics Revolution Nessa Carey

Prev Next

For I, the Lord your God, am a jealous God, punishing the children for the sins of the fathers to the third and fourth generation of those who hate me

Exodus, Chapter 20, Verse 5

The Just So stories published by Rudyard Kipling at the very beginning of the 20th century are an imaginative set of tales about origins. Some of the most famous are those about the phenotypes of animals – How the Leopard Got his Spots, The Beginning of the Armadillos, How the Camel Got his Hump. They are written purely as entertaining fantasies but scientifically they hark back to a century earlier and Lamarck’s theory of evolution through the inheritance of acquired characteristics. Kipling’s stories describe how one animal acquired a physical characteristic – the elephant’s long trunk, for example – and the implication is that all the offspring inherited that characteristic, and hence all elephants now have long trunks.

Kipling was having fun with his stories, whereas Lamarck was trying to develop a scientific theory. Like any good scientist, he tried to collect data relevant to this hypothesis. In one of the most famous examples of this, Lamarck recorded that the sons of blacksmiths (a very physical trade) tended to have larger arm muscles than the sons of weavers (a much less physical occupation). Lamarck interpreted this as the blacksmiths’ sons inheriting the acquired phenotype of large muscles from their fathers.

Our modern interpretation is different. We recognise that a man whose genes tended to endow him with the ability to develop large muscles would be at an advantage in a trade such as blacksmithing. This occupation would attract those who were genetically best suited to it. Our interpretation would also encompass the likelihood that the blacksmith’s sons may have inherited this genetic tendency towards chunky biceps. Finally, we would acknowledge that at the time that Lamarck was writing, children were used routinely as additional members of a family workforce. The children of a blacksmith were more likely than those of a weaver to be performing relatively heavy manual labour from an early age and hence would be likely to develop larger arm muscles as a response to their environment, just as we all do when we pump iron.

It would be a mistake to look back on Lamarck and only mock. We no longer accept most of his ideas scientifically, but we should acknowledge that he was making a genuine attempt to address important questions. Inevitably, and quite rightly, Lamarck has been overshadowed by Charles Darwin, the true colossus of 19th century biology – actually, probably the colossus of biology generally. Darwin’s model of the evolution of species via natural selection has been the single most powerful conceptual framework in biological sciences. Its power became even greater once married to Mendel’s work on inheritance and our molecular understanding of DNA as the raw material of inheritance.

If we wanted to summarise a century and a half of evolutionary theory in one paragraph we might say:

Random variation in genes creates phenotypic variation in individuals. Some individuals will survive better than others in a particular environment, and these individuals are likely to have more offspring. These offspring may inherit the same advantageous genetic variation as their parent, so they too will have increased breeding success. Eventually, over a huge number of generations, separate species will evolve.

The raw material for random variation is mutation of the DNA sequence of the individual; his or her genome. Mutation rates are generally very low, and so it takes a long time for advantageous mutations to develop and to spread through a population. This is especially the case if each mutation only gives an individual a slight advantage over its competitors in a particular environment.

This is where the Lamarckian model of acquired characteristics really falls over, relative to Darwinian models. An acquired change in phenotype would somehow have to ‘feed-back’ onto the DNA script and change it really dramatically, so that the acquired characteristic could be transmitted in the space of just one generation, from parent to child. But there’s very little evidence that this happens, except occasionally in response to chemicals or irradiation which damage DNA (mutagens), causing a change in the actual base-pair sequence. Even these mutagens only affect the genome at a relatively small percentage of base-pairs and in a random pattern, so these still can’t drive inheritance of acquired characteristics in any meaningful way.

The overwhelming body of data argues against Lamarckian inheritance, so there’s very little reason for individual scientists to work on this experimentally. This isn’t surprising. After all, if you are a scientist interested in the Solar System, you could choose to investigate the hypothesis that at least some parts of the Moon are made of cheese. But to do so would mean that you wilfully ignored the large body of evidence already present against this – hardly a rational approach.

There’s also possibly a cultural reason that scientists have shied away from experimental investigations of the inheritance of acquired characteristics. One of the most notorious cases of scientific fraud is that of Paul Kammerer, who worked in Austria in the first half of the 20th century. He claimed that he had demonstrated the inheritance of acquired characteristics in a species called the midwife toad.

