Junk DNA: A Journey Through the Dark Matter of the Genome

One of the first Bible stories learnt by children raised in the Judaeo-Christian faiths is the creation tale from Genesis. In this story, God creates the earth and the heavens and all that is in them, and finally he creates Adam and Eve. After that, peopling the earth is down to those two and their descendants, with no further divine intervention apart from the obvious exception in the Christian tradition at the start of the New Testament.

The strong grip of the Adam and Eve story perhaps drives or possibly reflects our ingrained acceptance of a simple piece of biology. To create a child you need a man and a woman. It’s not possible, biologically speaking, for a new child to be generated by two men, two women or a woman on her own.

This biological certainty seems so obvious that we virtually never think to question it. But we should, because sometimes the most extraordinary biology lies hidden in the most apparently mundane of assumptions. We should also question it because humans, like all other mammals who bear live young, are in the only class of the animal kingdom where there is never a virgin birth. The mammalian egg needs to be fertilised by a sperm in order to create a new individual. In every other class, there are examples of females who can give birth to live young without ever having mated. It’s not just restricted to the lower classes such as insects. Certain species of fish, amphibians, reptiles, and even some birds can do this. Mammals can’t, suggesting that this restriction on virgin birth arose relatively recently, following the separation of the mammalian and reptilian lines over 300 million years ago.

Researchers extracted mouse eggs that had been fertilised and took out the nucleus. They reconstituted the eggs using nuclei from eggs or sperm and implanted them back into the uterus of receptive female mice. The results are shown in Figure 10.1.

Live mice were only born after the eggs were reconstituted with a nucleus from both an egg and a sperm. If two sperm nuclei were used, or two egg nuclei, the embryos developed for a little while but then couldn’t develop any further. In genetic terms this is really peculiar. In all three experimental systems, the reconstituted egg contained the correct amount of DNA. In terms of DNA sequence there is no actual difference between the DNA from the sperm and the DNA from the egg, particularly because the experiments were designed so that the egg and the sperm each contributed an X chromosome. This created a strange paradox. In all three experimental situations, the DNA sequences involved were exactly the same. But live young only developed if those DNA sequences were donated by a male and a female.¹

We can be very confident that this requirement for both an egg and a sperm isn’t something restricted to mice, because of a human condition called a hydatidiform mole. A woman may appear to be pregnant, gaining weight and frequently suffering from extreme morning sickness. But a scan will show the presence of an enlarged abnormal placenta, full of fluid-filled lumps, and no embryo. This is a hydatidiform mole and it is detected in about 1 in 1,200 pregnancies, although in some Asian populations this figure can reach 1 in 200. The structure will abort spontaneously at about four to five months post-fertilisation, although in societies with good prenatal care clinicians will remove it earlier to prevent the development of potentially dangerous tumours.

Genetic analyses of the abnormal placenta have been very informative. They show that in most cases, hydatidiform moles arise when an egg that for some reason has no nucleus is penetrated by a sperm. The 23 chromosomes in the sperm are copied, to create the normal human chromosome number of 46. In about a fifth of cases the mole is formed when two sperm penetrate one of the unusual nucleus-free eggs simultaneously, again generating the correct number of chromosomes. Just like the mouse experiments, the hydatidiform mole contains the correct number of chromosomes but they derive just from one parent, and this again leads to a severe failure in developmental pathways.

Normally, the fact that cells express two copies of a protein-coding gene gives that cell a kind of insurance policy. Even if one of the copies suffers a mutation or perhaps is inappropriately silenced through abnormal epigenetic modifications, the cell has another copy to fall back on. But if the cell has switched off one of the copies through imprinting, this leaves it more vulnerable to the effects of a random shutdown of the other copy. The fact that for some genes the cell is willing to take this risk tells us that there must be substantial benefits to imprinting that outweigh this disadvantage.

But the same isn’t really true for the male. It doesn’t really matter to him if his offspring draw so much nutrient from their mother that she never reproduces again. In evolutionary terms, all he wants is for his offspring to be as well-nourished and strong as possible, so that they have the greatest chance of reaching sexual maturity and passing on his genes. He is likely to breed with other females, as relatively few mammals mate for life.

