An ideal topic for PhD students and Post docs.
Dr. Alexander Meissner is an Associate Professor of Stem Cell and Regenerative Biology at Harvard University. He is also the Principal Faculty at the Harvard Stem Cell Institute, and Robertson Investigator at the New York Stem Cell Foundation.
Dr. Meissner is the Senior Associate Member at the Broad Institute, which is a collaborative, multidisciplinary initiative between Harvard University and the Massachusetts Institute of Technology (MIT).
AM: Thank you for the introduction and it is my pleasure to be here today and tell you a little bit about our recent work on studying epigenetic regulation, epigenetic dynamics during both stem cell differentiation, as well as normal development. So just as a quick background, the reason we are so interested in this problem is the very fundamental question of how do you actually go from a single cell, such as a zygote, to build a very complex multicellular organism, such as a human, without actually undergoing any changes in the coding genome sequence? So how can you create this diversity and maintain it over a lifetime without actually changing the genetic blueprint? Now, because it is often hard to study in vivo, what we usually do is take advantage of stem cell models, in particular embryonic stem cells. And these are pluripotent stem cells that can give rise to all these diverse stem cell types and, as such, provide a very elegant model to study the dynamics that occur during the regulation here.
More recently, through work from Shinya Yamanaka and others, we also have now the ability to actually return any one of these somatic cell types back to a pluripotent state, again giving us another dimension of how we can look at these regulatory dynamics that maintain and specify cell types and solid entities. Of course, also, more recently, work has shown that you can also not only just go back and forth between the pluripotent and differentiated state, but you can also reprogram and transdifferentiate between all kinds of somatic cell states. But, in all cases, what we're looking at is, again, the same genome sequences maintained through all of these cell types and, as such, this is generally referred to as epigenetic reprogramming, because you're not changing the genome sequence.
Now, to just give a little bit of background on what we mean when we talk about epigenetics, this is a classic cartoon on the right now that is used to introduce this topic. So what you can see is that DNA can be modified directly on the cytosine basis, and this is typically referred to as DNA methylation and will be a major part of the talk today. But then, as you know, the DNA is also wrapped around histone proteins; these histones are themselves subject to dozens of different modifications, and all of these are impacting genome regulation. Today, we'll mostly focus on these two layers, however, it should be noted that the DNA can be higher chromatin structures and further packaged into condensed chromatin; and even compartmentalised within the nucleus. All of these, again, contribute, at some level, to the genome regulation.
Just to go back to the first part, and talking mostly about DNA methylation, so what we're really talking about is this very tiny, small chemical modification on this methyl group that's added to the cytosine base. It's noted that the cytosine base or this methyl-cytosine or cytosine does not change the coding function of cytosine. What has been studied over the last decade, as we know from studies in the nineties that DNA methylation is essential for normal development. All the catalytically active enzymes that are present in mammalian cells have been identified, and those are DNMT1 which is generally referred to as the maintenance enzyme that targets human methylated DNA, and copies methylation off the DNA replication. In addition to that, there's two so-called DNA methyltransferase and this is DNMT3A and 3B that can target unmethylated and DNMT methylated DNA in the genome.
So what we know from these studies in the nineties is that DNA methylation is clearly essential for development, because any one of these enzymes when knocked out causes a lethal phenotype. So if you look at the right, basically this is a wild-type embryo and if you delete the maintenance enzyme DNMT1 it causes an early lethal phenotype where the mice die around day eight to ten of embryonic development. If you knock out DNMT3B, which is shown on the bottom here, the embryo survives slightly longer and would die around embryonic day 15. Then if you knock out DNMT3A, the mice actually get born but are usually very runted and die within weeks after birth. However, if you knock out both DNMT3A and 3B combined, the phenotype resembles very closely to the DNMT1 knock out with a complete fatality around day eight of embryonic development.
Again, what this shows is you need DNA-methylation for normal development. What, obviously, could not answer is where in the genome is the methylation actually required in order to sustain normal differentiation and development? So when we look at the targets of methylation, again, I mentioned the cytosine base, but, in fact, in the mammalian cells at least, and there's some are different from plants and other organisms, the majority of methylation is found in the CpG dinucleotide. You can see the symmetry here provides a very elegant mechanism that allows you to replicate methylation patterns on the newly-synthesised strands, or the old strand will retain the methylation information, and this can be then copied to the newly-synthesised strand from both copies.
