Because the pluripotent state of ES cells is critically dependent on the expression of high levels of the master regulators, it’s perhaps unsurprising that the master regulators themselves are controlled by super-enhancers. This creates a positive feedback loop, which is shown in Figure 12.2.
Positive feedback loops are relatively rare in biology, mainly because they can be difficult to get back under control if they start to go wrong. Luckily, the protein-coding genes regulated at super-enhancers are extremely sensitive to small perturbations in binding of master regulators and a number of other factors. This may mean that even a small change in the balance of some of these factors may be enough to interrupt this positive feedback loop, and allow the cells to differentiate rather than remain pluripotent. After all, it doesn’t usually take much of a nudge to make a Slinky fall down the stairs.
Figure 12.2 The positive feedback loop driving high-level sustained expression of master regulator genes.
Super-enhancers have also been reported in tumour cells, where they are associated with critical genes that drive cell proliferation and cancer progression.{245} One of the genes that is regulated by such a super-enhancer is the same one that we encountered earlier in this chapter, which drives Burkitt’s lymphoma. There are also super-enhancers in some normal specialised cells. These bind cell-specific proteins that define cell identity.
Most of the events described so far involve enhancers that are relatively close to the genes that they target, usually within 50,000 base pairs. It’s relatively easy to visualise how this happens, through the long non-coding RNA and the Mediator complex acting to anchor the enzyme that copies DNA into messenger RNA. But there are a lot of situations where the enhancer and the protein-coding gene that it regulates are a really long distance apart on the chromosome, up to several million base pairs away. This is the difference between trying to pass the salt to someone who is on the other side of the table from you at breakfast, and trying to pass it to someone who is at the other end of a soccer field. It’s quite difficult to visualise how this long-range interaction of gene and enhancer can happen. Neither the long non-coding RNA nor the Mediator complex is large enough to span such a huge distance.
In order to understand this process, we have to be a little more sophisticated than usual in the way we think of the genome. Much of the time it’s very helpful to describe DNA in terms of a ladder, or railway lines, because it helps us to visualise the two strands and the way they are held together by base pairs. But the problem with this is it makes us think in very linear terms. We probably also think of DNA as being quite a stiff molecule because subconsciously we are comparing it with solid artefacts from our more familiar physical surroundings.
But we already recognise that DNA isn’t a stiff molecule, because we know that it can be squashed up and compacted really dramatically to fit into the nucleus. So let’s explore that a bit more. If we take the double-stranded nature of DNA as a given (so as not to complicate the picture) we can imagine a section of our genome as being like a very long piece of pasta, maybe the longest piece of tagliatelle ever created. This is marked in a couple of places by food dye, representing the enhancer and the protein-coding gene. Looking at Figure 12.3 we can see two scenarios. When the pasta is uncooked, it’s very inflexible and the enhancer and gene are far apart. But if the pasta is cooked, the tagliatelle becomes flexible. It can fold and bend in all sorts of directions and these can bring the dyed regions representing the enhancer and gene together.
Some parts of our chromosomes are repressed and shut down almost permanently in different cells, to switch off genes that never need to be expressed in that tissue type. Our skin cells, for example, don’t need to express the proteins that are used to carry oxygen around in the blood. These genomic regions are completely inaccessible in the skin cells, curled up tight like an over-wound spring. But there are huge regions that aren’t in this hyper-condensed state and where genes are accessible and can potentially be switched on. In these sections the DNA is like the cooked pasta, like having the longest piece of tagliatelle in the world, filling an entire pot on its own. It bends and swirls in the cooking water, throwing out loops and arcs.
Figure 12.3 Simple schematic to show how folding of a flexible DNA molecule can bring two distant regions, such as an enhancer and a protein-coding gene, into close proximity to each other.
In this way, a protein-coding gene and its distant enhancer may come very close to each other. The long non-coding RNA and the Mediator complex then hold the two loops together and ensure that expression of the gene is driven up. Another complex also has to work with the Mediator complex to carry this out.[35] The additional complex is one that’s also required for separating duplicated chromosomes during cell division, so it’s well equipped for dealing with large-scale movements of DNA. Mutations in some of the genes that encode members of this additional complex cause two developmental disorders, called Roberts Syndrome and Cornelia de Lange syndrome.{246} The precise features of the affected children can be quite variable, probably depending on the exact gene that is mutated, and the precise mutation in that gene. Typically, the children are born small and remain relatively undersized; they have a learning disability and frequently present with limb abnormalities.{247}
The extent of this looping mechanism is quite remarkable, and may not just be restricted to enhancers. It may also be used to bring other regulatory elements close to genes. In a study of three human cell types, analysing just 1 per cent of the human genome, researchers identified over 1,000 of these long-range interactions in each cell line. The interactions were complex, most frequently involving regions that were separated by about 120,000 base pairs. Often the regulatory region looped up to a gene that wasn’t the nearest one to it. In fact, in over 90 per cent of the loops, the nearest gene had been ignored. Think of this as needing to borrow a cup of sugar and visiting someone half a mile away instead of popping in to your next door neighbour.
And if we continue the neighbour theme, the relationships were outrageously promiscuous. Imagine a 1970s partner-swapping party on steroids. Some genes interacted with up to twenty different regulatory regions. Some regulatory regions interacted with up to ten different genes. These probably don’t all occur in the same cell at the same time. But what they show is that there is not a simple A to B relationship between genes and regulatory regions. Instead, there is a complex net of interactions, giving a cell or an organism an extraordinary amount of flexibility in how it regulates its overall tapestry of gene expression.{248} Although there is still plenty to be unravelled about the networks and how they operate, it would appear that while the junk DNA that forms promoters switches on our genomic engines, it’s the junk DNA that forms long non-coding RNAs and enhancers that converts that engine from one powering a Sandero to one that can accelerate the Veyron down the freeway of life.
Remarkable though the looping between individual regulatory regions and genes undoubtedly is, there is an even more dramatic set of long-range interactions that occur in cells. To understand the significance of this, a short social history lesson may be in order. In the early part of the 19th century in Britain, the bulk of all textile work was carried out as a cottage industry. Essentially, individuals worked in their homes on small-scale production. If you had mapped out the locations of textile production in a given region, you’d have a map with lots of individual dots on it, where each working cottage was located. Fast-forward about 50 years and into the Industrial Revolution and the same study would create a very different picture. Instead of a fairly homogenous dotted distribution, like a pointillist painting, you’d find a map with just a few large spots showing the location of big factories.