The number of imprinted protein-coding genes is fairly small, about 140 in mice.2 They occur in clusters of between two and twelve genes and many of these clusters are quite similar to those in the human genome.³ Perhaps unsurprisingly, the number of imprinted genes is much lower in marsupials where the young are only nourished in utero for a rather short period.⁴

The most critical component in each cluster is a region of junk DNA that controls the expression of the protein-coding genes. This critical component is called the imprinting control element, or ICE. It’s a little like lighting a room with twelve light bulbs. If you want to adjust the level of light in the room, you could use a range of bulbs with different luminosities, and you could have a separate switch for each. But that’s a fairly labour-intensive way of controlling the overall light level. Much better to have all twelve bulbs on one circuit and control them simultaneously with either an on/off switch or a dimmer switch if you want a bit more flexibility.

Imprinting critically depends on the junk DNA and its crosstalk with the epigenetic system. The ICE can be epigenetically modified. Expression of the long non-coding RNA is dependent on whether or not the DNA at its ICE is methylated. If the ICE DNA is methylated, this prevents expression of the long non-coding RNA. If the ICE has escaped methylation, the long non-coding RNA will be expressed. Essentially there is a reciprocal arrangement. If the long non-coding RNA is expressed, the genes in the cluster on the same chromosome will be switched off. If the long non-coding RNA is not expressed, the genes in the cluster on the same chromosome will be switched on. The long non-coding RNAs in the imprinted regions are sometimes extraordinarily long, the biggest being a staggering 1 million bases in length.5

Unfortunately, we are still a bit sketchy in our understanding of the exact mechanisms that the long non-coding RNAs use to repress the expression of the nearby gene cluster. It certainly seems to involve the epigenetic system again, resulting in the deposition of repressive epigenetic modifications on the protein-coding genes. If key epigenetic genes such as the major repressor that we met in Chapter 9 are knocked out in developing embryos, some of the imprinted genes that would normally be switched off are expressed.⁶ It’s not just restricted to the major repressor either, as knockout of other epigenetic genes that establish repressive histone modifications has similar effects.^7,8 This demonstrates the importance of the epigenetic system in carrying out the instructions of the long non-coding RNA. It’s likely this is because the long non-coding RNA attracts these enzymes to the imprinted cluster, thereby targeting the histone modifications to the protein-coding genes.

Epigenetic modifications are also present at the ICE itself. As we would expect, if the ICE DNA is methylated, the histone modifications are ones which are associated with switching genes off. If the ICE is unmethylated, the histone modifications are those which are associated with switching genes on. The pattern of epigenetic modifications on the ICE is completely consistent across the DNA and histone proteins.⁹

In the imprinting process, the critical determinant is whether or not the junk DNA forming the ICE is methylated or not. There have been suggestions that the methylation of the ICEs evolved when silencing of nearby parasitic elements such as those we met in Chapter 4 spread into neighbouring regions. This may have conferred a fitness advantage, and been selected for in subsequent generations.¹⁰ It’s intriguing that in the most primitive mammals, the egg-laying monotremes such as the duck-billed platypus and the echidna, there are uncharacteristically few parasitic elements near the regions where we would expect to find the ICEs in higher mammals.¹¹

This seems full of internal contradictions, but it becomes a little easier to understand if we once again visit the world of the musical. Not Oscar Hammerstein this time but Hal David, who was the lyricist who worked for a long time with Burt Bacharach. They wrote the songs for the 1973 flop film musical Lost Horizon. One of the songs from this became famous and contains a quite useful concept for us: ‘The world is a circle without a beginning and nobody knows where it really ends.’ Developmental processes become much easier to visualise if we think of them as neverending circles rather than in straight lines. The ‘put it on–take it off–put it on’ cycle in the generation of the imprinted ICE is shown in Figure 10.2. This shows how eggs always pass on a maternal pattern of ICE methylation. A similar process allows sperm always to pass on the reciprocal paternal pattern.

