Discussing Spatial Organization of the Human Genome with Professor Wendy Bickmore
June 24, 2019
Professor Wendy Bickmore is the director of the MRC Human Genetics Unit at the University of Edinburgh. Dr. Bickmore’s research focuses on how the human genome is organized in the nucleus and how this spatial organization contributes to the regulation of gene expression and other biological processes.
Dr. Stefan Dillinger, the host of Active Motif’s Epigenetics Podcast, recently caught up with Professor Bickmore at the EMBO Workshop on Chromatin and Epigenetics and they chatted about how gardening was her gateway to studying biology and some of the new work in her lab on how genomes are organized in human cells.
Listen to the full interview with Professor Wendy Bickmore on Active Motif’s Epigenetics Podcast.
Professor Wendy Bickmore on How Gardening Got Her Interested in Science
Dr. Stefan Dillinger: The first thing I like to ask our podcast guest guest is how did you become interested in biology and in pursuing a career in science?
Professor Wendy Bickmore: I think I became interested in biology because my dad was a really good gardener. In fact, his first job when he left school was as a gardener in one of these big English stately homes. He always had a wonderful garden, so I became interested in biology through plants and, still, a part of me would like to be a plant biologist!
Then, at school, I thought I wanted to become a medical doctor, and I really loved biology. Then, one summer holiday, I was reading a small paperback book by Steven Rose called Chemistry of Life, which was about biochemistry. When I read this book, something went off in my brain and I had this revelation, "Oh, I didn't realize that you could try understanding biology at the level of molecules and chemistry." I just found that utterly fascinating, so I decided to go and study biochemistry.
Stefan: Did you ever work on plants, then?
WB: No, I've never worked on plants, but I have a secret yearning to do so. I do have a lovely garden, though, so I guess I do work on plants.
Studying Gene Organization & Spatial Organization of the Human Genome
Stefan: I want to talk more about the science you're doing now on the spatial organization of the genome. This has been a matter of investigation for a long time, the concept of chromosome territories was proposed Carl Rabl back in the 1880s.
In the early 2000s, you and your team also did some beautiful experiments in this area and found that the positioning of the chromosomes in the nucleus might not be random. How did you approach this, exactly? What was your motivation to go into this area, and how did you then move forward?
WB: The experiments you’re referring to were actually the result of building renovation. I was a human geneticist at the time, doing my postdoc. I was trying to understand the linear organization of the genome, how genes were organized along the DNA sequence. In my first independent lab, the first experiment we actually asked was, "Where are genes within the genome?", because the genome was not sequenced at that time. We devised an experiment to address that and it came back with the very clear answer that genes are non-uniformly distributed along chromosomes, and between human chromosomes as well.
My favorite example is chromosomes 18 and 19, two small human chromosomes. Chromosome 18 is incredibly poorly populated by genes, it's our most gene-poor chromosome, and chromosome 19 is our most gene-rich. And yet, they're two DNA molecules that are essentially the same size, both about 85 million base pairs in length.
This where the building renovation came into the picture. The building my lab was in was being renovated and we got sent to what was called the Cytogenetics Department, where people made metaphase chromosome spreads, and banded chromosomes, and looked down the microscope. I had never looked down the microscope, really, before then.
One of the cytogeneticists said, "Have you ever seen chromosome?" And I said, "Not really. I haven't". So, she showed me, and it was just like, "Oh my gosh, aren't they beautiful!" They've got a pattern to them, bands, and so I became interested in these patterns at that point.
Fluorescence microscopy was just developing at that time, and fluorescence in situ hybridization technology was really starting to develop. So, then I thought, what if we paint chromosome 18, the gene-poor chromosome, with one color, and chromosome 19, the gene-rich one, with another color and hybridize them to cell nuclei and just ask, "How does the cell nucleus deal with these two DNA molecules that are so different in content?".
The resulting data was amazing. I just looked at it and I thought, "Oh my god, there's a pattern here. It's non-random." Chromosome 18 was most often peripheral, around the outside of the nucleus, and chromosome 19 was in the middle. That was the start of us investigating the spatial organization of chromosome territories.
Stefan: You then went on and asked the question, "What do the inactive and active parts of the gene-rich and gene-poor regions on the same chromosome look like?"
WB: Yeah, kind of focused in at smaller-scale resolution and began looking at different parts of individual chromosomes. We also asked, "Is there a non-random organization of whole chromosomes?", and "Is there a non-random organization of parts of chromosomes?"