If you look at the mouse and human genome, the mouse genome contains both 21 million CpG dinucleotides and the human genome about 28 million CpGs. So really what we are mostly interested in right now is understanding where are the modified CpG dinucleotides, and then which ones of those are participating in any regulatory events in the genome? So we have now technologies to very efficiently measure genome-wide DNA methylation patterns, and this just shows you a very zoomed out view of a large chunk of a chromosome where the methylation measurements are shown with the blue dots. Here, where every single blue dot has an individual CpG and its methylation value is shown on the Y-axis. What you can immediately see is that most of the genome is actually very highly methylated, and this will be true for pretty much any somatic cells that you will look at. But you can also see there are clear exceptions, and these are shown down here where you can see the dips in methylation, and you can see they align up very nicely with the so-called CpG islands, and they're referred to as CpG islands, because they tend to have a much higher CpG density than the rest of the genome.
In fact, the methylation that we observed is very CpG-density dependent with CpG dense regions generally being less methylated or lowly methylated, and the rest of the bulk genome which is depleted of CpG dinucleotides tends to be largely methylated. Again, what we're then interested in is not the static picture of what a normal cell type looks like, but, more importantly, when and where, actually, the CpG is changing the methylation value and state across different cell types? In order to address this question very systematically we took advantage of a very large data set where we had over 40 whole genome bisulfite sequencing data sets, covering at least 30 distinct cell and tissue types. So these included different shaded stem cells, but also many in vivo primary cell types such as brain regions, liver, hematopoietic cell types and so forth.
So what you can see in this pie chart is that actually only a small fraction of all the CpGs that we can measure show dynamic regulation or changes, where most of the genome actually contains static CpGs that are not changing across any one of these cell types that we've investigated here. Now, I'm just going to show you an example of what we actually look at here when we define these dynamic regions, and the dynamic CpGs. So what is shown here is actually in contrast to the previous figure, it's not showing you the methylation levels, that's actually showing you the variation in terms of methylation, so that it's measured variation across all of these 30 cell types. For example, if you have a region here that shows Region 1, that shows no variation and you can see it aligns up with the CpG island, and this is in line with the general idea that CpG islands tend to be unmethylated across all somatic cell types.
In contrast to that, for example, if you go to this particular region here it shows very high variation in terms of DNA methylation, and the gene that it shows here is actually OCT4 or POU5F1, which is a gene that's highly expressed in embryonic stem cells, but silent in most somatic cell types. To show you the underlying data that gives rise to this variation, I'm just going to show you a couple of these 30 cell types, so this is just four different cell types. So now it's showing on top the DNA methylation variation, and below that it's actually showing you the methylation measurement across these four cell types. What you can see is for this particular region of the OCT4 promoter, you can see the promoter region is unmethylated in the human embryonic stem cells. As you differentiate, within a few days you start seeing trickling in methylation here, and in somatic cell types this region is highly methylated. So because of this difference here, you have higher variation in terms of DNA methylation for this particular region.
Another thing that you can actually see here very nicely is that it's generally not a single CpG that's flipping back and forth between the unmethylated and the methylated state, but rather it's the entire region. For example, in this case, the promoter is in a slightly upstream region, so it's actually concordantly changing the methylation for many of the CpG dinucleotides. So, therefore, we can actually take advantage of this and we can summarise not just individual CpGs such as in the pie chart before, but we can actually define regions of dynamic regulation. When we apply this genome wide, we actually identify about 700,000 regions that show variation across the 30 cell types that we investigated here. This has the advantage that we can now actually take a closer look at these 700,000 regions and try to define both in the genome context where they are, what are their characteristic features, and also what do we actually find enriched within these particular regions?
So to start with what these regions are, if we look at the percentage of dynamic regions and look at classic promoter or transcription start site features, again, broken down here to promoters, who are then breaking the promoters down into high CpG, intermediate CpG, or low CpG promoters, or CpG islands, CpG island shores which are neighbouring the CpG islands, or the 5 per exons, you can see that they actually all contribute very little to the number of dynamic regions. In contrast, when you look at other features, in particular I want to point out the enhancers or DNA's hypersensitive sites, these are typically distal sites in the genome that are important for genome regulation. These tend to contribute mostly to where the dynamics in terms of DNA methylation are found.
