Cancer De Utero

>>> our first speaker is kurt harris, chief of the human carcinogenesis lab, and for background purposes kurt got his m.d. from kansas university school of medicine, subsequently he did residence in internal medicine at ucla, and he's received numerous awards, among

them the distinguished service medal, from the u.s. public health service. i was at his talk where he received the aacr princess takemodo award, published 500 manuscripts, chief of the journal of carcinogenesis. his talk today, integration of

cancer biomarkers in precision medicine. kurt? . >> curt? >> thank you very much. it's good to be back. what you see in front of you is a tapestry woven by claire and denise about ten years ago.

they made ten of them, about half of size of this screen. i use it as a metaphor in that i think each scientist is contributing one thread and where that thread is, how bright it is, is depending on peers and history. but you're not doing the whole

thing. we're just contributing. this is team science for collaboration kind of thing. so precision medicine, well, that's an interesting title. there are a number of things that recently joined the lexicon.

it was formulated by francis collins, the head of the nih, and that ad hoc imo, iom, institute of medicine committee, did a report on it in 2011 and i'll tell you a little bit more about that. the idea of personalized medicine has been around a long

time. this is a quote from hippocrates, it's far more important to know what sort of person the disease has than what sort of disease the perp has. interesting point. so let's talk about the traditional medicine, and

precision medicine will follow. traditional medicine, the kind that terry and i learned, was that lifestyle, medical history, family history, signs and symptoms, standard lab tests and imaging. precision medicine is multi-lay erred, individual-centric and

interconnect and combines all these various with epidemiologic and clinical information, shown perhaps better here, this is called the information commons, integrated, it takes big data biocomputer, developed a knowledge network molecular taxonomy to hopefully improve

diagnosis, health outcome and treatment and guide biomedical research that leads back into the information commons, and also guide prevention research. as i said, this is national academy of science, 2011. so now let's look at examples of this and i'll intersperse

inflammation, the main topic of this particular traco and talk about exposeosome, everything has a nice name, coined by chris wild ten years ago. this is from a cartoon that aaron and i prepared for commentary, involving the external environment, things

you're familiar with, tobacco smoke, infection, diet, so on and so forth, radiation and involves the internal related related to external, diet, obesity, chronic inflammation, and the two together lead to cancer biomarkers, and the cancer genome and epigenome to

build the driver genes that push toward cancer therapy, but the understanding of carcinogenesis can also provide mechanistic cancer biomarkers of risk in prognosis. so chemicals that cause cancer, you're quite familiar with those, going back to the chimney

sweep, aspergillis, cigarettes, they cause cancer and mutation in genes such as the p53 gene. in some cases there's a pattern and some specificity in the type of mutation that occurs. now let's go to inflammation, which is the other side of the exposome.

ininflammation and infection can increase cancer risk. it can be inherited, hemochromeatosis, crohn's disease, ulcerative colitis, and familial pancreatitis that can run in families like the jimmy carter family. more frequently, the

inflammation is due to -- the infection is acquired such as viruses, bacteria, parasitic organism. and this leads to about two million human cancers per year related to infection. this comes from the world cancer report that is produced by

w.h.o., the international agency for research on cancer, it just came out that i and many other people contributed to but doesn't take into account chemical, physical and metabolic examples. it could be acid reflux that causes barrett's esophagus and

increases risk of esophageal cancer. obesity, at many sites, and smoking contains 65 chemical carcinogens, an inflammatory agent. any of you who have taken a puff of a cigarette know it causes a marked response in terms of your

respiratory tree and chronic smoking is also associated with marked inflammation, chronic inflammation. so obesity, obesity is chronic inflammatory disease increasing risk of cardiovascular, neurological and cancer. there are two periods in which i

think some really interesting mechanistic work was done. the first period was around 2003, in which these two groups found that macrophages were infiltrated into body fat, and were activated. and then just a few years ago, it was found that adipocyte

undergo synesescence, and you have activated macrophages producing pro inflammatory cytokines in feedback loops. so that's one explanation. another explanation is, well, free radicals, which i probably won't say much about. so let's look at health

disparity in the role of pro inflammatory cytokines, measured in serum, in terms of risk factor for cancer and in this case lung cancer, diagnosis of cancer, or prognosis of cancer. so these are studies from our group primarily, in which we found that in this paper that il

8 and crp were independent co-variants in predicting cancer risk, lung cancer risk, five to eight years prior to development or diagnosis of cancer i should say. at the time of diagnosis, there was still high levels of il-8 and il 6 in afro americans and

european-americans. prognosis, there was poor survival if il-6 and il-8 were increased. tnf alpha in these two il 10 and il 12 in african-americans. let's look at il-6 and il 8. this is 100% survival, 0%

survival, five years, ten years, 15 years. this is stage 1 lung cancer. stage 1 lung cancer is the earliest stage in which there's no evidence of metastasis, margins are clear, usually small tumors the below 5 centimeters, stage 1a is below 3 centimeters.