Stefan: Was this around the time when this whole chromosome "looping" concept emerged?
WB: No, I think looping came a bit later on. I think people only thought of looping at high resolution later on, as Hi-C maps got better at high resolution, for example.
Correlation Between Location of Genes within the Nucleus and Regulation of Expression
Stefan: You later showed that the position of genes within the nucleus doesn't necessarily alter their expression in every instance. So, are there genes at the periphery that are expressed? In general, what aspects of a gene determine how its location within the genome impacts expression?
WB: It doesn’t look like there are any simple rules, so there's still a lot to learn in this space. Some genes care about where they are in the nucleus, other genes don't seem to care. Although it's quite clear that being at the nuclear periphery is more conducive to being repressed, or lowly expressed, you can still get decent amounts of transcription there and we've seen RNA FISH signals at the periphery.
We don't fully know what the rules are, in terms of whether it's particular types of promoters or enhancer-driven genes that give you the difference.
Stefan: Do you have an idea about what kinds of factors influence gene expression activity at different locations within the nucleus? Are there different factors that regulate expression at the periphery and in the middle of the nucleus, for example?
WB: We have started to try and pick apart the relationship between nuclear localization and gene regulation in a couple of different ways. For example, I spent 10 years or so of my independent laboratory time cataloging the way the genome looked in different cells and states of differentiation. Although that was fun and we learned a lot of new things, at some point I became a little frustrated because all we could ever do was make correlations. We looked where the gene is active, we looked where the gene is inactive. Are the cells a stem cell or differentiated cell? We see differences and we try to infer that they might have some functional relationship, but you can never prove that by just looking.
So, we decided we had to find ways to intervene in the system and see what happens when we move a gene to a different location in the nucleus, or what happens to the location of a gene when we activate it. That's when we started working on ways to tether particular genes to the nuclear periphery and look at the consequences on their activity, and, vice-versa, we also developed synthetic transcription factors to try and activate a gene that's at the periphery of the nucleus and then ask what happens.
The results of these experiments were very striking. When we tried to synthetically activate a gene that's at the nuclear periphery, we found that you can activate it, so it's not refractory to activation. But when it is activated, the first thing it does is move off the periphery.
So, transcription can drive organization of the nucleus as well as nuclear organization influencing transcription. It's a bit of a chicken-and-an-egg, who comes first situation. And what happens is also probably different for different genes.
H3K122 Acetylation & Other Histone Globular Domain Modifications
Stefan: Another thing you looked at is enhancers. This was in a Nature Genetics publication in 2016. We know that enhancers are usually identified by histone modifications like H3K27 acetylation and others. But you also looked at the modification of the globular domain of histones as enhancer marks. Could you briefly comment on that?
WB: Yeah. This is an example of why it's good to be at conferences. I was at a conference, I think it was a Cold Spring Harbor conference, and Robert Schneider was giving a talk. Robert works on globular domain modifications, from residue lysine 115 onwards, which I wasn't aware of at all because I was just focused on histone tails. He described antibodies he'd generated that pick up acetylated lysines in the core of the nucleosome particles, and not the tails.
Rob works in both yeast and mammalian systems, and he generated beautiful reagents for studying them. He said that the really interesting thing about these modifications is that they're the only histone modifications that affect transcription, in vitro, on a chromatinized template. In a purified system in vitro, histone tail modifications don't matter because there aren’t any epigenetic readers.
But if you modify the globular domain, it matters to transcription machinery in vitro because it destabilizes the nucleosome. It changes the physical properties of the nucleosome. I thought this was really interesting and was underappreciated in the gene regulation field because we'd been so focused on tails, and the readers of histone tail modifications.
So, I suggested to Rob that we take a look at where you would find modifications like K122 acetylation or K64 acetylation, across the mammalian genome. It turns out that they beautifully mark enhancers. I think the globular histone modifications an even better mark than K27 acetylation.
Stefan: Are there any differences between enhancers that are marked by the K27 trimethylation modification and the globular domain-marked enhancers?
WB: We think there are. There are a lot of similarities, but K122 acetylation looks more specific to enhancers than to the body of genes, whereas the K27 acetylation modification will spread into gene bodies. We found some enhancers in ES cells that appear to have K122 acetylation but not K27 acetylation, and we're still trying to understand why that may be. Because the interesting thing is that both modifications are probably catalyzed by the same enzymes, probably the histone acetyltransferase p300.