We can then take a closer look at the features here, and I just want to point out three characteristics that we observe for these dynamically regulated regions. Number one; that is quite striking as they're very small in size, so over 70 per cent of the region is actually less than 1 kb in size. They also tend to be very far away from transcription start sites, so you can see a large percentage is actually more than 100 kb away from the nearest transcription start site. You can also see that the majority of them are actually fairly CpG-poor, and so just as a reference this is giving you in dark with a density of CpG content of CpG islands, and you can see in contrast the dynamic regions actually are fairly CpG-poor compared to that. So just to summarise: so we find these regions are far away from the transcription start site, they're very small in size and so meaning these are very focal changes in DNA methylation, and these regions tend to be CpG-poor.
The next one we wanted to look at - what we find in these regions - and so for this we took advantage of a very large data set that was actually generated through ENCODE, and we actually took their transcription factor binding sets or binding data, and then just overlaid where these 161 transcription factors binding the genome and how is that overlapping with dynamic regions? What you can see is quite striking, you can actually see that the majority, over half of the dynamic regions are enriching at least for one, if not more, or two, or more than three different transcription factor binding sites. So there's a striking enrichment for transcription factor binding sites within these dynamic regions, and that's pretty significant because, again, this number is only about 10 per cent of all the known transcription factors; assuming that we will have data for all of those you would probably find that the majority, if not all of these dynamic regions enriching for transcription factors.
We can also then take a closer look at these, and we can actually classify them and so what you find, and I'll just briefly explain this density part here. So you can see on the Y-axis the median methylation level within these cell types, and then you see the maximum difference in terms of methylation between them. So, again, a cell type, for example, anything that would be down here, again, it's unmethylated so it has a low median methylation level, and actually does not change methylation across any one of the 30 cell types. Then overlay it, and this is actually the transcription factor binding site density, and so you can see that over 60 per cent of the transcription factors tend to enrich in these unmethylated regions that are canonically unmethylated, do not change. However, pretty significant numbers are also found in this region that shows the most dynamics, and so these are sites that will be, for example, unmethylated in one or two cell types, but methylated in the majority of other cell types, giving rise to this high median methylation, but then a high variant across different cell types. So, again, these are the ones we're particularly interested in, because these tend to be tissue-specific regulatory events.
So just to show you that the enrichment for these transcription factor binding sites is actually not random, we're now looking just at all the sites that are specifically unmethylated in one cell type, and methylated across the other one. So these are hypermethylated regions, and then, again, we're overlaying them with the transcription factor binding sites, and what you find - and I'll just point at two examples - is, for example, for the human embryonic stem cells, regions that are specifically unmethylated in these cells tend to be most enriched for OCT4, SOX2 and NANOG binding sites. Similarly, if you look, for example, at the adult liver you find that, again, regions that are unmethylated in the liver tend to be enriched for binding sites of FOXA1 and 2, or HNF4 factors, although these are very known regulators for these particular cell types. Again, so just them that you see, these unmethylated regions and specifically enriched for transcription factors that are known to play key roles within these cell and tissue types.
So just to summarise, the first part focused mostly on DNA methylation and what we find when we look at this large data set, we find that about 22 per cent of all CpGs are dynamic in their methylation state. Again, we can see from doing some saturation analysis that even adding more cell types will not significantly increase this number. We see that these dynamic CpGs generally coincide with the gene regulatory, or putative gene regulatory elements that are highly enriched for transcription factor binding sites. We can see that, again, we can use this methylation actually in the dynamic regions to distinguish different classes of transcription factors that may, for example, be sensitive to DNA methylation and thereby regulated and contribute to this regulation.
Next, I want to actually start using what I introduced in the beginning, this human embryonic stem cell model, where we can actually take advantage of the early dynamics as these cells are actually differentiating into different cell types. Then, again, integrate the DNA methylation measurements also with a multi-layered view of the histone modification dynamics. But, just briefly, so the human embryonic stem cells can be maintained in this undifferentiated state by carefully balancing the activity of different signalling pathways. So the reason, obviously, that we are not only interested from a mechanistic basic science viewpoint of these cells, is obviously that there's a lot of therapeutic implications of a better understanding how pluripotent cells can be differentiated. So this just shows you - so the class diagram that highlights what the promise of these pluripotent cells could be in humans. Where, again, now also through the work of Shinya Yamanaka that I mentioned, that you can actually take a somatic cell from a patient and derive these pluripotent cell lines in vitro, you can now imagine that you would differentiate them into specific cell types, and then you can use these particular cell types for both disease modelling as well as eventually for cell replacement therapies.