Kammerer reported that when he changed the conditions in which the toads bred, they developed ‘useful’ adaptations. These adaptations were structures on their forelimbs called nuptial pads, which were black in colour. Unfortunately, very few of the specimens were retained or stored well, and when a rival scientist examined a specimen he found that India ink had been injected into the pad. Kammerer denied all knowledge of the contamination and killed himself shortly afterwards. This scandal tainted an already controversial field¹.

One of the statements in our potted history of evolutionary theory was the following, ‘An acquired change in phenotype would somehow have to ‘feed-back’ onto the DNA script and change it really dramatically so that the acquired characteristic could be transmitted in the space of just one generation, from parent to child.’

It’s certainly hard to imagine how an environmental influence on the cells of an individual could act at a specific gene to change the base-pair sequence. But it’s all too obvious that epigenetic modifications – be these DNA methylation or alterations to the histone proteins – do indeed occur at specific genes in response to the environmental influences on a cell. The response to hormonal signalling that was mentioned in an earlier chapter was an example of this. Typically, a hormone like oestrogen will bind to a receptor on a cell from, for example, the breast. The oestrogen and the receptor stay together and move into the nucleus of the cell. They bind to specific motifs in DNA – A, C, G and T bases in a particular sequence – which are found at the promoters of certain genes. This helps to switch on the genes. When it binds to these motifs, the oestrogen receptor also attracts various epigenetic enzymes. These alter the histone modifications, removing marks that repress gene expression and putting on marks that tend to switch genes on. In this way, the environment, acting via hormones, can change the epigenetic pattern at specific genes.

These epigenetic modifications don’t change the sequence of a gene, but they do alter how the gene is expressed. This is, after all, the whole basis of developmental programming for later disease. We know that epigenetic modifications can be transmitted from a parent cell to a daughter cell, as this is why there are no teeth in your eyeballs. If a similar mechanism transmitted an environmentally-induced epigenetic modification from an individual to their offspring, we would have a mechanism for a sort of Lamarckian inheritance. An epigenetic (as opposed to genetic) change would be passed down from parent to child.

Heresy and the Dutch Hunger Winter

It’s all very well to think about how this could happen, but really we need to know if acquired characteristics can actually be inherited in this way. Not how does it happen, but the more basic question of does it happen? Remarkably, there appear to be some specific situations where this is indeed taking place. This doesn’t mean that Darwinian/Mendelian models are wrong, it just means that, as always, the world of biology is more complicated than we imagined.

The scientific literature on this contains some confusing terminology. Some early papers refer to epigenetic transmission of an acquired trait but don’t seem to have any evidence of DNA methylation changes, or histone alterations. This isn’t sloppiness on the part of the authors. It’s because of the different ways in which the word epigenetics has been used. In the early papers the phrase ‘epigenetic transmission’ refers to inheritance that cannot be explained by genetics. In these cases, the word epigenetic is being used to describe the phenomenon, not the molecular mechanism. To try to keep everything a little clearer, we’ll use the phrase ‘transgenerational inheritance’ to describe the phenomenon of transmission of an acquired characteristic and only use ‘epigenetics’ to describe molecular events.

Some of the strongest evidence for transgenerational inheritance in humans comes from the survivors of the Dutch Hunger Winter. Because the Netherlands has such excellent medical infrastructure, and high standards of patient data collection and retention, it has been possible for epidemiologists to follow the survivors of the period of famine for many years. Significantly, they were able to monitor not just the people who had been alive in the Dutch Hunger Winter, but also their children and their grandchildren.

This monitoring identified an extraordinary effect. As we have already seen, when pregnant women suffered malnutrition during the first three months of the pregnancy, their babies were born with normal weight, but in adulthood were at higher risk of obesity and other disorders. Bizarrely, when women from this set of babies became mothers themselves, their first born child tended to be heavier than in control groups²^,³. This is shown in Figure 6.1, where the relative sizes of the babies have been exaggerated for clarity, and where we’ve given the women arbitrary Dutch names.

Figure 6.1 The effects of malnutrition across two generations of children and grandchildren of women who were pregnant during the Dutch Hunger Winter. The timing of the malnutrition in pregnancy was critical for the subsequent effects on body weight.

