 Oh yeah. Good. So I'm very glad to have come to this meeting because I've learned an enormous amount. I'm not in end code and I don't do end code type experiments. So it's actually been very useful. I've used end code and it's been very useful to use it and to see how people are thinking about using it. So I think I'll be talking about a much more focused topic than many of you have talked about. It's going to be about DNA modifications. And I hope that I can bring some perspectives to what I think so far has been a very small fraction of what end code has been doing. So that's how to move the... Yeah, just this. Thank you. So as probably all of you know, it's a very short introduction to DNA methylation. Methylation tends to be, oops, in CPG contexts and symmetrically methylated. That's about 90% of the... Or more of the CPG's and differentiated cells in the mammalian genome probably saving neurons, which Joe Ecker has done a lot of work on non-CPG methylation. Anyway, the methylated CPGs are in the major groove available for binding by various kinds of metal binding proteins. And the way the standard way for DNA methylation to be erased is in fact just DNA replication, which replaces the methylated cytosine on the replicated strand of the plane cytosine, then DNMT1 puts it back. DNMT, UHRF1 complex. And basically, if you prevent DNMT1 from putting the metal group back, what you get is a slow, progressive, replication-dependent loss of DNA methylation. Now, since the tets were discovered in our lab about six years ago now, it's been clear that while DNMT's put on the metal cytosine, another mechanism for removing the DNA methylation mark involves this progressive oxidation of metal cytosine by-tet proteins to 5-hydroxymetal, 5-formal, and 5-carboxy cytosine. And these are all, they're all epigenetic marks, and of course they can come in combinations so that what you have is you can have, what you can have is different kinds of methylation marks here. So there are 25 possible combinations for epigenetic readers of different kinds to look at. Now, all of these marks, if they're opposite from the methylated cytosine, tend to prevent DNMT1 from doing the maintenance methylation. So they would, in fact, interfere with maintenance DNA methylation and be a way of facilitating replication-dependent DNA demethylation. And there are also some methods that involve base excision repair proteins like thymine DNA glycosylase, which normally takes a base, a mismatched TG dinucleotide and excises it, but in this case can also recognize the perfectly well base paired 5-formal and 5-carboxy cytosine and excise those, and then those are converted back to cytosine by the base excision repair machinery. So there are a couple of other mechanisms that have been suggested for demethylation that I won't go into because they're not as well validated. So where is 5-HMC in the genome? It's highest in gene bodies of the most highly transcribed genes, and I think you can see that over here. That's the top, 25 top quartile of 25% of most highly expressed genes. And you can see the unexpressed genes are really not enriched for 5-HMC in this standard DNA affinity enrichment method, but the lowest expressed genes you can see are slightly enriched but not highly. It's the most highly expressed genes that are the most highly enriched. And if you look at enhancers marked by K4ME1 and K27 acetylation, the color gradient gives you how much 5-HMC there is in these enhancers, and you can see that it's the most active enhancers is defined by high levels of both marks that have the highest levels of 5-HMC. And these are all highly transcribed regions of the genome, gene bodies as well as active enhancers. It is just heard from Blankingen who just spoke, sorry. It's all right. And this thing that you've shown here in T cells has in fact been shown for neurons and during spermatogenesis, during erythroid differentiation. So it's actually true in many different cell types. So this is also shown in this RNA-seq 5-HMC picture where you can see just called them cell type 1 and cell type 2. They have two different kinds of cells in the thymus, and you can see that the 5-HMC levels and the RNA levels match very nicely. So these two are expressed in cell type 2, and this one is expressed in cell type 1. So basically what we were trying to do was to understand the functions of tetanzymes and mammals, and for the purposes of this talk, I'll try to relate it to what we know about DNA methylation changes. So there are three tetanzymes. We've made mice with conditioned floxalleles of all of them, and we've found that the individual mutations give you relatively mild phenotypes, depending on whether you're looking at embryonic development or particular tissues. And what we do know is that double and triple mutations that we made different types and different cell types actually give you fairly striking phenotypes such as embryonic lethality and ablation of specifically lineages. So for instance, when we delete tetanzymes in hemopoietic cells, we lose red blood cells, we lose these cells, depending on what cre we use. And what I'm really going to focus on today is the relation to cancer, and I'll go