Colon Cancer Adenocarcinoma

>> good afternoon, everyone. welcome to the first of the academic year wednesday afternoon lectures, given this afternoon by a distinguished professor from hopkins, andy feinberg. those of you new to nih, hope you'll make a practice of coming

here on wednesday afternoons, we line up quite a remarkable number of speakers, often times asking them to put forward not just stuff that we all know about but provocative new ideas, today will be no exception. the list of some 35 speakers lined up for this coming

academic year is available on the web and the various posters that you might note and get on your calendar. welcome to all those watching on the web, masur is nicely full. there are hundreds of you who are trying to multi-task there in whatever location you're in

the middle of and welcome, glad you could join us in this fashion. it's my great privilege to as nih director to introduce many wednesday afternoon lecturers, when i have the chance to do so it's always fun to remind myself a little bit what field they

have been working in. in this instance, this is somebody i know quite well. andy feinberg and i were faculty together at the university of michigan back in the 1980's, and early 1990's, and have remained friends and colleagues ever since.

he is in fact a remarkably creative scientist who has over the course of several years, know a couple decades, contributed in significant ways to our understanding, particularly of epigenetics, a field which perhaps when he started into it was almost

nonexistent or considered to be squishy and lamarckian, and now, of course, is the subject of a great deal of interest and great deal of nih funded work and great deal of insight, understanding about how modification in terms of methylation of dna and what

happens with proteins a bind dna, have a profound impact on biology and medicine. andy started as a geek in the mathematical arena, hung up on fibonacci numbers. he got undergraduate training at yale and moved to hopkins

program. andy and bert defined hypomethylation as a signature of cancer that had not been previously noted, subject thpublished in 1982. after that he headed to the university of michigan, his lab and my lab were in ajaysentsadjacent

buildings, he and i were also both trained as medical geneticists that we would share experiences going to the clinic, sometimes in ann arbor, i believe it was there he encountered the syndrome which ultimately became a significant

insight for him and for the rest of us about how epigenomics can play a role in medical illness in a way not previously appreciated. eight spending eight years at michigan, he went back to hopkins, currently the gilman scholar, and the director of the

center for epigenetics, a member of the institute of medicine, elected to that role, and proud to say he's also fairly recent applicant for pioneer award, way to provide creative investigators with freedom to pursue ideas that inside that not fit the ro-1 mechanism in

which he used in a variety of interesting ways to look at a fantastic epigenetic model for evolution and disease studying honeybees. i don't know what he's going to put in front of you but i'm sure you'll find it interesting. please join me in welcoming dr.

andrew feinberg. >> thank you. i'm thrilled to be here, i'm incredibly grateful to francis for that exceptionally kind and generous introduction, and to louie for nominating me to give this today. i wanted to point out that the

beckwith leaderman patient was brought to my attention by francis, with his other skills is an incredibly gifted clinical geneticist, researching the patient had beckwith, and it fit my research and it was transformative for many years. i'm going to talk about the

epigenetic basis of common human disease. antoine, my microphone is good? thanks. this is where i work in the center for epigenetics, a google earth view. so like all hopkins buildings, it looks like a parking

structure but in fact there are laboratories in there, and so all generalettists are interested in phenotippic variations, interested in all of them but we're funded to look at the ones relevant to human i would argue if you asked the

basis of phenotippic variation, how is a human different from a chimpanzee, with respect to dr. goodall, differences are modest compared to something like a plant or epithelium, which is what i worked on before bert's lab, they are entirely explained by information we have.

we don't know how to interpret it but we have it all. we know what the complete sequence is of the species and that defies what th defines whatthe differences are. a more complicated question, what's the difference between the brain and heart and liver

and colon? here are all these different tissues, in fact it's easy to see that the stomach, say, of a chimpanzee is far more different than the eyeball of a chimpanzee and the stomach of a chimpanzee than the stomach of a human being and yet the tissues know

what they are. they have information that defines their function, and tells them what to do and they remember what they are the cells divide. that's what we really mean by epigenetic information. there are modification of the

genome that occur during development and define tissue-specific differences, other than the dna sequence it self which is identical across the tissue. it's even more complicated than that because you may know dr. ferucci also has an address in

florence, italy. i was over there not that long ago, and i encountered this 8-foot guy, david, i invited him to be the perfect example of human development, to visit us in baltimore. and he went down the street and had one of these double whopper

cheeseburger things and i'm sad to say he wound up like this. the point is that our environment shapes in a remarkable way our phenotype, but not through the genes themselves, but through influencing, how the epigenetic might take place.