I think this is still an emerging area, how structural changes within the nucleosome particle might be involved in enhancer function.
Stefan: Does this also have to do with pioneer factors?
WB: Well it could do, yes, because they obviously probably have a role in destabilizing nucleosomes as well.
Prefer listening over reading? Check out the full interview with Professor Wendy Bickmore on Active Motif’s Epigenetics Podcast.
How Enhancers in Topologically Associating Domains (TADs) Regulate Gene Expression
Stefan: You also recently uploaded a paper to bioRxiv, which was titled Developmentally regulated Shh expression is robust to TAD perturbations. Would you share some of your findings?
WB: Yes, that's right. It's on bioRxiv, which I'm a big fan of. All my papers go up on bioRxiv, I think it's wonderfully democratizing. I find the frustration of publishing in the peer-reviewed literature is that it's so slow! You can spend a year or more trying to get your paper published so people can read it. It's just about communicating results. Put it on bioRxiv, people see it, they talk about it, they tweet about it, they cite it in their papers. I think it just moves the whole of the field forward quickly.
So, back to this paper about TADs, topologically associating domains. This is the result of a long-term collaboration I've had in Edinburgh with a colleague of mine, Bob Hill, who is a developmental biologist interested in developmental gene regulation and, particularly, regulation of sonic hedgehog, this very important embryonic morphogen. It's a classic example of developmental regulation and enhancer-driven regulation.
Sonic hedgehog expression is controlled by at least 13 enhancers that we know about, each of which drives expression at a different time and place of development. These are not super-enhancers, these are individual elements that genetics says, "Okay, this element drives expression in this batch of cells in the brain, or this element drives expression in the limb. This element drives expression in the epithelium of the gut".
Bob is famous for identifying the limb enhancer of sonic hedgehog a million base pairs away from the gene it regulates, so it's a fascinating case of long-range gene regulation.
The intriguing thing about sonic hedgehog and its enhancers is that all the enhancers, although they can be up to a million base pairs away from the gene, they're all in the same TAD as sonic hedgehog within the nucleus. So, it fits everyone's paradigm that TADs are there to constrain the activity of enhancers, to allow enhancers to communicate to the correct target gene and not incorrect, inappropriate genes. Which is a fantastic model. Very appealing, and a beautiful example that might link structure and function.
We decided we'd better test this model by breaking the sonic hedgehog TAD. So what Bob's lab did was to delete the TAD boundaries of sonic hedgehog.
TAD boundaries are usually marked by CTCF sites in particular orientations in the genome, but TADs also have some substructures within them. So, there are sub-TADs and you'll find CTCF sites that might define those, and we decided to take them out one-by-one on the genetic level.
We took the CTCF sites out in both embryonic stem cells and also in mice, as well. So, we made knockout mice homozygously missing CTCF sites.
We showed, by chromosome conformation capture technologies and by imaging, that we had indeed perturbed the TAD structures in the way you might assume by taking these out. So that all looked good.
We took one of them out one at a time. So, for example, when we took out the left boundary, we shift the boundary into another site.
So, for example, if we took the left boundary out, so sonic hedgehog's right up against one end of its TAD, and then took out the next boundary, the remaining left-most boundary then slides in past sonic hedgehog to another CTCF site. Now that puts sonic hedgehog outside of its own TAD. All the enhancers are in the rest of the TAD.
This creates a system where you have a TAD boundary between the gene and the enhancer, allowing us to ask, "What happens to sonic hedgehog regulation in the mouse when you move the gene outside the TAD that contains the enhancers?" The answer, at least by in situ hybridization, was "nothing". We saw completely normal patterns of sonic hedgehog expression in mouse development and, moreover, the mice were viable, fertile, and they looked absolutely fine. They don't have a phenotype that we can detect.
We also don't see any ectopic activation of genes by enhancers when this happens, either. You don't pick up strange patterns. So, what we've done to the TAD so far doesn't seem to perturb enhancer-promoter communication.
It’s probably not that there's no effect whatsoever. Maybe if the mice were stressed, or they're in the wild, there would be some subtle changes in regulation which would make them at a disadvantage. What we conclude is that you don't absolutely need TAD integrity to get correct enhancer-promoter interactions and activation.
Stefan: What do you think the functions of the TADs are then?