Of course, one of the questions this schema raises is how the developing eggs and sperm identify ICE regions and how they ‘know’ which should be methylated and which unmethylated. This is an area of very active research and it may be different for each ICE, and between male and female germ cells. Some of it is frankly still a mystery but there are certain features that have been elucidated. We know that in the maternal germline, i.e. the cells that give rise to eggs, the process is critically dependent on the enzymes that can add DNA methylation to previously unmethylated CpG motifs.*¹² After that, the pattern is actively sustained by an enzyme whose job it is to maintain existing methylation patterns.**¹³ Other proteins are also likely to be involved in establishing the correct methylation patterns, some of which are likely to be selectively expressed in the developing germ cells.

How do the enzymes in the germ cells recognise the ICE regions among all the other genomic DNA? Again there are gaps in our knowledge, although it has been suggested that certain repeated sequences in these special junk DNA regions may play a role.¹⁴ These are quite poorly conserved at the sequence level between species, but may look more similar when we consider their three-dimensional structures. The cell may have a way of recognising them through their shape, rather than their sequence.¹⁵ This is similar to the findings for long non-coding RNAs we saw in Chapter 8.

Although there are obviously plenty of questions that remain about imprinting, we are confident that this is absolutely the reason why both sexes have to contribute to the offspring. In a complex set of breeding experiments using genetically modified mice, researchers showed in 2007 that they could generate viable mice by inserting two egg nuclei into one fertilised egg. The reason they were able to do this was that they artificially altered the pattern of imprinting at two regions in the mouse genome. In one of the egg nuclei, they had created a methylation pattern that looked like the normal paternal pattern, not the maternal one. This fooled the developmental pathways into believing that the genetic material was from a male rather than a female. This demonstrated a particularly strong role for these two imprinted regions in controlling development. It also showed that the only real block to bi-maternal reproduction is the DNA methylation pattern at key genes. It disproved a previous hypothesis that sperm were required because the sperm themselves carried certain necessary accessory factors such as particular proteins or RNA molecules required to kick-start development properly.16

Going back to Figure 10.2 we can see that imprinting patterns may change during development. Imprinted control of gene expression seems to be particularly important during development. In mice, for example, most of the 140 or so imprinted genes are only imprinted in the placenta. In adult tissues both or neither copy of the genes may be expressed. This confirms that control of growth during early development was probably the major reason why imprinting evolved. There seems to be almost a geographical reason for this. In the imprinting clusters, the genes nearest the ICE may remain imprinted in all tissues but the ones further from the control centre may only be imprinted in the placenta. Selected cell types in the brain seem to be particularly likely to retain imprinting, although there is no clear consensus on why this would be favoured evolutionarily in most cases. There have been suggestions that the long non-coding RNA produced from the ICE attracts DNA methylation to the nearest genes but attracts histone modifications to the more distant genes in the cluster.¹⁷ Because histone modifications can be more easily altered than DNA methylation, this may provide a mechanism for releasing more distant genes from imprinting as tissues mature.

So, imprinting occurs, and we have insights into at least some of the mechanisms by which this happens. In light of the theory that imprinting has evolved to balance out the competing evolutionary drives of the mother and foetus (and thus indirectly the father), it’s not surprising that the majority of protein-coding genes controlled by imprinting are ones involved in foetal growth and infant suckling, along with metabolism.¹⁸ It’s also not surprising that when imprinting goes wrong, defects in growth are the commonest symptoms.

Studies of imprinting disorders really took off in the 1980s, when it was first becoming possible to identify genes associated with inherited diseases. The techniques involved finding families with more than one individual affected by a condition, and then analysing these families to narrow down the region on a chromosome that caused the disease. We can do this pretty easily now because we have the sequence of the normal human genome and access to very cheap sequencing technologies. But back in the 1980s, finding a mutation which caused a disease was akin to being asked to find a specific broken light bulb when all you knew was that it was in a house in America. It took years of work by large teams of scientists to identify the mutations underlying a condition.

