 Well, thank you very much, and I thank the organizers for inviting me, whether I've made contributions or not, maybe, but it's been long, I've been around a long time for whatever that counts. So I am not a genomicist, I'm not a bioinformatician, I'm a cancer biologist, but I understand what you're hearing here and what ENCODE can contribute to what we as cancer biologists and translational investigators may be able to do using epigenetics to both understand the biology of cancer and the management of cancer is absolutely gargantuan. You've heard elegant examples introduced. I think that in point of fact, we're really in the infancy for doing this. We're in the infancy of my talk, too, since I don't have any slides. I can chalk talk it. I can chalk talk it. I probably wouldn't do justice, but I saw where those peaks were at the transcription start site for the looping, so I could try. All right, there we go. So I'm going to comment on this and some points maybe we need to know, most of them probably in the future yet. But since Waddington really had his idea about epigenetics as heritable states, which the environment can switch without changing the base sequence of DNA, you've heard that and how integral is the genome. The key for the cancer biologist, I think, and the translational cancer researcher is, that unlike reaching in and reprogramming most mutations directly, which can be difficult, certainly with drugs, could be done, genetic changes while heritable are reversible, normally with signaling and in cancer. And this is a key for what we might do about using our knowledge about them to do something about cancer management, not just understanding the biology. Now, for me in the field and for others, it's dizzying because you've been hearing how we've expanded our notion of what constitutes an epigenome. When we started our work, we were promoter-centric entirely. And the example that we worked with the most was that these unmethalated CPG island regions by the white lollipops here would abnormally assume DNA methylation during the course of cancer initiation and progression. We've followed that for a long time and we know hundreds of genes in every patient's cancer, hundreds almost in every patient's individual cancer take on this change. We don't know who the passengers and the drivers are for many of the genes that are affected by this mark, which would be associated with a strong transcriptional repressor at the promoter. And certainly some tumor suppressor genes can clearly do this change and be silenced as an alternative to mutations, so it really can involve some important genes. But now we know who are these regions, what's the relationship of these promoter regions in that DNA methylation change to DNA methylation and chromatin changes at the enhancers? Are the same genes being affected? What's the timing? Different genes? It's all about, I think, learning the balance between the regions you've been hearing about, what's happening in enhancers, and genes such as these. If this is the gatekeeper at the promoter and it's closed by a tight repression mark, is it not responsive to the enhancers? The three-dimensional changes you heard about are important here. All of this, I would posit to you, we really need to learn as cancer biologists to develop really where we're going in the epigenetics. And I'm going to talk about control of long non-coding RNAs, which is a critical role of DNA methylation in our genomes. It's associated with transcriptional repression of transposons and line elements and viral elements, and I'm going to comment on this in the last part of my talk. So it's about balance, I would posit to you. And that balance we need to understand during tumor initiation and progression. We know cancer is a disease of genetics. We know about the driver proteins in this famous model of progression, for example, in colon cancer that accrue KRAS mutations, P53 mutations, and others. But we constantly have to remember that starting right back here, before you develop an invasive carcinoma, and this is true in the colon and probably for most regions of the common cancers, well before an invasive carcinoma is developed, years sometimes, the epigenome is severely altered. And many of those promoter hypermethylation events for DNA that I just described to you can be well defined back here in hyperproliferative stages or premalignant adenomas in the colon. And so what is the balance, again, between these mutations and what's happening in the epigenome? And some of the things that we need to know for those first two points, again, I'll stress. I think we really need to know much more about the relationships in balance during cancer initiation and progression between some of the DNA methylation events I've just initially mentioned and chromatin changes in key genomic regions. How do they relate to one another? The enhancers, the promoters, gene bodies, non-coding regions, we've got to integrate this. And again, I think we're coming, we have the technologies emerging, sometimes technically challenging, particularly in primary tumors, but we're at a stage now I think we can really make tremendous progress in beginning to try to understand this. Now Harry Esuwan is here at this meeting and he's been working as a faculty member in our group for some time now trying to explore the DNA methylation relationships and beginning to chromatin that I think interdigitate these areas I'm talking about. And let me give you one example that I think is very important. I'm going to posit to you that changes in chromatin and DNA methylation affecting these given regions can really tell us about the compartments from which tumors come and their relationships to the specific tumor subtypes. And therefore, if that's true from the translational standpoint, something about perhaps the way we best manage these tumors considering this information. And so our group about maybe eight