WB: Ah, that's a very interesting question. TADs are very conserved, so they likely do have a real role in biology. It's probably either that they have some subtle role in the regulation of gene expression, which would be important in evolutionary time, or that they have a function that’s not really related to transcription, per se.
I always like to think about evolution when trying to understand function because what you don't want to do in evolutionary time is break the cis relationship between a gene and its enhancers. You want to keep them together on the same bit of the genome. You don't want synteny breaks between the gene and the enhancer.
Perhaps TADs are much more about a structural, architectural element of the genome that keeps these blocks of chromatin intact, preventing them from getting broken, for example. I think that could explain what we see in the genome, in that enhancers and their target genes seem to be in the same TAD.
Stefan: Did you look at the individual enhancers or did you just look at the block, in this study?
WB: This new model is just a thought experiment, so far. But when we broke the TAD in the previous experiments we discussed and looked at expression, we tested all the sites of sonic hedgehog expression. So, as far as we know, every enhancer still works fine.
Stefan: Where are you going next with these experiments?
WB: We want to push it further. We took out individual CTCF sites, now we want to combine them. If we just pile them up and take them all away, and completely get rid of the TAD, will that give us a phenotype?
Then we want to really go in more quantitatively and look at transcription itself, of sonic hedgehog, and see if there's a slight delay or a change in the levels. In the lab a favorite tool is to go back to our artificial transcription factors, so we can do synthetic activation of sonic hedgehog from in the TAD, and in the disrupted TADs, and really see what's going on. I think these experiments will really start unpacking what a TAD might and its role in biology.
Stefan: Do you think there might be another way of regulating the TAD, other than CTCF? Are there any other factors involved?
WB: Yeah, not all TADs in our genome have CTCF sites at the edges, but a lot of them do. Polycomb makes very nice TADs by itself, and it looks like the promoters of highly-expressed genes can be TAD boundaries as well.
I'm sure that there'll be other mechanisms too. If we assume that TADs are made by cohesin-mediated loop extrusion, and I think the evidence is good that they are, there are probably other things in the genome that can stall that and stop cohesin moving down, and that would make TAD boundaries. So yeah, I'm sure there'll be other things.
Epigenetics and Cellular Senescence – A Role for the Nuclear Pore
Stefan: You have also studied aging and senescence, and the role of heterochromatin. Some of this work was published in 2015 in Genes and Development. How does chromatin behave during senescence and aging?
WB: I've always really loved the work from Thomas Cremer and Irina Solovei in the rods, the photoreceptors, of nocturnal mammals, where they showed the radial organization of the genome that we normally see, chromosome 18 around the edge, 19 round the middle, gets completely inverted. So, it's still radial, but it's back-to-front.
The heterochromatin goes to the middle and the active genes are around the outside, and I thought, "That's incredibly cool!" But it’s really difficult to study in vivo because it only happens in a few defined cell types postnatally in a living animal. So, tough to manipulate.
I was thinking, "Well, are there any examples I can think of where something as dramatic as that might happen?" At my institute, the MRC Human Genetics Unit happens to be in the same building as the Cancer Research Center, where Juan-Carlos Acosta works on oncogene-induced senescence. Knowing his work, I became aware of senescence-associated heterochromatin foci, which are a bit like rod photoreceptor cells. The heterochromatin regions in these foci are in the middle of the nucleus. Of course, this is an experimentally controllable and tractable system. Actually, we teamed up with JC to investigate this and what might cause it.
Why would heterochromatin be released from the nuclear periphery? In the rod photoreceptor cells, Irina Solovei showed very nicely that it's because you lose the glue that keeps the heterochromatin at the periphery, which is Lamin A, Lamin B receptor. Now, that's not the case in senescence, you still have Lamin A around the periphery, so that didn't seem to be the right model.
Then, we were thinking for a while, "Well, maybe there's some other nuclear membrane protein that changes during senescence and releases heterochromatin." We were pondering that in a lab meeting and then I thought, "No, this is getting awfully complicated. It can't be right." I like simplicity.
We had completely forgotten that there's more to the nuclear periphery that just the lamina. There are holes in it, the nuclear pores, and there usually isn’t heterochromatin right around the nuclear pores. This is presumably because you need to be able to get the RNA out of the nucleus.
The classic electron microscopy images show that there's a gap in the heterochromatin where the pore is, so it looks like the pore physically excludes heterochromatin. So, if you changed pores in some way, could you change heterochromatin distribution?