A number of groups were looking into a disease called Prader-Willi syndrome. Babies born with Prader-Willi syndrome have a low birth weight and poor suckling responses. Their muscle tone doesn’t develop properly until after weaning, so the babies are quite floppy. As the children get older, their appetite becomes completely insatiable and as a consequence they develop early and extreme obesity. The children also suffer from mild mental disability.19

A completely different set of researchers was working on a condition with very different symptoms. This is called Angelman syndrome. Children suffering from this condition have small, under-developed heads, severe learning disabilities and are very late at moving on to solid food. The children are prone to outbursts of laughter for no reason, but thankfully the previous appallingly insensitive description of these patients as ‘happy puppets’ is falling into disuse.²⁰

Imagine building a railway across a continent, where one set of workers starts in the east and builds westward, and the other starts in the west and builds eastwards. At first the workers are in completely different territories, but as time goes on they begin to get closer and closer to each other, and eventually there is a point (assuming all has gone well) where they meet, drive in the last spike, shake hands and have a drink. This is pretty much what happened to the researchers investigating Prader-Willi syndrome and Angelman syndrome. The difference, of course, compared with our railway analogy is that the scientists never expected to meet. They thought they were building independent railways, to completely different cities, and yet they each ended up in exactly the same spot as the other.

As the mapping of the chromosomal regions responsible for Prader-Willi syndrome and for Angelman syndrome gathered pace, it became obvious that the two disorders were located in the same region of the genome. At first, the most obvious assumption was that the disorders were caused in genes that were different from each other, but in very close proximity, like two adjacent shops on a street. But eventually it became clear that the disorders were caused by a defect in exactly the same tightly defined region.

Both conditions had the same underlying genetic cause, a loss of a small region on chromosome 15. The parents of the affected children didn’t suffer from either disorder and when researchers analysed their chromosomes, they discovered these were completely intact. The loss of the key region of chromosome 15 happened during formation of eggs or sperm.*

It was really bizarre that the deletion of small part of a chromosome could cause two conditions that were so different from each other. But the conundrum began to make more sense once researchers demonstrated that it wasn’t just the fact that this small region of chromosome 15 was missing that was important. What mattered was why it was missing. Seventy per cent of children with Prader-Willi syndrome inherited the abnormal chromosome 15 from mutated sperm cells. Seventy per cent of children with Angelman syndrome inherited the abnormal chromosome from mutated egg cells. A little later scientists discovered that 25 per cent of the patients with Prader-Willi syndrome had two perfectly intact chromosomes; nothing was missing. The problem in these patients was that they had inherited both their copies of chromosome 15 from their mother, instead of one from each parent.** In a smaller percentage of Angelman syndrome, the patients had two perfect copies of chromosome 15, but both inherited from their father.

These inheritance patterns make sense only in the context of imprinting, as shown in Figure 10.3. In all the abnormal situations, the cells are lacking an imprinting control region from one parent. This results in abnormal expression levels of the genes that should normally be kept under tight parent-of-origin control, and this leads to pathology including under- or over-growth.

Researchers have been able to narrow down the problems that result in these conditions even further, by analysing the genes that are controlled by the imprinting control regions. In about 10 per cent of cases of Angelman syndrome, the patients have inherited all the appropriate DNA from each parent. The problem they have is that there is a mutation in the DNA from their mother. This is not in the ICE but in a gene controlled by the ICE. This is a protein-coding gene, which is normally expressed only from the maternal chromosome. The version on the paternal chromosome is kept silent by imprinting. If the maternally derived gene can’t create protein because of a mutation, it means the cell can’t produce any of this protein at all, and this leads to pathology.*

The situation in Prader-Willi syndrome is rather more peculiar. A small number of patients have been identified who are lacking just one of the genes that is found in the critical region on chromosome 15. This gene doesn’t code for a protein. Instead, it codes for a batch of non-coding RNAs, all of which have similar functions.^21,22,23 These functions are involved in the control of yet another class of RNA molecules that don’t code for proteins. It seems to be the absence of this one non-protein-coding gene that is critical for the majority of symptoms in Prader-Willi syndrome.

Consider the implications. A region of junk DNA (the ICE) controls the expression of a piece of junk DNA that encodes a long non-coding RNA. This long non-coding RNA in turn critically regulates the expression of a gene that codes for a batch of non-coding RNAs. And the role of these non-coding RNAs is to regulate other RNAs that don’t code for proteins. When we think of it in these terms, it becomes quite difficult to argue that junk DNA has no function.