the person who coined the term, epigenetics, was conrad waddington in the 1950's at came cambridge university, he said they arrived from genotype through programmed change and interaction with the environment, pulled into a particular pattern of, say,

tissue development. his original mono graph was something like this, where you have apleuria potent cell that becomes a liver or cell but people pointed out water runs in the other direction, literally, his friend, piper, the landscape artest, redreist, redrew theball rolling

down the hill. the environment might push things up this way, they eventually roll to this one or this one, each of the different tissue types controlled according to waddington, by your sequence. it's called panellization, the

modern definition of epigenetics is different and more flexible and more information-based. that is modifications of dna are associative factors content maintained during cell division other than the sequence. i alluded to that earlier. to give you an idea, here is a

cell, a gene, that's making an rna active, and a gene transcriptionally silent. one thing is dna methation, cpg, but also know there's noncpg methation, we know of no mechanism to copy that. this is copied. don't have time to go into the

mechanisms but an enzyme does that. that's associated in general with gene silencing and thought until some recent studies that we're going to talk about to be enriched or almost entirely at dense regions with many cpg's called cpg islandings.

there's histone modifications. there are some of these posttranslational modifications of histones, you see the dna, associated with activation and other ones associated with silencing, and they are quite different. it's either this modification

for activation or the silencing but never for both. and then there are proteins that recognize the complexes, and then those are triauthothorax, polycomb, and the density of nucleosumee not like my daughter packs her suitcase, which is interesting and no order to it.

but in an ordered way. now, i'm going to talk about things bothering me four or five years ago, haunting me. one is the missing heritability of common scenes. disease, common variants explain 1 to 20% of genetic variants of a disease.

it's puzzling. many good explanations, including where variants, things that we can't just identify because of the way that our studies are -- epidemiologic cohorts, it's an active area of study, something people wonder about.

there was a provocative paper in science translational medicine from my former mentor, bert fogel stein. he didn't mean harm but it was a twin study and showed that it's difficult to explain the contribution to disease greater than the factor of about 20%,

and there must be a very strong therefore environmental component. there's humorous give and take, all well meaning really, in the comments that aid period after that paper came out. appeared after epigenetics play a role.

as france francis alluded to, inthe popular science, not so much anymore in the scientific community, there might be a lamarckian role, and that idea is that -- you have to read lamarck himself, the way he described it, there are fluxors, caused by environmental

exposure, so the giraffe occurred because some horse had to stretch his neck to eat a leaf and the fluxors got transmitted to project anprogeny and that carried on. there's a picture of lamarck. i'm not saying there isn't some component of transgenerational

inheritance, but it would be stablely transmitted, here is the important thing, over and over and over again despite germ 1 reprogramming and undergoing evolutionary solution is difficult for me. i'll come back to that in just a moment.

so about 12, 13 years ago, i became very interested in whether or not one could start to explore the epigenetic basis of common disease generally, including cancer, but other diseases as well. cancer is one where people have finally come to accept an

epigenetic component but what about all the other diseases? particularly where environmental exposure might be important. and in 2004, a student, hans and danny, an epidemiologist i work closely with, we didn't call it epigenetic epidemiology. we had another term.

that term belongs to stephan, but our suggestion was that you could integrate environmental exposure, dna variation in modifying the epigeneti upygenome. it would integrate into the epigenetic to define phenotypes, that doesn't mean it explains all or most of genetics but

might explain some, particularly that related to the environment. that was the idea. we said in here, i think a neat point, you could get quantitative traits over modifying the epigenome without invoking lots of individual variants.