So, therefore, it's very critical to understand just how these initial decisions are made, and how the cells are actually differentiating and committing to specific fates. Again, as I mentioned, we can also instead of maintaining the undifferentiated sites, we can actually tip the balance of these signalling pathways and thereby differentiate the cell specifically into three embryonic germ layers. For this, again, we took protocols that are well-established for creating all three embryonic germ layers: endoderm, mesoderm, as well as ectoderm. So when you do this using established conditions, what you find is that within a few days and so, again, we fixed them after doing extensive transcription of profiling, we figured out that the five time points seems to give us regional distance from the pluripotent state and well-established marker expression, such as, for example, for the endoderm we can detect very readily SOX17 and FOXA2. For ectoderm, for example, you find, again, many cells part of SOX2 as well as PAX6. Again, we've extensively corrected it using gene expression signatures RNA-Seq and so forth, and so I won't talk about that, but jump directly into the epigenetic dynamics.
Just to remind you, what we're looking at now is in addition to the DNA methylation, which on the left side is shown at the top here using whole gene bisulfite sequencing, we're now looking at various histone modifications. We do this using ChIP sequencing where you use an antibody specific for your modification of interest, and you can pull down the regions where this particular modification was found in the genome, and then you can actually use the DNA pieces that came down with these fragments and use them as tags to identify where in the genome the particular antibody was finding the enriched histone modification. So we can align those back to the genome, and then basically create there what you're used to seeing, these particular enrichment maps where anywhere in the genome where you find the histone modification, you do see this particular enrichment peek and so where you don't have modifications you can see this background signal.
So just to introduce the modifications that we're looking at here, what we use is a couple of modifications - active as well as supressed - if we want them. So we're looking at H3K for monomethylation, which as you can see on the top here, tends to enrich at both promoter as well as distal regulatory types such as enhancers. You can see at the promoter region you can find enrichment of H3K4 trimethylation. Whilst looking at H3K27 acetylation, which, again, you can see is how they enrich active enhancers. I should mention that this gene we're looking at is NANOG and the cell type is a human pluripotent cell. So, as you can also see from the RNA-seq down here, NANOG is how they express them, so this is consistent with seeing this type of enhancer signature, and that's how they enrich for K27 acetylation. You can also see that this gene has been transcribed as it enriches over the gene body for H3K36 trimethylation, and is depleted for any repressive modification such as H3K27 trimethylation and H3K9 trimethylation.
To just show you a couple of examples where we find these repressive modifications, here is an example of a gene that's not expressed on the human embryonic stem cells, and it's called goosecoid. It's been induced upon differentiation and you can see it's lacking a lot of the key characteristic active markers here, and is instead enriched for the repressive modification H3K27 trimethylation. In particular, the trimethylation of K9 you find often at regions such as imprinted genes where you can see both, and now you have the intermediate methylation values here and you can see now enrichment for K9 methylation over this imprinted region.
Just to summarise this overview here, what you can actually see is, first of all, you can see that typically when we find the system modifications and it doesn't matter whether that's the repressive ones or the active ones, you can always see they associate with this dip in DNA methylation; we can see depletion of DNA methylation. We can also see with using just a combination of histone modifications I explained here, that these are very well associated with a particular transcriptional state. Again, there's different ways of utilising repressive modifications, including the K27 repression, or repression by K9, or DNA methylation. So these are the modifications we'll be looking at now across the differentiation into the three germ layers. First, what we do is, we actually define genomic regions of interest and so these are very simply defined as regions that either show presence for ChIP-seq peaks for any one of the modifications that I've just introduced. Or these are regions that are unmethylated, show intermediate methylation values, or are highly-methylated as shown on the right here.
Using that, we actually define not just the six histone modifications and DNA methylation, but we also include a couple of well-established combinations of modifications, such as, for example, the H3K4 trimethylation and K27 trimethylation, the so-called 'bivalent' state. We look now at combinations of modifications, as well as individual combinations, plus the different states of DNA methylation. So combined that gives rise to 11 different states of interest, and so now we can basically look at these and look at genomic features similar to the methylation studies I showed you before. This is, again, a classic gene region where your transcription start site and the promoter is defined as a third proximal region. Then distal is anything that's more than 10 kb up to 50 kb away, and then if it's more than 50 kb away both upstream or downstream, it's considered intergenic. If you now ask where are these dynamic epigenetic regions, including the chromatin dynamics, you find, again, that most of these are actually distal and far away from the transcription start sites. So they're either distal or intergenic, so that makes up more than 60 per cent of the dynamics are quite far away from the transcription start site.