The effects on the birth weight of baby Camilla shown at the bottom left are really odd. When Camilla was developing, her mother Basje was presumably healthy. The only period of malnutrition that Basje had suffered was twenty or more years earlier, when she was going through her own first stages of development in the womb. Yet it seems that this has an effect on her own child, even though Camilla was never exposed to a period of malnutrition during early development.

This seems like a good example of transgenerational (Lamarckian) inheritance, but has it has been caused by an epigenetic mechanism? Did an epigenetic change (altered DNA methylation and/or variations in histone modifications) that had occurred in Basje as a result of malnutrition during her first twelve weeks of development in the womb get passed on via the nucleus of her egg to her own child? Maybe, but we shouldn’t ignore that there are other potential explanations.

For example, there could be an unidentified effect of the early malnutrition which means that when pregnant, Basje will pass more nutrients than normal across the placenta to her foetus. This would still create a transgenerational effect – Camilla’s increased size – but it wouldn’t be caused by Basje passing on an epigenetic modification to Camilla. It would be caused by the conditions in the womb when Camilla was developing and growing (the intrauterine environment).

It’s also important to remember that a human egg is large. It contains a nucleus which is relatively small in volume compared to the surrounding cytoplasm. Imagine a grape inside a satsuma to gain some idea of relative sizes. The cytoplasm carries out a lot of functions when an egg gets fertilised. Perhaps something occurred during early developmental programming in Basje that ultimately resulted in the cytoplasm of her eggs containing something unusual. That might sound unlikely but egg production in female mammals is actually initiated early in their own embryonic development. The earliest stages of zygote development rely to a large extent on the cytoplasm from the egg. An abnormality in the cytoplasm could stimulate an unusual growth pattern in the foetus. This again would result in transgenerational inheritance but not through the direct transmission of an epigenetic modification.

So we can see that there are various mechanisms that could explain the inheritance patterns seen through the maternal line in the Dutch Hunger Winter survivors. It would help us to understand if epigenetics plays a role in acquired inheritance if we could study a less complicated human situation. Ideally, this would be a scenario where we don’t have to worry about the effects of the intra-uterine environment, or the cytoplasm of the egg.

Let’s hear it for fathers. Because men don’t get pregnant, they can’t contribute to the developmental environment of the foetus. Males also don’t contribute much cytoplasm to the zygote. Sperm are very small and are almost all nucleus – they look like little bullets with tails attached. So if we see transgenerational inheritance from father to child, it isn’t likely to be caused by intra-uterine or cytoplasmic effects. Under these circumstances, an epigenetic mechanism would be an attractive candidate for explaining transgenerational inheritance of an acquired characteristic.

Greedy fellows in Sweden

Some data suggesting that male transgenerational inheritance can occur in humans comes from another historical study. There is a geographically isolated region in Northern Sweden called Överkalix. In the late 19th and early 20th centuries there were periods of terrible food shortages (caused by failed harvests, military actions and transport inadequacies), interspersed with periods of great plenty. Scientists have studied the mortality patterns for descendants of people who were alive during these periods. In particular, they analysed food intake during a stage in childhood known as the slow growth period (SGP). All other factors being equal, children grow slowest in the years leading up to puberty. This is a completely normal phenomenon, seen in most populations.

Using historical records, the researchers deduced that if food was scarce during a father’s SGP, his son was at decreased risk of dying through cardiovascular disease (such as stroke, high blood pressure or coronary artery disease). If, on the other hand, a man had access to a surfeit of food during the SGP, his grandsons were at increased risk of dying as a consequence of diabetic illnesses⁴. Just like Camilla in the Dutch Hunger Winter example, the sons and grandsons had an altered phenotype (a change in the risk of death through cardiovascular disease or diabetes) in response to an environmental challenge they themselves had never experienced.

These data can’t be a result of the intra-uterine environment nor of cytoplasmic effects, for the reasons outlined earlier. Therefore, it seems reasonable to hypothesise that the transgenerational consequences of food availability in the grandparental generation were mediated via epigenetics. These data are particularly striking when you consider that the original nutritional effect happened when the boys were pre-pubescent and so had not even begun to produce sperm. Even so, they were able to pass an effect on to their sons and grandsons.