years ago now, Howard Cedars' group in Israel, Peter Laird's group, Peter Jones' group and others, began to realize that 60 to 70% of those genes that take on that promoter DNA methylation abnormality that I told you about, actually traced back in their chromatin history and embryonic stem cells to control by the polycomb repression system and in a state that Brad Bernstein called bivalent chromatin, a poised state. And that should be a plastic state. When the signaling comes to an embryonic stem cell to leave cell renewal and commit to cell lineage, it may want to turn off sitting genes. It's not going to use in a given lineage and remain in a polycomb state or maybe wanting to be active and take on the active market is along with polycomb in these poised genes. Plasticity is the key. You've got to make those switches properly. Sometime during tumor progression, I just mentioned to you a host of genes go through a process where they substitute the market, the promoter for this kind of chromatin with DNA methylation in normally unmethalated CpG islands. So these regions are not normally controlled by DNA methylation. In cancer, they have evolved that change. So again, what is the balance here? This is a very tight transcriptional mark. Not available to the enhancers anymore? Did the enhancers make a preliminary change that bias to this? We need to know these answers and tumors can be anywhere in this progression molecularly. And so you can have genes anywhere along the spectrum that have been making this convergence from this normal plasticity mark towards this more silenced mark of DNA methylation. And since many of these are developmental genes where the embryonic stem cell uses them to balance between staying in cell renewal and committing to cell lineage, you could easily imagine if you artificially silenced these genes that that would be something that could help lock that cancer cell or subsets of cancer cells within the population in an abnormally cell renewal state and block it from moving towards differentiation. And I would posit again the relationships we've been talking to, we need to understand more holistically the epigenetic changes that might relate to this. How does it look in TCGA? It's important. Several investigators have seen this profile here that Harry and his group see in TCGA. So what you're looking at is an infinium 450 hybridization plot for DNA methylation and only the regions in the promoter at CPG islands are on this plot. Green means the probes that are in those regions are unmethylated and red means they're moving towards more methylation. But I want to show you breast cancer here. The red, these are all triple negative breast cancers here. And you can see for this group of genes, most of which would not be methylated at this region in normal tissues, the triple negatives are not, as opposed to the luminal, more differentiated cancers that have a huge amount of methylation in these genes. Some investigators believe that triple negative tumors are more involved with having genes controlled by polycomb. I can tell you that that may be the case, but if you look at these very same genes, many of them in the triple negative which are not methylated are the ones that are methylated in the luminal. So you've made these switches and could this tell us, is all of this derived from the cell from which triple negative comes, the primitive cell perhaps in the breast epithelium versus a more committed progenitor that gives rise to a more differentiated phenotype? All questions I think that are very important. This is TCGA data for the Nature paper on adenocarcinoma that Peter Laird, who's the PI and the group put together including people at Hopkins, again now yellow's methylated, same probes, type of probes, yellow's methylated, blue's unmethylated, normally unmethylated. I think that the tumors have various patterns for their DNA methylation. And here you find a whole group of adenocarcinomas that have a huge number of the genes methylated, this group doesn't. And who are those genes? If you look at them, they're transcription factors for development like GATA4, they're anti-wint genes like secreted frizzle proteins in the nuclear protein SOC17. They're developmental genes, again, where that same kind of a profile for polycomb and bivalent control in embryonic stem cells now shows up in a big group of genes in the tumor as hypermethylated genes. How did they get there? These are the questions. Another link to genetics you've heard a little about, and many of you know that the deep sequencing efforts in various projects including the genome atlas have turned up one of the most frequent types of mutations that we see in human cancer are in genes encoding proteins that are regulatory proteins for the epigenome. And going around this wheel, they include proteins which move the nucleosomes around in the histones themselves at various marks. Proteins that are in the erasers and the writers, the proteins that put on acetylation, take off acetylation in the histone tails, put on methylation, take off many, many mutations here. And I would posit for the most part, we don't understand the phenotype in the cancer for the epi... Is it an epigenetic change that we see associated with these mutations? You would posit that it should be given the nature of the genes. We understand a few of them now, but you can look at the exercise here that's going to have to go on, I think in the next few years, to understand the relationships in mutations in proteins that could really be associated with profound alterations in the epigenome at the regions that you've been hearing about. And so, again, I would say just as above here, we really have to understand these mutations. Do the mutations create epigenetic changes? Or do they inherit a series of changes that happened early? And once that mutation sees that, it helps lock those in and gives the cell a selective