Prader-Willi and Angelman syndromes are not the only human conditions whereby defects in imprinting lead to abnormalities in growth plus associated problems such as learning disabilities. Another reciprocal pair of conditions are Silver-Russell syndrome,²⁴ an under-growth condition, and Beckwith-Wiedemann syndrome,²⁵ which is characterised by over-growth. The two conditions are caused in some patients by parent-of-origin issues at the same region of chromosome 11. It’s a particularly complex imprinting locus, with lots of genes involved and more than one ICE.

Similar relationships can be found at other chromosomes. Children who inherit both copies of chromosome 14 from their mother are growth-restricted in the pre- and post-natal periods but later become obese.26 But if both copies of chromosome 14 are obtained from the father, an abnormally large placenta develops and the child is born with different problems including defects in the abdominal wall.^27,28

For most of these disorders, there are also rare examples of the condition developing because of epigenetic mistakes. There are small numbers of patients who have inherited the correct DNA from the correct parent. The DNA is not mutated and yet the patients develop an imprinting condition. In these rare cases, the establishment and maintenance of the imprint in the zygote and in early development has usually gone wrong. This can result in an ICE being inappropriately methylated or non-methylated and switched off or on when it shouldn’t be. This once again demonstrates the importance of the cross-talk between junk DNA and the epigenetic machinery.

In 1978 a little girl called Louise Brown was born. If you had seen Louise Brown you would have thought she was a perfectly ordinary baby. No doubt her parents thought she was the most remarkable baby in the world. What parent doesn’t think this about their child? But on this occasion Mr and Mrs Brown could be forgiven for making this claim because they were right. Louise Brown’s birth was front-page news all around the world, because she was the first test tube baby.

Her mother’s egg had been fertilised by her father’s sperm in a dish in a lab and then replaced into her mother’s womb. This procedure was used because Louise’s mother’s fallopian tubes were blocked and she couldn’t conceive naturally. The successful birth of Louise Brown opened a new era in treatment of human infertility. It has been estimated that since that first baby over 5 million children have been born using assisted reproductive technologies.29

There have been claims that assisted reproductive technologies may result in higher levels of imprinting disorders, especially Beckwith-Wiedemann, Silver-Russell and Angelman syndromes. The concerns arise because the embryos are being cultured in the laboratory during the critical period when imprinting gets established. It may seem strange that we don’t know if there really is a problem or not. Surely with 5 million children to analyse it should be quite straightforward to perform the calculations? But the problem is that imprinting disorders are rare, only occurring naturally at rates of one in several or even tens of thousands. When you are analysing events that are so rare, the statistics can be skewed really easily.

Remember Concorde, one of only two supersonic plane models that ever entered commercial service? For decades it was the safest passenger plane in the world, because there had never been a fatal crash. But following the tragic accident at Paris Charles de Gaulle airport in 2000 in which 109 passengers and crew were killed, it became one of the most unsafe planes, statistically speaking. Of course, this was simply because there were relatively few Concorde flights compared with most airliners and the passenger numbers were small (it was a surprisingly bijou plane inside). Therefore, one event could have a major effect on the statistics if these were calculated in a fairly simplistic fashion.

It’s just the same with imprinting disorders. If you would normally expect to see 50 cases for every 5 million children born, how do you interpret it if you detect 55 among the children born via assisted reproductive technologies? Has the medical intervention led to a 10 per cent increase in imprinting disorders, or is this just statistical noise?* We also have to bear in mind that infertility itself may lead to a slight increase in imprinting problems, which is simply unmasked by the assisted reproductive techniques. It’s possible that sperm or eggs from people with low fertility are more likely to carry imprinting defects, but these only become apparent because they are able to have children thanks to medical technology. In the past, they wouldn’t have been able to reproduce so we wouldn’t have seen the effects of the imprinting defect.³⁰ It’s one of those confusing situations in biology where what we think we see is possibly distorted because of what’s out of sight.