However, there are some caveats around this work on transgenerational inheritance through the male line. In particular, there are risks involved in relying on old death records, and extrapolating backwards through historical data. Additionally, some of the effects that were observed were not terribly large. This is frequently a problem when working with human populations, along with all the other issues we have already discussed, such as our genetic variability and the impossibility of controlling environment in any major way. There is always the risk that we draw inappropriate conclusions from our data, rather as we believe Lamarck did with his studies on the families of blacksmiths.

The heretical mouse

Is there an alternative way of investigating transgenerational inheritance? If this phenomenon also occurs in other species, it would give us a lot more confidence that these effects are real. This is because experiments in model systems can be designed to test specific hypotheses, rather than just using the datasets that nature (or history) provides.

This is where we come back to the agouti mouse. Emma Whitelaw’s work showed that the variable coat colour in the agouti mouse was due to an epigenetic mechanism, specifically DNA methylation of a retrotransposon in the agouti gene. Mice of different colour all had the same DNA sequence, but a different degree of epigenetic modification at the retrotransposon.

Professor Whitelaw decided to investigate if the coat colour could be inherited. If it could, it would show that it’s not only DNA that gets transmitted from parent to offspring, but also epigenetic modifications to the genome. This would provide a potential mechanism for the transgenerational inheritance of acquired characteristics.

When Emma Whitelaw allowed female agouti mice to breed, she found the effect that is shown in Figure 6.2. For convenience, the picture only shows the offspring who inherited the A^vy retrotransposon from their mother, as this is the effect we are interested in.

If the mother had an unmethylated A^vy gene, and hence had yellow fur, all her offspring also had either yellow fur, or slightly mottled fur. She never had offspring who developed the very dark fur associated with the methylation of the retrotransposon.

By contrast, if the mother’s A^vy gene was heavily methylated, resulting in her having dark fur, some of her offspring also had dark fur. If both grandmother and mother had dark fur, then the effect was even more pronounced. About a third of the final offspring had dark fur, compared with the one in five shown in Figure 6.2.

Figure 6.2 The coat colour of genetically identical female mice influences the coat colour of their offspring. Yellow female mice, in whom the agouti gene is expressed continuously, due to low levels of DNA methylation of the regulatory retrotransposon, never give birth to dark pups. The epigenetically – rather than genetically – determined characteristics of the mother influence her offspring.

Because Emma Whitelaw was working on inbred mice, she was able to perform this experiment multiple times and generate hundreds of genetically identical offspring. This was important, as the more data points we have in an experiment, the more we can rely on the findings. Statistical tests showed that the phenotypic differences between the genetically identical groups were highly significant. In other words, it was very unlikely that the effects occurred by chance⁵.

The results from these experiments showed that an epigenetically-mediated effect (the DNA methylation-dependent coat pattern) in an animal was transmitted to its offspring. But did the mice actually inherit directly an epigenetic modification from their mother?

There was a possibility that the effects seen were not directly caused by inheritance of the epigenetic modification at the A^vy retrotransposon, but through some other mechanism. When the agouti gene is switched on too much, it doesn’t just cause yellow fur. Agouti also mis-regulates the expression of other genes, which ultimately results in the yellow mice being fat and diabetic. So it’s likely that the intra-uterine environment would be different between yellow and dark pregnant females, with different nutrient availability for their embryos. The nutrient availability could itself change how particular epigenetic marks get deposited at the A^vy retrotransposon in the offspring. This would look like epigenetic inheritance, but actually the pups wouldn’t have directly inherited the DNA methylation pattern from their mother. Instead, they’d just have gone through a similar developmental programming process in response to nutrient availability in the uterus.

Indeed, at the time of Emma Whitelaw’s work, scientists already knew that diet could influence coat colour in agouti mice. When pregnant agouti mice are fed a diet rich in the chemicals that can supply methyl groups to the cells (methyl donors), the ratios of the differently coloured pups changes⁶. This is presumably because the cells are able to use more methyl groups, and deposit more methylation on their DNA, hence shutting down the abnormal expression of agouti. This meant that the Whitelaw group had to be really careful to control for the effect of intrauterine nutrition in their experiments.