advantage. What's the chicken or the egg relative to these mutations in genes that control the epigenome? And I think that's a very important question that we need to look at. And I want to turn to some thoughts about cancer therapy. And cancer therapy, epigenetic therapy is now a term that you hear a lot about. And what does that really constitute? Well, if you have a mutation like an EGFR mutation in non-small cell lung cancer, that's an addicting mutation in about 5% to 10% of the patients. We have drugs now that can block that mutation and give tremendous early responses in the world's deadliest disease. The problem that happens is all too soon resistance emerges. And we need to understand the nature of that resistance. Some of it may be a new mutation, but in many cases it doesn't look like it is. And so that resistance looks in some investigator's hand to have a strong epigenetic component. And we need to define that. And defining it may give us tools that will either enhance the drugs that block the mutations to prevent that resistance. And so all of these pathways here, which are key tumor-driving pathways around the famous Hanna-Hann Weinberg wheel, the asterisk means just for that one change I told you about, the promoter DNA methylation, multiple of those changes exist in virtually every pathway, sometimes simultaneously in the same patient's tumor. So you could imagine that you might want to reverse some of those abnormalities if they're causing abnormalities of signaling in these pathways. I'm going to tell you about one to end the talk about the fact that tumors can evade the immune system. And how do they do that? And how does this relate to the epigenome? I'm going to tell you a story to end the talk about that. But I think most of the time we're talking about not targeting a single gene if we're going to reverse the epigenetic changes. We're really talking about targeting that epigenetic program. And the question that we ask in epigenetic therapy I think most of the time is, in targeting that program can we reprogram the cancer cell from an epigenetic standpoint? There are older drugs in the clinic trying to do this. And I think I skip one slide or maybe let's go forward here. Let me skip, go back one. I just want to tell you as you go around the nucleosome for the proteins that put on the histone tails, acetylation and methylation and DNA methylation in the genome, the so-called writers, for the proteins that take off those marks, the erasers, for the proteins that read those marks, or for the proteins that move the nucleosomes around. There are small molecules now. And a beautiful example is in Toronto in the structural consortium that Cheryl Aursmith and others. The molecules are out there and available to us now in preclinical to test some of these concepts. Probing the response of the epigenome to these molecules, I think is going to evolve the same kind of genome-wide surveys and looking at relationships that we've been talking about. So let me tell you, again, just make a point here that I believe, again, the above things that we've been talking about, we're really going to be exploring many of those same things when we're looking at agents which target the epigenome for the possibilities in cancer therapy. And I think this is going to be a quite big area which most of the major pharmaceutical companies are now engaged in. So we've been in this, the last part, and I'll tell you one story about how this may be valuable. I am fortunate enough to co-league a stand-up-to-cancer project with my colleague Peter Jones. And I stand up to cancer, which is a clinical trial program based on molecular paradigms. So we have clinical trialists involved, and I'll show you. We have a laboratory component at several institutions. It's to try to bring molecular paradigms for epigenetic changes that are ready to explore in a clinical trial and perform correlative science that says why we brought them to the trial and what's happening during the trial. We're at our end of infancy. We're going to be using, I predict, over the next few years many of the tools that you've been hearing about. And we have a lab component, and I'm going to tell you one story that's come out of the lab from collaboration with my colleague Cindy Zanow. Very talented post-doctoral fellow, Kate Cipionelli, working with a host of others in this program. In the stand-up to cancer, our first trial was to take an old drug, azocytody, which is a drug that directly blocks DNA methylation. So we were aiming to block these DNA methylation changes that the promoters primarily. You'll see we've had to broaden in our thinking about what we're actually doing. And we pitted that with a low dose of a histone deacetylase inhibitor because histone deacetylation accompanies DNA methylation in association with gene silencing. And we were using low doses, and we were tackling advanced stages of non-small cell lung cancer. That, as some of you may know, is our biggest killer. 150,000 people a year just in this country alone die of non-small cell lung cancer. A very aggressive and difficult disease. And we had some early successes in a couple of patients, but we saw a chance collision. This is a lung lesion here in a patient who had progressed on the low-dose epigenetic therapy, but happened to go on to in a subsequent trial, what is arguably now one of the most exciting cancer management strategies, immune checkpoint, blocking immune checkpoint, blocking immune tolerance. And all of the small number of patients, five of them, but all five of them passed a point in progression where only 20% of the patients on the immune checkpoint with the lung cancer got alone. All five passed that and did better, and three of them went into long-term years of remission. Now it could be the immunotherapy, but the question then we had to think about and put into clinical trials, which it is now to validate early on, did the