Similar, again, to the DNA methylation dynamics alone, we can see that these regions are actually very focal, so the changes are very small in size where you have a range of about 600 base and it goes to about 4 kb in size where these dynamics and modifications are found. This gives you an overview of the different states across the different cell types, and so this is quite busy but it's meant to mostly show you an overview of what the particular states are in embryonic stem cells on the left. Then how they are changing as they different shade into ectoderm, mesoderm or endoderm. So there's a couple of things you can immediately see: first of all, there's a lot of regions, about 300,000 regions exhibit an enrichment for any one of these particular states. But the more amazing numbers actually that you find within this differentiation, and, again, I should remind you that there is only five days of differentiation from leaving the pluripotent state. We find that about 50 per cent of the regions are actually showing dynamic regulation, and so you can see that it's particularly for the endoderm it's quite striking where you can see this very unique barcode here that's been generated, and that makes the endoderm. But all of the other germ layers, there is too very distinct from the pluripotent state, and so the dynamics that we observed are very lineage-specific.
So an additional data set that I just want to briefly introduce before actually zooming into some of these dynamics, is that we also have ChIP-seq data for some of the core transcription factors, in particular OCT4, SOX2, as well as NANOG. Again, if you look at the binding of these classic core pluripotency factors, you also find that many of them are actually enriched at these distal intergenic sites. Also, what's quite established is that there's unique targets for the facts, but there's also a lot of shared targets and all three factors are co-binding in the same regions. So, typically, what people have seen as, again, these factors are binding their own promoters, they're binding to the factors that are expressing pluripotent cells. They also tend to be highly enriched at the developmental transcription factors that are silenced until they seem to be both regulating repressed targets, as well as activated targets.
Now, I just want to show you a few examples of the dynamics that we observed, and so the first one is centred around those that are actually active and transcribed, and that are being silenced during the differentiation. So this is actually a fairly interesting and unique example here where you see lineage-specific gain of DNA methylation. So the gene is called DBX1 and you can look at the actual proximal region of the gene, which actually harbours a few CpG islands. You can see that there's actually very little dynamics going on, as you would expect, because we know the CpG island sequences are fairly static. But the regions of interest are actually further away from the actual gene, and you can see two different ones that are highlighted with the light grey bars here. We can see the region is actually unmethylated in the embryonic stem cells, it retains this unmethylated state in the ectoderm, but then starts to gain methylation in both the mesoderm and the endoderm.
What's interesting is that this gene is not expressed, even though it's unmethylated in this region in the embryonic stem cells, but it becomes induced as the cells are differentiating towards the ectoderm, but it's not expressed in the two alternative germ layers that do gain methylation here. When we started to look at this closer, what we found is that when we actually look at this particular region we actually find that it's highly enriched for binding of the core pluripotency factors. So that was a little bit surprising, to some extent, because, again, why would these pluripotency factors bind the gene and the region that's not actually utilised in the stem cells, and it's just a distal to the actual gene, and then this region actually retains that unmethylated state and becomes induced in the ectoderm? One thing we could speculate, that in addition to the classic known functions of OCT4, SOX2 and NANOG of activating genes and also repressing developmental transcription factors, that it might play a role with these distal sites where it actually might be involved in maintaining, almost like a placeholder maintaining open chromatin that only would be utilised at later stages of differentiation, but where the factors might be sensitive to methylation. So you need to keep this free of methylation already in the pluripotent state, in order to have it available and accessible and downstream as the cells are differentiating.
So just to summarise this part, so what we see is when we differentiate the embryonic stem cells into the three embryonic germ layers we see a lot of epigenetic dynamics, and all of these lead to very unique and distinct signatures in the three germ layers. We also see that most of the dynamics occur in putative distal regulatory regions, and we see in terms of DNA methylation dynamics that the core components of the pluripotency network, such as OCT4 and NANOG, and that's been well-established and known, are being silenced by DNA methylation. But, on top of that, we can see, and I just showed this one example where you see this specific gain of methylation in two of the germ layers, but not one, and the one where it remains unmethylated, is the one where the gene is being induced. Lastly, again, based on that we speculate or could speculate that the pluripotency factors may also act as these placeholders for maintaining loci that need to be accessible downstream upon differentiation.