if these in serum are low, relatively good prognosis. as you can see, 25% of these people who have surgery, curative surgery presumptively die of micro metastasis. if both are high, 50% mortality at five years. this earliest stage cancer and

the cytokines in the serum are predictive of risk and recurrence. lung cancer is the most common lethal cancer in the world. this is the surgeon general report from 2014. again, lots of us contributed to that.

and so it's 50 years, they were asking the question how many premature deaths were there over those 50 years. and you would be astonished at the answer, over 20 million people died premature deaths due to smoking over a 50-year period, not only cancer but

cardiovascular disease and others. for lung cancer, this is the plot over the last ten years or so. in males it's going down. in females, we hope it's plateaued off or is going down. but it's still a very large

number of people who die of lung cancer due to smoking in the united states. now, a number of years ago, in fact a long time ago, there was a meeting in cape sunion in host doctors on cancer. and one of the people who presented his results for the

very first time was professor hirayama. and he made the first observation that secondhand smoke, passive smoke, increased the risk of lung cancer. at this time, very few people believed it. and but since then, it's quite

obvious, 50 or 100 -- between 50 and 100 papers now showing an increased risk of people who are around cigarette smoke and develop -- have increased risk of lung cancer. this is another quotation from hippocrates, this one i particularly like.

they used flowerily language. some men have constitutions that are like wooded mountains running with springs, others like those with poor soil and little water, still others like land rich in pastures and marshes, and yet others like the bare dry earth of the plain.

all right. so when i was at this meeting, in greece, it occurred to me after listening to hirayama's lecture that infants, and maybe in utero might have increased risk of developing lung cancer. that was 1981. we couldn't do this study until

2009, an that's susan, now at ohio state. why couldn't we do it earlier? couldn't do it earlier because women didn't smart smoking with frequency until the second world war. we needed enough time for infants to grow up to be 50, 60

years of age, so they would get lung cancer, if they were going to have lung cancer. and then lastly, to only look at never smokers, it's 10 to 15% of people coming into the clinic today. susan tested the hypothesis childhood exposure and genetic

alterations increased lung cancer risk in never-smoking adults. the conclusion was parental secondhand smoke exposure from the father or the mother, during childhood, is associated with dose dependent increase in lung cancer risk in never-smokers in

two cohorts. this is especially true with those who have the haplotype that leads to high levels of protein involved in innate immunity and complement pathway, so these people normally have -- it's about a third of the population, that includes a

third of us in this room, have this haplotype, the hyper reactive innate immune system. a startling finding was the early age of onset of cancer in the never smokers. the once that were exposed by their parents, they developed lung cancer in their 50s.

the ones that weren't exposed but were exposed to secondhand smoke elsewhere or never smoked developed lung cancer 10 years later in the mid-60s. genome, going down the list here. so lung cancer again the traditional view many years ago

was you had small cells, lung cancer, squamos cell, adenocarcinoma. 1987, k-ras was discovered. in 2004 egfr was discovered. they were almost exclusively in never smokers. in 2014, many more quite rare mutations were found primarily

again in never smokers, and egfr is here and k-ras. what isn't here is p53. a 53, tumor suppressor, is mutated in maybe 60 to 70% of lung cancers depending request it's squamos cell, which is higher, or adeno, which is lower.

epigenome, there are many aspects. it with be demethylation, micrornas or noncoding rnas, chromatin remodeling, there's many possibilities, but i'm going to focus onmicrornas and the interaction between

micrornas and the inflammation. these are small microrna, evolutionarily conserved. the person that gets most credit is victor ambros who got $3 million at the breakthrough meeting with the hollywood people in los angeles just

recently. so he has a bigger smile now. and gary was a colleague at the same institution who actually helped identify targets of microrna. these studies were done in elegance in the early 1990s. in which they discovered

micrornas, but micrornas are evolutionarily very old. you can find them in plants. the other interesting thing is that their targets are multiple. like hundreds of targets. so when we do experiments in the laboratory, we do everything possible to have no off-target

effects. but micrornas, that's how they function. they function by affecting many targets in many pathways, and sometimes in the same pathway. so it's evolutionary conserved to work that way. so ten years later, this guy

carlo croce, a colleague, made a seminal observation of two micrornas that were lost, 15 and 16, that was the first paper. and since then, there are about 10,000 papers on cancer and micrornas, an equal number in other disease states.