through it very quickly. So at about the same time that we were identifying the enzymatic activity of tetanzymes, a different group in France showed by looking at a very small deletion in this 4Q24 region, which is often mutated in milder dysplastics and drones and so on, that one small deletion contained only the Tet2 genes. And this was followed by flurry of whole axon sequencing that basically said the mutations were all across the Tet genes. And we and others have shown that these are very likely all loss of function mutations of Tet2. Since then, there have been any number of papers relating all sorts of cancers, including solid cancers to mutations in Tet1 and Tet2, or to loss of hydroxy metal cytosine, which is a readout for all Tet activity. And so the question became for us, we work with model organisms, not so much to relate the mutations in human cancers to what's going on, but to see whether we can induce an acute cancer of different cell types in a model organism by acutely deleting Tet proteins. And this, in fact, turned out to be true. Basically, we asked the question, is acute loss of Tet function associated with cancer? And the answer is yes. We can't delete just one Tet protein, we have to delete two. And haven't yet been able to do it with three because of the embryonic lethality, but two Tet proteins definitely with Tet2 and Tet3 so far. It's what we've tested because they're present to the highest extents in differentiated cells. And we've taken it out, taken out Tet2 and Tet3 together with either with amyxcite CRE, followed by double-stranded RNA injection, or with CRE-RT2, entomoxifen injection. And the effects of deletion, if we do it either way, which is an acute deletion, is first seen in the hemiportic stem cell compartment. And what we get is an aggressive my lord leukemia. If we delete Tet2 and Tet3 with CID4 CRE in T cells, what we get is aggressive antigen-driven T cell leukemia. In both of these cases, the cancers are cell intrinsic, polyclonal, transmissible indefinitely to recipient mice, and the cancer develops really rapidly. And so we have to assume that the initiating event in each one of these cases is the acute and profound loss of Tet function, which derives from losing this hydroxymethyl cytosine and other oxidized metal cytosines in the genome, or maybe changes in DNA methylation. So that was one of the questions we asked. We took Lsk cells from the hemiportic stem precursor cells, Lsk cells, from mice that had either been either controlled mice or mice that had been before deletion, or mice after the acute temoxifen or CRE-RT2 treatment, and did whole genome-bysulfite sequencing on it. And now I'm completing a comment to the students. So one of the things I'm going to be really presenting this data for is to solicit your opinion, whether we're doing it right. And basically what we've done is taken three control samples, three biological replicas of control Lsk cells in three knockouts, and sequenced to fairly high depth, so that's 200 million reads, most of which are mappable, and the coverage in each case is about seven-fold, so that we have an overall coverage of about 21, 20-fold in each one of those cases. And that translates to pretty good coverage of CPGs in the genome. And that was validated, our choice of this seven-fold and so on was validated by this paper, which is a Nature Methods paper, I believe, that tells us that you need at least two replicas, and you need a certain depth of coverage in order to get a sequencing. So what do we find out? We find out that the canyons of DNA methylation, that's a terminology used by Peggy Goodell, or what Bing Ren has called valleys of DNA methylation, or some of these canyons and valleys correspond to CPG islands in the genome, do in fact, if we look at the control versus tech knockouts, do in fact shrink in terms of getting extra methylation of the edges, and this has been called methylation of CPG island shores for what it's worth. And that's true of both of these genes, and if we look at what both of these genes do, both of them are down-regulated in tech knockouts, relative to wild type. The promoter regions of both of these genes are actually present within the valleys, and I think that's shown here. And so one could take this handful of genes which shows this behavior and say, well, we've shown that there are methylation-associated gene-expressing changes, and therefore this sort of explains the phenotype. I would rather say that if you look at the whole genome level, however, the changes we see are really mild. So for instance, here are three replicates. So those are three lines of the controls and three lines of the tech knockouts, and what you can see over here is all across the genome you see an increase in methylation, but it's not really very striking at this average level at all, and that's also true for active enhancers. It's a little higher there, but basically the level of DNA methylation that we see is not change in DNA methylation is not very high. We can also look at this in the previous slide, I was looking at it at all genes, so that's 21,000 genes, a lot of those are not expressed, and we then focused our attention on the differentially