they could integrate, or together with variation, into a quantitative straight. we put in a grant i was thrilled to get funded for centers for epigenetics at johns hopkins, the genome institution, tinstitute to look for tools, a program created by francis when he was director, to

develop and apply tools across the genome and populations of patients. the first thing we did was with the help of rafael rosario, develop an array-based acronym, charm, to appeal to the moniker for baltimore, charm city, it's actually baltimore, people

snicker because of tv shows, but it's a great town. and that enabled us to interrogate the genome at a greater scale than people had done before, that the typical thing was 14,000 and 40,000 of these islands. this is up to four million

sites. and then we didn't invent whole genome bisulfite sequencing but ben langley invented bow tie and a brilliant statistician now at johns hopkins, developing tools to reduce the cost of analysis, a clever way to improve extracting epigenetic

information from sequencing. the latest bonnet, there's a center at hopkins for keeping track of stars in the sky, and alex leads that, the idea is to try to use the graphical processors, cheap little processors for $50, and so we're hoping to use that power because

i'll tell you, the biggest bottleneck in our research, i hear this from other investigators too, probably a paper on this, the computation. it's not the experiment anymore. it's just the computer time and analysis. so i'm going to show you the

second big puzzle that was eating at me from a study, really our first big study, of the epigenome of normal and cancer cells and this is the result of the charm analysis, we're looking at millions of sites, dna methylation, across the genome.

the question, how is colon cancer different from the normal coal mucosa, we're asking charm. red is more meth lated blue is more methylated. you see they are very different. here is a great big surprise. if you look at the diagram over here, he's are the sites.

if you ask an autopsy specimen how well they differ, they do it completely. the way of saying that formally, the cancer differential meth why latemethylated, colon cancer, it isn't normal, it's the same thing that goes back to my original analogy telling the

eyeball it's not a stomach. why should it be general like that? how does that relate to function? it was a puzzle. the other thing i noticed is that, yes, the changes are the same but the epigenetic targets

are hypervariable. there's a great deal of variability, you see it in the pattern of tumors compared to normal, even in the normal itself, and look at this. there is clearly a great deal of variability in the normal tissue as well.

that could be caused by the genome sequence, right? sequences with dna methylation. if you look at inbred mice, the samples right here and those samples right there, look at the variability. these are brother-sister inbred strains, they still show

variability, even though regions are conserved from human to mouse, there are variable legions from animal to animal but the locations are conserved, from one species to another. that's really strange results. it was eating at me, as well as the is missing heratibility.

i don't know you can rely on lamarckianism. i call this my epiphany. it's a way o saying a good ideaat church. i was sightseeing with my son in london. what really happened, we wanted to go up big ben, there's a sign

that says in polite english, if you're not british, get lost. that's not the words they used. right next door is westminster abbey, and they had a sign out front that said, it's the 150th anniversary of origin of species, darwin is dead but it's still fun to see him.

i changed the language on the poster. i found myself standing on darwin's grave. and right next to darwin's grave is newton's grave. what i'm not showing you is there's a velvet rope right here, and that's because the

british do not want you to put your filthy feet on isaac newton's grave but they don't have such misgivings about darwin and they also decorate newton's grave with cherubs and things like that. they really like him a lot. and just above darwin's grave is

a plaque to durak, one of the people who founded quantum theory. i thought ma maybe he was buried standing up. the laboratory said, no, no, it's an urn. i was explaining this to people at the university of miami, they

said he's buried in bonita springs, florida. this is a commemorative plaque. there's nothing like that here in biology. so maybe there's something about this that's the core element of quantum theory that would apply to biology and maybe solve this

conundrum about where epigenetics fits in and where that variability comes from and maybe that would make sense in evolution. this is the idea i had. so that you could have fantastic epigenetic vair yack variationas a drivings for.

you don't need this in single cell organisms. there are mutational mechanisms that do what i'm saying in bacteria. but for a complex species, multi-celled organism, i thought it would make sense. bear in mind these are not

quantum mechanical differences but the idea is just like -- to be completely correct, an electron fits in this one little place, for a chemical reaction to take place it has to go into one particular place over here, so it can be shared between two atoms, it would never happen.

the only way it ever happens, if there's some sort of probability field and some interaction. similarly, i think it makes more sense in developmental biology if the ligan's and reseptemberrors, if there's a gradient or variability in the expression of both so some

cells, signalling molecules from the root or bud would interact with receptors there at the right concentration in the mesenchyme. maybe that's the reason so many cells die during development. you need stochasticicity. if that's the case, the degree

of stochastic variation could be controlled by general it they would lead inherently by the name of sequence to epigenetic variability. genetic variants that increase epigenetic plasticity could increase. the environment is one way, and

then totally opposite in an unpredictable way, it goes back and forth from time to time, after periods of selection, you would select such variants. i'm going to show a move to illustrate. here is the conventional theory. let's say you have 100 people.