I want to show you one example or two examples, actually, two types of examples for the opposite direction of where you go from a heterochromatic or a silent region, to euchromatic region. The genes in this case, the schematic shows an express gene, and so this is the first trend I'll show you. This is a gene involved in neural differentiation, and you can see it's in this region here it's highly methylated in embryonic stem cells. You can see it's very specifically demethylated in the early ectoderm, but retains high levels of methylation in both endoderm and mesoderm. You can see this here in the individual CG data, or on the heat map here you can see this very clear loss of methylation within this particular region. So there's lots of methylations also associated with the change in the expression where it becomes specifically induced in the ectoderm.
That's quite interesting and worth pointing out, is that the loss of methylation, just looking at this example, it's very germ layer specific. So you can see there's very little overlap between regions that are losing methylation in all three germ layers, whereas most of the loss of methylation is unique to the individual germ layers. Again, I pointed out that this gene is changing methylation and losing methylation, and it's being induced to be expressed, and so this is quite a few genes that we see that behave that way. We also see other dynamics that I want to briefly explain, and so, again, this is just a small snapshot of this big heat method I showed you before now focusing only on these dark blue regions which are called highly methylated regions, so those are highly methylated. But, if you look at how they behave as they differentiate them into the different germ layers, you can see all of these switch to alternative states. Many of them switch to, for example, K4 monomethylation and this would include examples like the gene I just showed you that becomes induced upon differentiation.
However, a fairly surprising trend that we saw is very unique, so we can see the gain of this red bar here which is H3K27 trimethylation, and you can see how unique it is for each of the germ layers. That's a surprising switch, because, again, this would be a region that's highly methylated in the stem cells, but now it starts to enrich for H3K27 trimethylation, which is another repressive modification in a very germ layer specific manner. More importantly, if we look at the features of these regions they tend to be fairly CpG-poor, and so this is again the reference just giving CpG islands, and then these regions are changing in their K27 trimethylation state. So that's interesting, because, again, in at least pluripotent cells and through several models it has been suggested that particular polyclonal which targets K27 trimethylation to the genome, tends to be highly enriched in CpG dense regions. So this will be a different trend here where you see a gain of K27 trimethylation in CpG-poor regions.
Now, this is sort of the summary, but to show you one specific example, you can see here a region that is showing you only the K27 trimethylation on the bottom here and not the DNA methylation, but this would be a region that's highly methylated in the cells, and you've specifically now switched to this gain of K27 trimethylation in the early endoderm population. I think I chose this region because these are key genes involved in a hepatocyte differentiation and function. So if you look at an adult liver, the K27 trimethylation disappears again and these genes become induced, and are now enriched for active markers such as H3K4 trimethylation. So, again, now coming back to that question, so why would a region actually switch from high levels of DNA methylation to another repressive modification such as H3K27 trimethylation?
So when we started to look at certain ways of how this could actually be happening specifically, we looked at a couple of different transcription factors, and one that stood out was FOXA2. We can see FOXA2 binding, and FOXA2 is one of the early factors I mentioned being induced in this early endoderm population. It's actually specifically located at the sites that switch DNA methylation to histone modifications, such as K27 trimethylation. So the reason that that's interesting is because FOXA2 belongs to this class of transcription factors that are known as pioneering transcription factors that have the ability to actually access heterochromatic regions. So what one could speculate here happens, is that FOXA2 is actually able to, and might actually be important for accessing these silent heterochromatic regions. Switching them over to a more facultative heterochromatin marker such as H3K27 trimethylation, that then would make these regions still repressed, but more accessible to lineage-specific transcription factors that would not be able to bind regions and that are highly DNA methylated.
Again, we don't have any supporting evidence for that, but it's a very appealing model to think about how you would transition these highly methylated states into lineage-specific states that are not yet being induced for transcription. Just to summarise all of this, so what I've talked to you about is, again, the last part that we see the very lineage-specific loss of DNA methylation. You could also see that there's a strong bias that the ectoderm shows most of the loss. I showed you that DNA methylation decreases particularly in distal regions, and these regions then tend to enrich now and switch to certain histone modifications. Again, either active ones where the gene is being transcribed, but also silent ones such as the H3K27 trimethylation. In particular, this last part here that I mentioned, what we kind of think about now referred to as epigenetic priming, because, again, these regions are not actually switching on gene expression, but they're primed for later stages of differentiation.