epigenetic therapy somehow prime for the immune checkpoint therapy? We'll talk a little bit about immune checkpoint therapy and what actually constitutes it. And so we brought our trial of low-dose azacytidine from a lab paradigm in which we were taking very low doses relatively of a drug like five-azacytidine with foxy and amethylation and just exposed the cells transiently to it and then stopped the cells from seeing the drug. We wanted to know what did their memory, did it affect their epigenetic memory, and we would rest the cells, they were viable at this point, put them into various assays, but the key point that has really proven valuable for us, we would do genomics. We need to do much more and deeper kinds of genomics. But at time points after stopping the drug, we would look at DNA-wide, genome-wide DNA methylation, look at expression, looking at proteins, looking and now doing proteomics, and we would build a signature that said, what did the azacytidine do to the signature of that cell and that transient exposure? And then we could take the basal, at least, levels of those signature and those transcripts and probe TCGA. And now, beginning probably with enhancer positions and ENCODE in many projects, the point is we started looking at databases for what that basal signature did. And what we quickly learned relative to the trial for immune checkpoint is people had known that azacytidine up-regulates immune parameters in cells. But what I guess we added to this was how coordinated this up-regulation is. Because what we learned was that antigens, neoantigens on the surface of the cancer cells are up-regulated by the drug. The proteins that assemble those antigens in the MHC1 complex and endoplasmic reticulum up-regulated by the drug. Pro-inflammatory responses with cytokines. And I'm going to mention to end the talk about the up-regulation of an old pathway for viral, sensing viral infection. And this is what we saw and this is what we've linked to possibly why this drug might prime. So what we thought we maybe were doing with the drug is reversing immune evasion. And the concept of immune evasion, popularized by Schreiber and others, is that as tumors evolve, early on they actually may do things like cytokine production, tumor antigenics being up-regulated on the surface, promote an immune attraction signature, which would actually bring T-cells and other macrophages and others to the tumor environment. But of course that would be against the tumors' job description to stay alive. And so some tumors, either from the get-go or during progression, instead of maintaining that immune attraction signature, stay under the rodar and lose it. And they lose it by down-regulation of antigens and various cytokines and various pro-inflammatory signals. But one thing they also do is develop immune tolerance. Signals that can render T-cells as effector non-reactive immune cells rather than T-effector cells. And this is the basis for immune checkpoint therapy, which really arose out of elegant pre-existing basic science showing the receptors on tumor cells that could interact, or ligands on tumor cells in general, they could interact with receptors on macrophages or T-cells and that chronic interaction promoted signaling that made the T-cells tolerant. And what immune checkpoint therapy and the main is doing right now is using antibodies against those receptors and ligands to break that interaction to reverse the immune tolerance. And we had seen a signal that included up-regulation of one of these ligands that can mediate tolerance, actually PD-L1 by ASA Cytody. So we continued to mine databases and work and developed the 300 gene signature, which we called ASA-inducible immune gene sets, or AIMS. And then we took that further and Dan B. Carvalho's group, we have a paper in press, his paper is co-impressed with this, came to the very same conclusion, that the key central pathway of what the drug may be doing is this double-stranded RNA viral sensing program in the cytosol. And blocking key constituents of that pathway is shown by the X's here, including an interferon beta readout when you stimulate this pathway, can block the ASA responses and turn out the lights, I guess. It's pretty powerful. And so this is the pathway, and I can't show you all the details that we discovered, and that we then wondered what could be triggering that pathway? Where were those transcripts for double-stranded RNA and viral-like becoming from? And we thought about the fact that we've got all these endogenous retroviral elements in our genome, over 8%. Many of them are silenced normally in association with DNA methylation around their promoter-like LTR areas. A fascinating group of proteins that are non-coding in many regions, but coding in some, repeat families of these herbs that we knew nothing about. And a long story short is that we showed, indeed, by probing them individually with some collaborators in Germany who had made PCR reactions against the non-repeat sequences in these herbs we were up-regulating these transcripts. We don't know that this is the only double-stranded RNA. It is probably not that we're inducing. But many herbs are up-regulated, as shown in the red in these primary ovarian cells here and low here, and they group with an up-regulation of this viral-defense signature in the primary tumors. So now we can go back to databases again. We can ask some questions about this whole signature. If we take the key genes in the viral-defense signature, they're basal levels in the RNA-seq from the TCGA, and put them through and do clustering on those primary RNA-seq signatures, this is what we see, which is so tantalizing to us that it relates to the immune primary. The red would be the high signature, the intermediate, and low here. And this is ovarian cancer, in our ASA site of Dean work. And what you see here in the red, these are all the tumors for ovarian cancer in the