i'm going to talking about sizes of people but i'm not really talking about people, i'm talking about circles that i'm calling people. these are not actual people, okay? but let's say you have 100 people, and they have varying

sizes noted by the size of the circle, some small, some big, controlled generalet really so the red has three gaines genes for big, blue three genes for being small. under the normal way we think about evolutionally biology, the large survive, if there's

abundant nutrition and resources because you'll get a bigger guy that about win the jousting competition and the next generation will have larger people. life on earth is not fair. you might select for being big but nutrients, there might be a

number generation, you'll be selected for small and big and small. play out the movie and see what happens. what would happen is you would be selecting for big, big, big, bigger, bigger, bigger. now you select for small.

too bad everybody died here and now they are little. and now they get big again. and you get these huge changes in population. we didn't know that there haven't been these large diops over development, the happens in an existing

population. what about models? now you select for the variability itself. it's just the variants, really. you have genetic sites that are selected for variants, not just for meades. it's going to be like it was

before and you'll have a huge chain, after enough fluctuation back and forth like this, eventually look what happens to the variants of genetic selection. you wind up with a population that is extremely heterogeneous that includes some people who

are big and some people wh withgenes for being small and small people with genes for being big, that makes is understand in genetic data. how do you test this idea? the prediction of the idea is that there would be a thing, a variable methylated region. this is a charm slot more moth

methylation, less methylation. they do exist. you do one of these annotation analysis and find out that they are involved in anterior/posterior pattern formation, development. central nervous system, gut, i've never seen a chart like

that with so many important processes that are all showing up for the same phenomenon. so how might this be relevant to disease in the first disease that seems to make sense is cancer. cancer involves repeated changes in the microenvironment.

in a way, a cancer cell surviving the colon, selecting -- being selected for being able to live in that space that it's innovating, and selecting for being able to travel through the blood and the lung and the liver, subjected to

huge varying environmental changes. like oxygen being one of the principal ones, very hypoxic, it becomes i'm of hyperoxic. we did a study to see whether or not sequences are changing more by variants, the level of dna methylation.

this is a principal component slot. it indicates by how close you are, how similar you are to other sampels, measuring across normal tissues cluster together, breast, thyroid, lung, if epigenetics where you have a defined change that goes in a

precise place to normal to cancer, from the normal lung, say, to new pattern of lung cancer but that's never what you see. what you see is there. you get a huge variability in the methylation pattern at the very site defined normal

tissues. it's like a normal group of what i was showing earlier, visually in that first heat map i showed at the beginning of the lecture. and with as few as 25 cpg's, using this method, you can distinguish cancer. when we continue this work, and

then take it to the level of sequencing, we got a new insight into what i discovered with birds in 1982 in this "nature" paper. there's a picture of us back then. here is a interesting interesting thing for young

people in the audience. when i look at myself in the mirror not morning, that's what i think i see. it's really interesting. so this is what we saw. this is sequencing now. this is a million bases dna, highly methylated, low

methylation, this is normal colon that's blue, coal an coloncancer is red. this is hyper variable too. and not only is it hyper available, these are individual genes, there's normal, here is the variability, even when genes are turned off sometimes, most

of the genes involved in invasion metastasis, it's a half of a genome in these big blocks and represents a third. this is a lot of stuff in a genome. these are giant dmr's from the nomenclature i gave earlier. what are are these big blocks?

they turned out to be something else that we've been studying earlier, in this paper, from the lab, we describe the large organized chromatin, di methylation, trimethylation. these are normally -- they have the increase chromatin methylation and correspond with

the region identified associated with a nuclear laminate, they are highly corresponding, that's what this represents as well. importantly, we showed they are not present -- they develop in differentiation and there are tissue-specific differentiations in them.

i had a postdoc in the lab, an accidental discover, oliver macdonald, a pathologist, i suggested he look at the -- he looked to see the response, emt, epithelial transition, induced involving changes of dna there was no evidence of this at all.

oliver, of course, stained everything, being a pathologist. he said, andy, by the way, look how pale the nuclei are on exposure. i knew he would know that, he knows everything. you could see the tgf data reduces the methylation

globally. you can see the lock specific here. there they are, visualized by em. and they disappear on this exposure and come back when you take this off. so it's something that's normal

reprogrammable in response to tgf data. i suggested a dynamic landscape different than what waddington said. it's not totally gene determined but it's adaptable. these valleys can become much flatter and allow transition

from one to another, and that is incruised by chromatin structure, based on things that other people have done, that i pointed out in the review paper, chromosome-chromosome interacting as well. there's plasticity that's important in normal development.