So we think, again, and I hypothesise the factors, pioneering factors such as FOXA2 might actually play a critical role in this, as they are able to transition these regions of high DNA methylation to the small facultative heterochromatic state, but then it's accessible for later stages of differentiation. So with this I just want to thank a number of people in my laboratory that have worked on these different projects, in particular Michael Ziller who worked on the first part of the dynamic CpG analysis. Then Casey Gifford who pioneered the differentiation into the embryonics and germ layers, and worked with Michael together on this epigenetic dynamic story; and all other members of the laboratory and my funding sources. Thank you for attending and I'm happy to take questions later.
SF: Thank you Alex for such an interesting talk. I'm sure you will have plenty of questions from our listeners. Hello, I would like to take this opportunity to tell you a bit more about some of the resources and products that Abcam has available for epigenetic research. We have recently launched a new epigenetics microsite at abcam.com/tag/epigenetics. Here we have grouped all our epigenetic resources in one place to make it quicker and easier for you to find our latest tools covering the key mechanisms of DNA methylation, histone modifications and non-coding RNAs. This includes articles, selected products, protocol hints and tips, as well as relevant webinars and events. You can request or download copies of our posters from our website. Please feel free to contact us if you have any suggestions, or comments for a new poster or protocol ideas. If you have any questions regarding Abcam products, please contact our scientific support team who will be very happy to help you with any queries you might have. We have scientific support teams in the US and Hong Kong, in Europe and in Japan.
I would like to highlight that we have multiple language support in German, French and Spanish, as well as in Portuguese, Chinese and Japanese, so please do not hesitate to contact us. Also, please be aware that we recently opened an office in Shanghai as well. If you use large quantities of one product, you might want to look into buying in bulk. Bulk-buying will help you to save money and minimise the variability in your experiment. For more information about our buying options, please contact our sales team at email@example.com. I would now like to find out Abcam's range of epigenetics products, in addition to an increasing range of ChIP-grade antibodies, we offer relevant biochemicals, for example, HDAC inhibitors. We also offer a range of HDAC and SIRT activity kits, as well as ChIP kits that have been designed so that you can spend less time doing the experiment, and more time thinking about the design and outcome. We also have kits for immunoprecipitation of methylated DNA, which I will elaborate later on.
Today I will just mention our range of EpiSeeker ChIP kits for cross-link ChIPs. Our One Step and Plant ChIP kits have been optimised for mammalian plant DNA, respectively. These kits do not contain a preselected antibody, and are therefore adaptable to any target of choice. The EpiSeeker range also includes kits optimised for methylated or acetylated histone modifications. They can be used either for cells or tissue starting material. We have also kits for immunoprecipitation of methylated DNA; these kits contain specific antibodies to enrich methylated or hydroxylated DNA, respectively. More details on these kits can be found on our website at abcam.com/EpiSeeker. You may ask what are the advantages of using the EpiSeeker ChIP kits instead of following the conventional ChIP method? Well, the reaction takes place on a 96 well plate, so they're easy to standardise. It only takes five hours. The kit contains all main reagents, except the cross-linking reagent, and except for the general kits, the kits all contain a preselected ChIP-grade antibody which has been optimised. The precipitated DNA can further be used straightaway for downstream processes, such as ChIP-on-ChIP or ChIP-sequence.
As I just mentioned, we offer methylated DNA in monoprecipitation kits. These kits contain specific antibodies to specifically enrich methylated and hydroxymethylated DNA, respectively. As we heard from Alex, this would on occasions play an important role in differentiating gene expressions, important transfer pop up tools to investigate their function. So our immunoprecipitation kits are also very useful to complement with other DNA methylation experiments, such as bisulfite modifications where it's not possible to differentiate between methylated and hydroxymethylated cytosines. Lastly, I would like to highlight two upcoming Abcam conferences, and these are probably of interest to you. First, the Crossing Boundaries: Linking Metabolisms to Epigenetics meeting that is taking place in Boston in May. More details are here. Second, the Chromatin and Epigenetics: From Omics to Single Cells meeting that is taking place in Strasbourg in October. If you would like more information about these meetings, please visit the meeting website at abcam.com/strasbourg2014.
Also, as a thank you for attending the webinar, we are offering a buy one ChIP-grade antibody get a second antibody for half price promotion. After this webinar you will receive an email with the details about the special offer, and also a PDF version of the slides. Without further delay, I will pass you back over to Alex who is ready to answer the questions we received during the webinar. Thank you very much for your attention.