In one of those experiments that simply aren’t possible in humans, they transferred fertilised eggs obtained from yellow mothers and implanted them into dark females, and vice versa. In every case, the distribution of coat patterns in the offspring was the same as was to be expected from the egg donor, i.e. the biological mother, rather than the surrogate. This showed unequivocally that it wasn’t the intra-uterine environment that controlled the coat patterning. By using complex breeding schemes, they also demonstrated that the inheritance of the coat pattern was not due to the cytoplasm in the egg. Taken together, the most straightforward interpretation of these data is that epigenetic inheritance has taken place. In other words, an epigenetic modification (probably DNA methylation) was transferred along with the genetic code.

This transfer of the phenotype from one generation to the next wasn’t perfect – not all the offspring looked exactly the same as their mother. This implies that the DNA methylation that controls the expression of the agouti phenotype wasn’t entirely stable down the generations. This is quite analogous to the effects we see in suspected cases of human transgenerational inheritance, such as the Dutch Hunger Winter. If we look at a large enough number of people in our study group we can detect differences in birth weight between various groups, but we can’t make absolute predictions about a single individual.

There is also an unusual gender-specific phenomenon in the agouti strain. Although coat pattern showed a clear transgenerational effect when it was passed on from mother to pup, no such effect was seen when a male mouse passed on the A^vy retrotransposon to his offspring. It didn’t matter if a male mouse was yellow, lightly mottled or dark. When he fathered a litter, there were likely to be all the different patterns of colour in his offspring.

But there are other examples of epigenetic inheritance transmitted from both males and females. The kinked tail phenotype in mice, which is caused by variable methylation of a retrotransposon in the Axin^Fu (Axin fused) gene, can be transmitted by either the mother or the father⁷. This makes it unlikely that transgenerational inheritance of this characteristic is due to intra-uterine or cytoplasmic influences, because fathers don’t really contribute much to these. It’s far more likely that there is the transmission of an epigenetic modification at the Axin^Fu gene from either parent to offspring.

These model systems have been really useful in demonstrating that transgenerational inheritance of a non-genetic phenotype does actually occur, and that this takes place via epigenetic modifications. This is truly revolutionary. It confirms that for some very specific situations Lamarckian inheritance is taking place, and we have a handle on the molecular mechanism behind it. But the agouti and kinked tail phenotypes in mice both rely on the presence of specific retrotransposons in the genome. Are these special cases, or is there a more general effect in play? Once again, we return to something that has a bit more immediate relevance for us all. Food.

The epigenetics of obesity

As we all know, an obesity epidemic is developing. It’s spreading worldwide, although it’s advancing at a particularly fast rate in the more industrialised societies. The frankly terrifying graph in Figure 6.3 displays the UK figures for 2007⁸, showing that about two out of every three adults is overweight (body mass index of 25 or over) or obese (body mass index of 30 or over). The situation is even worse in the USA. Obesity is associated with a wide range of health problems including cardiovascular disease and type 2 diabetes. Obese individuals over the age of 40 will die, on average, 6 to 7 years earlier than non-obese people⁹.

Figure 6.3 The percentage of the UK population that was overweight or obese in 2007.

The data from the Dutch Hunger Winter and other famines support the idea that poor nutrition during pregnancy has effects on offspring, and that these consequences can be transmitted to subsequent generations as well. In other words, poor nutrition can have epigenetic effects on later generations. The data from the Överkalix cohort, although more difficult to interpret, suggested that excess consumption at key points in a boy’s life can have adverse consequences for later generations. Is it possible that the obesity epidemic in the human population will have knock-on effects for children and grandchildren? As we don’t really want to wait 40 years to work this out, scientists are again turning to animal models to try to gain some useful insights.

The first animal data suggested that nutrition might not have much effect transgenerationally. The change in coat pattern of pups when pregnant agouti mice were given diets high in methyl donors didn’t transmit to the next generation¹⁰. But perhaps this is too specialised a model. In 2010, two papers were published that should at least give us pause for thought. They were published in two of the best journals in the world – Nature and Cell. In both cases, the researchers overfed male animals and then monitored the effects on their offspring. By restricting their experiments to males, they didn’t need to worry about the intra-uterine and cytoplasmic complications that cause such (metaphorical) headaches if studying females.