that's what i was saying earlier, you need a mechanism of normal development and response to injury and challenges for adapting your program, you could do this by altering up alteringepigenetic plasticity. there's a subsequent paper in "nature," it's not that hard to

model how this might occur, flattening the valley and doing things like -- i'm not going into the names. the essential idea, even if you had, say, a deterministic element pulling things back to a normal level of methylation, say like a hooking and spring kind

of very simple process, you could add to it a variable factor and that if that was relaxed, you would get this variability and if you model that, the data that you get in terms of distribution of dna methylation between normal and cancer, the reality of what you

this is data i think are published. i saw it a couple days ago in molecular medicine, this is also from winston templar. i want to show you this. it's amazing. we saw this change in dna methylation and use large scale

methylation, but this shows, and this is an example of breast cancer, here is the region of islands, and the nearby regions that we call shores, we define the hypo methylated shores in cancer and hyper methylated islands. in the cancer, if you compare,

say, the breast cancer outside the these block agents, there's high methylation, you can go to the island, it's low and back up normally. inside the block, these are islands inside the block, you get this hyper methylation of the island, hypo methylation of

the shore but nothing much happens outside the block. even the changes locally that people looked a, cpg islands, variable methylation seems to be able to be controlled by the chromosomal mechanisms. the blocks are kind of are

eating the islands and shores, at least mathematically. there's some data that i'll show you that's unpublished, that we've been looking at aging, and this is data from amy vanderber, from hopkins, and are the methylated blocks, we see them in nonmalignant tissue in

photo-aged older people, not really without the combination of both of those things. we think of the environment inducing these changes over time. and there's work -- this is like my most favorite paper to show you, because it isn't from me,

but it supports the ideas. this is work done published in genome medicine in 2012 showing that if you take biopsy specimens who women who later developed cervical cancer or didn't, why would you have that? because it's a real conundrum, even with the vaccines and

everything following me, making histologic diagnosis. women did biopsies, most don't have cancer when you do the biopsy but some develop cancer. you can develop who will develop cancer later using variability testing, you couldn't do that with methylation.

i think this occurs early, it's predictive of cancer development. we have an idea about that. the idea is this. you've all heard about the hallmarks of cancer, i'm not question their existence. but i'm suggesting that what

really might be happening is the center of this is the epigenome that needs to be stable for a given tissue, if something derails that tissue, and leads to increased plasticity, it will allow natural selection of those cells at the expense of the host.

and that model will only matter and take place if there's repeated changes to the environment, cancer arises from depleted cycles of injury and repair, many times are known to occur specifically because of that, cirrhosis of the liver, even colon cancer.

we know there are primary changes in the up epigenetic i would argue they are affecting the epigenome specifically. it's an evolutionary process like the one i was describing earlier. let me talk about other diseases in the final ten minutes, this

idea of combining studies to look at commonages with gwas. common genes. this is what was done. the end result was we identified a reason that changes dna methylation nearby the hla cluster with defined changes in amino acid sequence, we think it

has to do with expression of hla family of genes. this is validation of result by independent methods on independent samples, but here is what we really saw. now here is a density plot of the methylation for a genotype that does not lead to rheumatoid

arthritis, and here it does, here is for the heterozygous. it's increasing variability of methylation in that region and these individuals are well afterwards, and these individuals are sick. if you don't did the gwac, you don't see this.

it's below the radar screen. when you combine it with methylation, you can pick it up. this is a recent study with the help of the dlsa led by luigi to look at how -- what is the normal relationship across the genome that might be regulating dna mettlation methylation.

methylation association is very small distance but what's surprising about that is you can have snips down here that are regulating clusters of cg far apart, these are from several different regions, yet they follow in the population a similar pattern under the

control of the snips. it's because these are regions that have associated with each other in the not-my-daughter packing method. they don't necessarily lie within the same ld blocks. that means that the integration of epi genetic information could

reveal patterns of human disease and a supplementary figure shows how that might be. here are snips associated with metabolic genotypes but what they control in methylation is a region, we call it a gene, just because we like bad names for things, but here it is and it's

right in the region. i want to share a story we're about to send in a journal. this is something we have to address. that's a 100-year-old woman who looks exactly like my mother.