red that TCGA is called immunoreactive ovarian tumors. And the high viral-defense basal signature for the genes in the pathway encompass virtually all those tumors, and the low do not. This is colon cancer. And colon cancer has a very high group in an intermediate, and the blue here are right-sided colon cancers, which is important. So you have a high viral-defense signature encompassing all the right-sided colon cancers. Why is that important? It's because those tumors are full of mutations. They have a heavy, mutational load, which has now been linked to the other key feature that seems to auger good response to immune checkpoint therapy. So there's a collision here between the high viral-defense signature and the mutational signature for the cancer. And every tumor type we've tested has breakdowns that are important, including the poster child for immune checkpoint therapy, melanoma. So can we actually take this to the clinic? And the last slide show you that we're actually starting to get data in the clinic that suggests why this ought to work, whether it's going to work or not, another matter, to have epigenetic therapy with drugs like azocytidine So this is an actual clinical trial for Memorial Sloan Kettering where they had in advanced melanoma the tumor tissues in a trial for immune checkpoint therapy. And again, what we could do, either in the pre-bopsies, or this is all the biopsies here, they could put our viral-defense signature through that. And what you see in the red here is the high viral-defense signature correlated with the patients who had long-term benefit in the trial for the most part the blue low-signature associated with those that didn't. And in the gray is the mutational burden. So most of the ones with a high viral-defense signature also had a high-mutational burden, and most of the ones with a low signature did not. And if you quantitate gene levels, the viral-defense gene levels in the tumors are pretty good in separating the good responders here, the long-term benefiters So, again, augurs that may be the following model, and we're working on this in the lab obviously with mouse models now that have a syngenetic immune background, that if you use something like a DNA methyl transferase or a DNA demethylating agent, and you do things like reactivating these endogenous viral sequences, you trigger this pathway for double-stranded RNA and an interferon response, and you're immune-attractive, but you may be attracting ineffective T-cells, they may be T-effectors at that point. But if you come in in association with this and block the immune checkpoint, could this enhance the response and really get it way up in many of the common tumors? And only clinical trials were so, and here's a biomarker hypothesis of the things we could follow, including the IRV transcripts, the viral-defense genes, that were monitored during the trials, and I'll stop there and thank you. So I wanted to ask you what you thought the relevance of ASA-C was in solid tumors, especially lung cancer, since platinum-based doublets were the standard of care. And a quick look at clinicaltrials.gov shows 40 completed or recruiting trials with ASA-C, most of them in leukemia. And I'll show your trial that has completed enrollment. So I just wanted to get your thoughts on that. Well, I think this is the big challenge in the field. I think if you use the right dosing, I'm positing to you if you use it right in the solid tumors. For those of you who don't know, ASA-Cytidine and its congenar deoxy-ASA-Cytidine are FDA-approved in the pre-leukemic disease myelodysplasia. And it was approved after people brought the doses with less toxicity, maybe more on target, and they get great responses. In new agents now, up to 40-50% responses in some of the leukemias. The same signatures and changes, and if this immune change is operative, should be there in the major solid tumors that are in the leukemia. So the challenge is to bring it. I've shown you, we're getting some early signals. We're beginning to learn. There are other small molecules targeting the epigenome and can they go to solid tumors? I think we can think about it and we can use biochemical strategies to think how we might do this. Steve, I actually have one question. Sure. So the ERVs and actually a lot of different kinds of repetitive elements now light up in various kinds of data from ENCODE and abnormal cells and also lots in normal cells. And the ERVs in particular, a lot of them are mappable, so there's a lot of data. What I'm wondering is if you might have looked to see if there is any signature chromatin signature in normal cells that might predict which of these ERVs is being reactivated. I think this is the spot-on question for the types of things we're talking about. We do know DNA methylation. Normally, it's very interesting I mentioned the ERVs, as you may know are highly expressed in embryonic stem cells and some people think they have function either for generating viral protection of the embryonic stem cells or how they see their microenvironment and niches and then they generally turn off and many of them somatic cells are methylated. So what the tumors are doing, they're either losing that methylation and highly expressing some of these ERVs or they're keeping it and we would hypothesize that we're up-regulating those that kept it but there are probably other chromatin factors obviously in play here and recognizing these in the RNA-seq and looking for them I think is going to be important in other repeat sequences. And a final comment. Is this a broader RNA-editing situation that the cancer cell is dealing with and seeing that helps set these and we have some preliminary evidence that it's broader than this and maybe RNA-editing and even mutations in the RNA are they important in triggering this whole thing related? So I think there's a lot to think about. It's cool, it's great. Thanks again. Thank you. Okay.