One of the studies used a breed of rat called Sprague-Dawley. This is an albino rat, with a chilled-out temperament that makes it easy to keep and handle. In the experiments male Sprague-Dawleys were given a high-fat diet, and allowed to mate with females who had been fed an ordinary diet. The over-fed males were overweight (hardly a surprise), had a high percentage of fat to muscle and had many of the symptoms found in type 2 diabetes in humans. Offspring were normal weight but they too had the diabetes-type abnormalities¹¹. Many of the genes that control metabolism and how mammals burn fuel were mis-regulated in these offspring. For reasons that aren’t understood, it was particularly the daughters that showed this effect.

A completely independent group studied the effects of diet in an inbred mouse strain. Male mice were fed a diet that was abnormally low in protein. The diet had an increased percentage of sugar to make up for this. The males were mated to females on a normal diet. The researchers examined the expression of genes in the liver (the body’s major organ when it comes to metabolism) in three-week-old pups from these matings. Analysing large numbers of mouse pups, they found that the regulation of many of the genes involved in metabolism was abnormal in the offspring of the males that had been fed the modified diet¹². They also found changes in the epigenetic modifications in the livers of these pups.

So, both these studies show us that, at least in rodents, a father’s diet can directly influence the epigenetic modifications, gene expression and health of his offspring. And not because of environment – this isn’t like the human example of a child getting fat because their Dad only ever feeds them super-sized portions of burgers and chips. It’s a direct effect and it occurred so frequently in the rats and mice that it can’t have been due to diet-induced mutations, they just don’t happen at that sort of rate. So the most likely explanation is that diet induces epigenetic effects that can be transmitted from father to child. Although the data are quite preliminary, the results from the mouse study in particular support this.

If you look at all the data in its entirety – from humans to rodents, from famine to feast – a quite worrying pattern emerges. Maybe the old saw of ‘we are what we eat’ doesn’t go far enough. Maybe we’re also what our parents ate and what their parents ate before them.

This might make us wonder if there is any point following advice on healthy living. If we are all victims of epigenetic determinism, this would suggest that our dice have already been rolled, and we are just at the mercy of our ancestors’ methylation patterns. But this is far too simplistic a model. Overwhelming amounts of data show that the health advice issued by government agencies and charities – eating a healthy diet rich in fruit and vegetables, getting off the sofa, not smoking – is completely sound. We are complex organisms, and our health and life expectancy are influenced by our genome, our epigenome and our environment. But remember that even in the inbred agouti mice, kept under standardised conditions, researchers couldn’t predict exactly how yellow or how fat an individual mouse in a newborn litter would become. Why not do everything that we can to improve our chances of a healthy and long life? And if we are planning to have children, don’t we want to do whatever we can to nudge them that bit closer to good health?

There will always be things we can’t control, of course. One of the best-documented examples of an environmental factor that has epigenetic consequences, lasting at least four generations, is an environmental toxin. Vinclozolin is a fungicide, which tends to be used particularly frequently in the wine industry. If it gets into mammals it is converted into a compound that binds to the androgen receptor. This is the receptor that binds testosterone, the male hormone that is vital for sexual development, sperm production and a host of other effects in males. When vinclozolin binds to the androgen receptor, it prevents testosterone from transmitting its usual signals to the cells, and so blocks the normal effects of the hormone.

If vinclozolin is given to pregnant rats at the time when the testes are developing in the embryos, the male offspring are born with testicular defects and have reduced fertility. The same effect is found for the next three generations¹³. About 90 per cent of the male rats are affected, which is far too high a percentage to be caused by classic DNA mutation. Even the highest known rates of mutation, at particularly sensitive regions of the genome, are at least ten-fold less frequent than this. In these rat experiments, only one generation was exposed to vinclozolin, yet the effect lasted for at least four generations, so this is another example of Lamarckian inheritance. Given the male transmission pattern, it is likely this is another example of an epigenetic inheritance mechanism. A follow-on publication from the same research group has identified regions of the genome where vinclozolin treatment leads to unusual DNA methylation patterns¹⁴.

The rats in the studies described above were treated with high doses of vinclozolin. These were much larger than humans are believed to encounter in the environment. Nonetheless, effects such as these are one of the reasons why some authorities are beginning to investigate if artificial hormones and hormone disrupters in the environment (from excretion of chemicals present in the contraceptive pill, to certain pesticides) have the potential to cause subtle, but potentially transgenerational effects in the human population.

Prev Next