i pointed this out to her, she didn't talk to me for about a week. hamburgers, stuff that inflames the colon. by the way, when i see these pictures, i want to have a hamburger and a cocktail. oh well.

it's very difficult to study the environment. epidemiologic studies have expensive. they have tough to do. the real issue, you can't control the environment in human populations. there's no way, it's ridiculous

to think about doing it. how often are you going to really study that? a lot of my lab is now interested in figuring out the new model of the house. mouse. we took mice and give them a normal diet or high fat diet,

work of a graduate student, michael moltoff, andrew dockery. we did charm analysis. it's an inexpensive way for getting a lot of data from these animals. we asked what does the high fat do versus low fat in terms of dna methylation and diabetes and

insulin tolerance. we find lots of these dmr's related to high fat or low fat diet or one of these phenotypes. we found 625 in a population. then we said, all right, of these 625, 491 map onto our trauma in humans. of those, how many are

conserved? 80% are conserved from mouse to human, really astonishing. and then we found methylation changes using tissues that we got from our colleague, obese versus lean individuals, or people who were obese and treated by ga gastric bypass,249

showed methylation changes, even though they are so far apart, the mouse and human. we mapped them for diabetes. this is largely pancreatic, this is largely adipose. it's amazing we find any overlap. 30 significantly overlapping the

region, by permutation analysis. this is a dramatic example. here is the mouse data. here is the exact same region where the obese individual has the same difference and it goes halfway back, after bariatric surgery, just to show you that. so this is the result.

and what i've shown here is an ingenuity spot where i've shown our 30 genes and a couple that are pale to show a connection between them. i first thing i want to point out, two of these that we found using nothing to do with genetic analysis turned out to be top

gwa, we have to be on to something. those are shown in blue, two genes here. we found there were the other 28 or so were overlapped gwa that had not risen to genome significance by gwa but they are so close in pathway, here to

here, these to here and so forth. and the other ones shown in purple are conserved mouse gene and dmr, showing changes in the same direction. the plot itself suggests that there's something real about this mechanism, about this

process, and the ability to go from mouse to human that might tell us something that we didn't know about diabetes or obesity associatediabetes associatedphenotype. we took six for which there was not a paper in the literature suggesting a connection to fat metabolism, diabetes or obesity.

here is the data for two, the control, overexpression. five of the six showed they really are involving metabolism this way. i think this might be a way to uncover new targets that are if not the primary generators of the phenotype, they might be

druggable, might be able to come up with compounds to modify the phenotype. in summary, i think that epigenetics does tie into environment of genetic disease. i forget to say this until now. please do not confuse my enthusiasm for the ideas of

certainty that they are right. i don't possess such certainty. i just get excited about this stuff. epigenetics clearly drives not just cancer but has a role in common disease generally, that's emerging to be a true story. i didn't mention but in oliver's

study we were able to reverse it. some of the drugs currently used for cancer therapy might also have an effect on this as well. they won't come up in the usual toxicity but they are beneficial and the epigenetic epidemiology could offer faster cheaper

results in medicine, combined with animal studies i think are applausible way ta plausible wayto go. what this means is these are regulatory pathways. we're not selected to have disease, right? these are regulatory pathways corrupted by the things we eat

or by longevity but they are important functional and have been around a long time, at least 50 million years. thank you very much. i appreciate it. [applause] >> thank you for a wonderful romp through a lot of ideas, and

it's time for people to pose questions. there are microphones in the aisles, i'll ask you to use those because people watching on the web have a chance to hear the questions. please don't be shy. while they are thinking, andy,

in terms of what you're seeing in terms of this variability of methylation, quite striking when you look even at inbred mouse strains, so it's hard to blame that on dna sequence, what happens if you look, assuming you can, at single cells within a given animal?

is there as much variability between cells as between animals? >> we -- okay. so we don't know, because the technology is just coming along, we're working hard out of a lot of other labs, working hard. we think it would be probably,

at least for cancer, in the earliest stages, adenomas. with earlier work looking for conventional genetics changes, you see the emergence of lots of and then you get selection and then almost a homogenousization. there's a lot of advances in variability, even in the early

primaries. data suggests something like that would be going on at least in tumors but the really interesting question is how much is this variability present from cell to cell and i think that's something you have to answer. it's a great question.

