The third reason for scepticism is possibly the most important and it relates to DNA methylation itself. DNA methylation at the target genes in the brain is established very early, possibly pre-natally but certainly within one day of birth, in rodents. What this means is that the baby mice or baby rats in the experiments all started life with a certain baseline pattern of DNA methylation at their cortisol receptor gene in the hippocampus. The DNA methylation levels at this promoter alter in the first week of life, depending on the amount of licking and grooming the rats receive. As we saw, the DNA methylation levels are higher in the neglected mice than in the loved ones. But that’s not because the DNA methylation has gone up in the neglected mice. It’s because DNA methylation has gone down in the ones that were licked and groomed the most. The same is also true at the arginine vasopressin gene in the baby mice removed from their mothers. It’s also true for the corticotrophin-releasing hormone gene in the adult mice that were susceptible to social defeat.
So, in every case, what the scientists observed was decreased DNA methylation in response to a stimulus. And that’s where, molecularly, the problem lies, because no-one knows how this happens. In Chapter 4 we saw how copying of methylated DNA results in one strand that contains methyl groups and one that doesn’t. The DNMT1 enzyme moves along the newly synthesised strand and adds methyl groups to restore the methylation pattern, using the original strand as a template. We could speculate that in our experimental animals, there was less DNMT1 enzyme present and so the methylation levels at the gene dropped. This is referred to as passive DNA demethylation.
The problem is that this can’t work in neurons. Neurons are terminally differentiated – they are right at the bottom of Waddington’s landscape, and cannot divide. Because they don’t divide, neurons don’t copy their DNA. There’s no reason for them to do so. As a result, they can’t lose their DNA methylation by the method described in Chapter 4.
One possibility is that maybe neurons simply remove the methyl group from DNA. After all, histone deacetylases remove acetyl groups from histones. But the methyl group on DNA is different. In chemical terms, histone acetylation is a bit like adding a small Lego brick onto a larger Lego brick. It’s pretty easy to take the two bricks apart again. DNA methylation isn’t like that. It’s more like having two Lego bricks and using superglue to stick them together.
The chemical bond between a methyl group and the cytosine in DNA is so strong that for many years it was considered completely irreversible. In 2000, a group from the Max Planck Institute in Berlin demonstrated that this couldn’t be the case. They showed that in mammals the paternal genome undergoes extensive DNA demethylation, during very early development. We came across this in Chapters 7 and 8. What we glossed over at the time was that this demethylation happens before the zygote starts to divide. In other words, the DNA methylation was removed without any DNA replication[222]. This is referred to as active DNA demethylation.
This means there is a precedent for removing DNA methylation in non-dividing cells. Perhaps there’s a similar mechanism in neurons. There’s still a lot of debate about how DNA methylation is actively removed, even in the well-established events in early development. There’s even less consensus about how it takes place in neurons. One of the reasons this has been so hard to investigate is that active DNA demethylation may involve a lot of different proteins, carrying out a number of steps one after another. This makes it very difficult to recreate the process in a lab, which is the gold standard for these kinds of investigations.
As we’ve seen repeatedly, scientific research often throws up some very unexpected findings and so it happened here. While many people in epigenetics were looking for an enzyme that removed DNA methylation, one group discovered enzymes that added something extra to methylated DNA. This is shown in Figure 12.3. Very surprisingly, this has turned out to have many of the same consequences as demethylating the nucleic acid.
Figure 12.3 Conversion of 5-methylcytosine to 5-hydroxymethylcytosine. C: carbon; H: hydrogen; N: nitrogen; O: oxygen. For simplicity, some carbon atoms have not been explicitly shown, but are present where there is a junction of two lines.
A small molecule called hydroxyl, consisting of one oxygen atom and one hydrogen atom, is added to the methyl group, to create 5-hydroxymethylcytosine. This reaction is carried out by enzymes called TET1, TET2 or TET3[223].
This is highly relevant to the question of DNA demethylation, because it’s the effects of DNA methylation that make this change important. Methylation of cytosine affects gene expression because methylated cytosine binds certain proteins, such as MeCP2. MeCP2 acts with other proteins to repress gene expression and to recruit other repressive modifications like histone deacetylation. When an enzyme such as TET1 adds the hydroxyl group to the methylcytosine to form the 5-hydroxymethylcytosine molecule, it changes the shape of the epigenetic modification. If a methylated cytosine is like a grape on a tennis ball, the 5-hydroxymethylcytosine is like a bean stuck to a grape stuck to a tennis ball. Because of this change in shape, the MeCP2 protein can’t bind to the modified DNA any more. The cell therefore ‘reads’ 5-hydroxymethylcytosine in the same way as it reads unmethylated DNA.
Many of the techniques used until very recently looked for the presence of DNA methylation. They often couldn’t distinguish between unmethylated DNA and 5-hydroxymethylated DNA. This means that many of the papers which refer to decreased DNA methylation may actually have been detecting increased 5-hydroxymethylation without knowing it. It’s currently unproven, but it may be that instead of actually demethylating DNA, as reported in some of the behavioural studies, neurons really convert 5-methylcytosine to 5-hydroxymethylcytosine. The techniques for studying 5-hydroxymethylcytosine are still under development but we do know that neurons contain higher levels of this chemical than any other cell type[224].
Despite these controversies, research is continuing into the importance of epigenetic modifications in brain function. One area that is attracting a lot of attention is the field of memory. Memory is an incredibly complex phenomenon. Both the hippocampus and a region of the brain called the cortex are involved in memory, but in different ways. The hippocampus is mainly involved in consolidating memories, as our brains decide what we are going to remember. The hippocampus is fairly plastic in the way that it operates, and this seems to be associated with transient changes in DNA methylation, again through fairly uncharacterised mechanisms. The cortex is used for longer-term storage of memories. When memories are stored in the cortex, there are prolonged changes in DNA methylation.
The cortex is like a hard drive on a computer with gigabytes of storage. The hippocampus is more like the RAM (random access memory) chip, where data are temporarily processed before being deleted, or transferred to the hard drive for permanent storage. Our brain separates out different functions to selected cell populations in different anatomical regions. This is why memory loss is rarely all-encompassing. Depending on the clinical condition, for example, either one of short-term or long-term memory may be relatively lost or remain relatively intact. It makes a lot of sense for these different functions to be separated in our brains. Just try to imagine life if we remembered everything that ever happened – the phone number that we dialled only once, every word a dull stranger said to us on a train, or the canteen menu from a wet Wednesday three years ago.