For example, if the number or the density of pores in the nuclear membrane went up, you would increase the surface area which is excluding heterochromatin and decrease the surface area that attracts it. Would that be enough to reorganize the nucleus? Would that tip in the balance of forces be enough to give you reorganization? It turned out, actually, it was. We found that during senescence the nuclear pore density went way up, and we were able to show that the increase in pores was responsible for the formation of these senescence-associated foci.
Stefan: What causes this increase in the number of nuclear pores?
WB: Oh, that's a good question that is not answered. One thing we know about pores is they're incredibly stable.
Once they're inserted in the membrane, they kind of stay there for the whole cell cycle.
So, reading the literature it became apparent that what quiescent cells do, which are not going to divide, is they downregulate the transcription of nucleoporins, the components of the pores. Because they know they don't need to make more pores, so they stop transcribing the required components. You end up with a fairly stable nuclear pore density.
Senescent cells don't do that. They don't realize they're not going to divide anymore, so they keep transcribing nucleoporins, translating them, and making nuclear pores. I think it's as simple as that. It's a failure of regulation of nuclear pore density.
Stefan: That's very interesting. So that would also lead to unstable nuclei, at some point, right?
WB: It's possible, yeah. This is a really untapped area of biology, actually, the control of nuclear pore number.
Comparing FISH and Other Methods to Look at Chromatin Conformation
Stefan: Switching to another topic now, you once published a review article about the comparison between conformation capture methods, FISH, and immunofluorescence. How do they compare and how you can compare the different results they deliver? How did you do that and what is the conclusion that you drew?
WB: I think the conclusion we made was that for the most part, they tell you the same thing. At the time perhaps was a little shocking to the field, but I think the field's got more used to the idea now. Certainly, at course-grain levels of resolution, like where are compartments and things like that, the methods perform similarly.
But when you get to high resolution, I think you can pick up differences. Imaging is like using a ruler, you literally measure the physical distance between two sites.
Chromosome conformation capture methods like Hi-C are not necessarily the same thing. These methods require crosslinking, and I think the crosslinking efficiency of different bits of the genome could be different. In Hi-C, there's a stage in the protocol where you add detergents to strip off the chromatin before you do the ligation, so you've got a lot of free DNA ends floating around. In cases like this, you can get crosslinks between two loci that are not physically on top of each other, within 50 nm or so. They can probably be 100, maybe 200 nm, away from each other and you'll still get ligation, which would be interpreted as a physical interaction. But, of course, if you imaged that you would see that they were separate.
I think it's at that high level of resolution there can be subtle differences. Which can be really interesting to explore, will tell you something about the nature of the interactions you capture by Hi-C.
I'm a great fan of orthogonal methods to look at the same problem in two different ways because every method has its limitations and problems.
Stefan: Right, and good controls are always needed. Would you say that the methods have improved or changed since you published this article?
WB: Hi-C methods have certainly improved, and they've become much more high resolution. There've been subtle changes in the methodologies. FISH has also improved because now people have moved away from using large probes to oligonucleotide-based probes, so you can pick up really small bits of the genome. There are more fluorescent colors being used now too. Of course, the big thing that's happened to imaging is super-resolution optical imaging, which we've taken great advantage of.
EMBL has been a pioneer of super-resolution microscopy.
Stefan: This conversation has taken a journey through your scientific career so far, but can you maybe give a short summary about what you feel are your most important findings and what you would like to investigate going on from now?
WB: Well, I'm very excited by the senescence research. It's the first time in my career I've seen such dramatic 3-D organizational changes in such a short time window. In my head, I always thought that changes in the spatial organization of the genome within the nucleus would be relatively slow. I was surprised that heterochromatin can move such big distances so quickly.
Our most provocative current findings, which are on bioRxiv but not yet published, is that we don't see evidence that promoters and enhancers actually physically loop and touch each other. We see them often move apart from each other, so they can be hundreds of nanometers away from each other, which I think could fit very well with some of the movements towards the phase separation condensate models of transcription regulation.
I think both that question and the heterochromatin reorganization questions require us to move into live cell imagining. To me, that's the most exciting direction that I'd like to go in.
Check out the full interview with Professor Wendy Bickmore on Active Motif’s Epigenetics Podcast to learn more about her views on spatial organization of the human genome how organization of the genome within the nucleus contributes to the regulation of gene expression and other cellular processes.