>> thanks, andy, for the very interesting. some of the liquids, malignancies like aml show nothing on fish. there's the normal genotypes, normal sequence. does epigenetic theory apply to that, this role?

>> it might. we looked hard at that. in doing research on aml, what we're looking to see is whether there are defining signatures for the genotypes for survival. we think so, in fact. but this explosion of a high degree of plasticity and so

forth, we find very little evidence of that in aml. if there, it might be confounded by the large number of mutations that are present in mutations with the disease. as much as i would like to say liquid tumors are like solid tumors, data doesn't support

that as well. >> you spoke about pre-nature al plastic changes prior to morphologic changes. can you comment? >> i can't. it's not my study. that's why i liked it so much. i don't mean to make light of

the question. right off the top of my head, i don't remember that well. andrew testendorf, one paper on cervical cancer, one on breast cancer, genome medicine or genome biology. he was a group in singapore, the statistician of record for the

women's gynecological cancer registry. they had all these samples, doing originalally 27k arrays, recently 457 k array. the reason i really loved the paper is because, i mean, he read our first paper on the idea from westminster abbey, and the

suggestion, he cites that paper, that was generous of him, and said that he's going to develop a statistical tool for looking at the variability, and both in our nature genetic paper and his, we had to borrow a tool. this don't use this mathematical analysis, we use something

called the -- oh, i'm going to regret it for the rest of my life, i'm blocking on the name. the levine test. i did remember it. and he used his own testing, eborra, i don't know what it stands for, measures of hetero-- it's something the wall street

guys use to make all that money that we don't get. [ laughter ] >> andy, i asked you this question two years ago, i'm going to ask it. >> okay. >> if you compare a young individual and an old

individual, what do you think is going to happen? do you fine increased variability or not? >> yeah, so, well, we don't know. you know? that's the great experiment. here is why we don't know.

we don't know because there's an issue of cell types. so your cell type changes as you get older. the samples that we've been examining so far are from blood. there are still different cell types. until someone can do a study

which i think is a fantastic thing to do, i know you're doing it, to isolate purified cell populations, from individuals of different -- different individuals over time, you can't really answer the question. so i think it will change. my guess is that you would

have -- i'm just guessing, you would have some of this is plasticity that might account for the reduced adaptability of aging distress. a lot of aging problems are related to inability to respond to rapidly changing environmentings and go back.

little just a guess, not based on fact. >> i was wondering if you would elaborate on the missing heritability comparing the role of epigenetics and interactions between genes, which is also something that is missed by a gwap study.

>> yeah, so -- well, i don't they are fantastic studies. if you could do a gwa study and control the environment on the population, then i think -- that's the problem. i think it's not that there's a flaw in gwa, it's just very hard to look at an outbred

population, which we are and hope to remain, with many different exposures which also given the -- both of our gastronomy, we hope there will be diversity in diet that will persist. it's going to be difficult to get atticly th at particularlythe

it will be difficult to power i don't -- i wouldn't use the term mis misinherittability. it's not that it's missing, there's difference things you have to do to resolve different what i'm arguing in diabetes results, some of these things will be purely epigenetic, they

are downstream of the phenotype changing which doesn't mean intervention -- still one can benefit medically. certainly the genetic manipulation experiment suggested exactly that. you can manipulate the genes, you'll get a change in

structure. there are going to be things that are going to be still on genetics, but you need to look at all of that. you know, i mean, even in toads, a genetic experiment and not -- it is epigenetic but if you look at the genetic data, they point

to variants and enhancers. that's very much part of the epigenetic program too. a lot of language sometimes gets in the way of these, of -- you know what i mean. it's like gray science, these things integrate well. my goal and all of these years

i've been doing this, it's really even though i was doing cancer genetics, but certainly since we got our site grant and this recent work, you have to look at genetics and epigenetics as integrated together. you can't think about epigenetics without genetics.

i would argue often you neatneed epigenetics to understand purely genetic detail. >> a wonderful presentation and wonderful conversation amongst all of you with our speaker. there will now be a reception in the nih library where you can continue the conversation with

andy over coffee and cookies. everybody is cordally invited to come back next week. meanwhile, let's thank our speaker one more time.