expressed genes in both of these cell types, and you can see once again there's a minor change in the average level of DNA methylation. So what Edahi Avalos and Lucas Chavez, my informatic colleagues did, was basically to sum up the methylation at the transcription start site and in the three quarters of the end of this model gene body, and then plot those values in a sort of a dot plot where each dot here then represents one region of the genome, and what you can see over here is if you look at each one of these dots there's as the trend is towards increased methylation in the TET knockouts, but only about a 20 percent change in methylation overall, and if you look at up-regulated genes and down-regulated genes that are about twice as many up-regulated as down-regulated genes, and you can see that in both cases they gain methylation. So we're not seeing that methylation is making any change either at the transcription start site or in the gene body to overall gene methylation. There are some genes that follow the pattern, over here are genes that are up-regulated and lose methylation, and here are genes that are down-regulated and gain methylation, but overall there's really no particular pattern. So what do we need to make of this? I don't really know at this point. Are we to focus on the small subset of genes that just show the kinds of changes in quote unquote expect, or are we to look at the entire pattern and say that we may be changing the metabolic state of the cell? But we don't really know. However, that's probably a reasonable question for later, but what I would like to point out to you is that in fact the measurements that we're making with this whole genome by sulfite sequencing don't distinguish five mc to from five hmc, because what happens is that neither of these gets deaminated upon sodium by sulfite treatment. Similarly, C, Fc and Cac all do get deaminated in terms of deep on sodium by sulfite treatment, and therefore these are all read as C after amplification and conversion, and really since the numbers of Fc and Cac, at least in our CS cells, are very low, what we should at least be trying to do is distinguish tribe mc from tribe hmc. So the recommendations for this first part of the talk are basically to include oxymc or at least five hmc measurements in any of the DNA methylation analyses that in code might be contemplating doing in the future. So as I told you, five mc and five hmc can be distinguished by certain techniques such as tabcic and oxbsc, but these are really too expensive to do and on a routine basis they require about 10 times as much coverage as simply whole genome by sulfite mapping, so one would have to do it at an apricon level. And also remember that we are talking about five new cytosine bases, unmodified, five mc, five hmc, five fc, and five Cac, but the change is binary, whereas what we're looking at when we say a change in methylation level is a change at the population level. And there's some undefined proportion of change state. So the other thing we probably ought to do is include perturbations and kinetic measurements. These changes seem to be much more dynamic than we previously appreciated. And they may happen on very rapid timescales. And finally, I think one really should encourage the development of new sequencing methods to map modified cytosines in unamplified genomic DNA, and that would include ideally long reads, so as to allow unambiguous mapping of repetitive regions of the genome. One small section. Okay, so I want to bring up the idea that one should also be looking at primary cells that are available in small numbers. And in order to do that, what we did was look at a phenomenon called T-cell exhaustion, which is observed in chronic viral infections and in tumor infiltrating cells and appears to resolve for prolonged exposure to antigen. And it's important in the context of certain immunodeficiencies. And what I'm just showing you here is some attack-seek data that are looking at naive cells, effective cells, and exhausted cells in a region. And what you can see over here, looking at these tiny numbers of cells, only about five times than four cells, is that there are, in fact, regions that are accessible to exhausted cells in vivo, in the chronic viral infection context, and do not appear when you stimulate the T-cells in vitro. So I think something like this should be incorporated into the flow for attack-seek. And you can see over here that the same peaks show up in cells that are tumor infiltrating and also exposed to antigen and don't show up in visor standard cells in the tumor. And so I'll skip forward and just tell you about my collaborators. I think Irvind actually is a very important resource on the NIH campus. And these are the people who did the work on the hemopoietic cells. And this is the person who did the work on the T-cell cancer. And these are our prime climatic colleagues whom I've tended to mention. Thank you. We're still 10 minutes away from where we'll be. I guess we need a moment to do some recalibration of the AV. But our next speaker will be John O'Shea, who will ask us not what, well, ask not what you can do for ENCODE, but what ENCODE can do for you.