AM: One of the questions we received was, we find the DNA's hypersensitive sites actually overlapping with sites that are methylated at the same time. So that's actually a very good question, so what I was showing was mostly, was actually the overlap of all the dynamic regions that are highly enriched for the hypersensitive sites. So I guess the question is whether we actually see them co-occur in the methylated state? So the answer is I can't tell you exactly right now how many of those it would be, it depends a little bit on the regions and the way these regions are regulated. So just to give a specific example for regions, for example, that are bound by a factor like OCT4, those would be the ones that I showed that are unmethylated in one or two cell types, but then methylated in all the other ones. So in all those cases where those factors are binding, you would actually see the bound sites are clearly unmethylated, and many of those would be overlapping with hypersensitive sites, and so those would be in that category. There is, however, a small fraction that I didn't talk about that actually seems to be highly methylated, and still enriches for transcription factor binding sites. So that's consistent with some recent work where people have used a raise to basically determine whether certain transcription factors combine methylated DNA.
So, again, those regions you might actually find an overlap between the hypersensitivity that's caused displacement of histones and nucleus zones that could still retain methylation. But we haven't actually looked specifically at that, but it's a good question. So another one that's somewhat related to that, that we received, is basically do you always see the decrease in DNA methylation in regions that show enrichment first or modifications; meaning do they rarely or never coexist? So that one is a little bit complex to answer, but it depends a little bit on the modifications and so there's some clear rules that are well-established. So, for example, you would never find H3K4 monodi- or trimethylation at places where you have DNA methylation, because that has been shown by NICE work from Tim Bestor and others. Is that, again, there's actually some steric hindrance that the DNA methyltransferase can't actually access these regions that are now having modified histones at the lysine 4 residue, and so in those cases it's never co-occurring.
This other, like H3K27 trimethylation, which, again, you could say generally, and at least that's what we observed at CpG dense regions, and to take CpG islands you also never find K27 trimethylation co-occurring with DNA methylation, and they're highly anti-correlated. That's slightly different once you go away from the CG dense regions where you actually do find some overlap between these modifications. So it can't get fully explained, so it depends, and I guess the answer is complex because it depends a little bit on the context for certain modifications in terms of genomic context. For others, again, it's well-established as, for example, K9 trimethylation or dimethylation can actually specifically co-occur, or even help recruit DNA methylation. Again, some do overlap and others show a context-specific anti-correlation. Then the third - another question that we received relates to - and I actually didn't talk about this, but this one relates to non-CG methylation and also whether this changes for different germ layers.
So just to briefly mention, I introduced briefly in the beginning that the predominant target for DNA methylation in mammalian cells is CG methylation. But also if you look at all the cytosines, even in a non-CG context, those will be CpA dinucleotides, or a CpC or CpT. You actually do find that certain mammalian cell types do enrich for these non-CG, or asymmetric dinucleotides. In particular, it's usually the CpA that is found to be where the site is in its modified. Now, it's only certain cell types and so typically you would find that the embryonic stem cells: mouse and human, you find it in the early embryo, and, certainly, you find it in the brain. So, typically, what we do observe for a non-CpG methylation, it's generally only present in cell types that have DNA methyltransferase activity. We see it higher enriched in cell types that are non-dividing, and so, again, that's more sort of the brain you might find higher levels. So the reason is that you cannot actually replicate non-CG patterns at a target-specific site, because once you replicate your DNA one strand contains the methylation info, the other strand actually does not have any cytosines and so it can't cope with this information. So, as a result, the patterns are less stable and usually less maintained and the levels overall are much lower than for a CG methylation.
So that's the background, so just to specifically answer the question, we did look at that and, again, what you usually see when you're differentiating, for example, non-CG methylation levels decrease very rapidly upon differentiation. Most of the somatic cell types, again, we don't actually see detectable levels, but as I introduced, usually this correlates with the expression of the DNA methyltransferase and both DMP3A and 3B are down-regulated as well upon differentiation. So there's no more questions. Thank you for your attention and attending, and that's it.
Thank you Alex for your presentation, and thank you Sonja. Unfortunately, we have not been able to answer all the questions received today, but for those whose questions were not answered, our scientific support team will be in contact with you shortly with a response. If you have any questions about what's been discussed in this webinar or have any technical queries, please do not hesitate to contact the Abcam scientific support team and we'll be very happy to help you. They can be contacted at firstname.lastname@example.org. After this webinar we will be sending you further information about the webinar promotion, and also a PDF copy of the presentation site. We hope you have found this webinar informative and very useful to your work, and we look forward to welcoming you to another webinar in the future. I'd like to thank you again for attending, and good luck with your research.