so he really started the interest and found that micrornas are differently expressed in human cancers and we and others have been asking questions about risk, diagnosis, prognosis and therapeutic outcome and interaction with inflammation and other risk

factors. so they have many targets. they have primarily two functions. one is they bind to messages, and these messages are inhibited in terms of their translation. or they bind to the message and cause instability of message

rna. they also bind to proteins including rnps, rna inhibitory proteins, and inactivate them. and then lastly, and i'll return to this at the end, they can act as liganding. particularly mir 21 can bind to toll receptors in humans and

mice and activate nf-kappab and increase il-6, dnf alpha and inflammatory response associated with nf-kappab inflammation. so we got interested in this in lung cancer, and this is in nozumu, who asked the question about diagnosis and prognosis and found microrna profiles

were different between primary lung cancer and corresponding so you can distinguish adeno, squamos, small cell. to go back to stage 1 lung cancer, increased mir-21, 511 and 10 act as oncogenes, and decrease in let-7, a tumor suppressant gene, confirming

previous results. now, we also collaborated with carlo croce to ask which are the micrornas that were known. there were 280 known, now there's close to 2000. at that time there was 280. we asked the question in breast,

colon, lung, prostate, pancreas, stomach, which of the micrornas was most highly upregulated. one of these came to the top of the list, mir-21. i mentioned it in the last slide. since this initial paper was

published, it is upregulated in 18 major cancer types, a biomarker of poor survival in 14. this is one of the things we and a lot of other people do is look at multiple cohorts, these are 100% survival, 0%, this is time and months, high levels of

mir-21 in each of these populations is associated with poor prognosis. as i said, there were 14 different answer types that are associated -- mir-21 levels are associated with prognosis, these are 10, we studies lung but pancreas goes on and on.

each has been validated by several other studies. this is quite a common finding. and it's true in colon cancer, and these are two populations that aaron in our group studied. a maryland cohort and hong kong cohort. there have been three other

studies that have valentine date dated those results since then. how about therapy? this is charlie heidelberger who died about 20 years ago. he discovered it 50 years ago or more, still being used in the clinic today for a variety of cancer types.

and mir-21 high levels, poor prognosis, for colon cancer in japan and german cohorts, and it's confirmation of a study in a previous slide in hong kong and the u.s. four different cohorts that have found the same thing. so before going to the

combination of micrornas and inflammation, there is a lot of work that's been done by many people, in terms of providing a mechanistic underpinning of mir-21 in human cancer. mir-21 is a noncoding gene, that can be amplified to lead to increase, controlled in part by

dna methylation, so a decrease in dna methylation leads to increase. we show vegr and k-ras in the same pathway increase mir-21. david found that several of the inflammatory cytokines, including ih-6, stat3 is a promoter of mir-21, this leads

to phosphorylation of stat3, active at promotes lots of genes including mir-21, in collaboration we did a study of stress, that increases mir-21. what's downstream from mir-21? there's many, many downstream targets. the one listed on this slide,

ones validated in the laboratory. they are not just algorithm predicted targets. these are real targets. and these are selected on the basis that they are involved in cancer in various ways. high levels of mir-21 actually

decrease the protein level, of all of these different gene products. and as i mentioned, mir-21 activates inflammatory pathway, and cachexia. this is a cartoon that was prepared, it makes several points.

one is that cancers can produce micrornas, and package them in micro vesicles, they call an exosome, they can transfer the micrornas to immune cells such as macrophages and toll receptor leads to increase in nf-kappab, producing these cytokines that can increase

growth and metastatic potential. they can also be transferred to muscle cells and activate mir-21, that can increase nf-kappab and cause apoptosis of muscle cells. so why is that of any curiosity? well, those of you who may be associated with cancer care know

that for certain cancer types, such as pancreas and sometimes lung cancer, there is a syndrome called cachexia, and it's essentially wasting of muscle and fat, so these people die quite a terrible death, and there's nothing so far you can do for them.

so far. and it looks like one of these micrornas might be the cause of this, and that antimir-s against this might have some therapeutic value. we'll see. so let's go back to colon cancer and jane and aaron did these

studies. the hypothesis was that inflammatory risk we looked at 41 inflammatory-related genes, would correlate with colon cancer specific mortality independent of tumor stage, there's contribution by

cytokines and inflammatory protein in the adjacent colon as well as in the cancer. so adjacent colon, these were increased, including il10, associated with normal metastasis. it's an antiinflammatory cytokine.

in colon cancer these were increased, including il-23, which increases the proliferation of th-17 cell and production of il-17. and so il-17 frequently is associated with increased cancer progression. what happens if you combine the

inflammatory biomarkers in mir-21? is it a better prognostic classifier of either one than the other? so inflammatory risk, the cytokine genes that i just mentioned, kaplan meyer plots 100% survival, 0%, and

median, colon cancer specific, 30 months. the combination gave interesting results. if they were both low, good prognosis. if both were high poor prognosis, median was 17 months.

this is asking the question, is each co-variant independent the other co-variant? these are hazard ratios of mir-21, independent of it, and also they are both independent of tumor stage. so what might be the principle for this?

this is what we proposed, that each kind of biomarker, for example the noncoding rna or coding rna, they combine and give you this, that you've already seen, and they can both, with some degree of precision and accuracy, in agreement identify people who have cancer,

or have a good prognosis, i should say. but each of these biomarkers, this misclassified some people. so the coding rna that's classified these, noncoding rna classified these, so the two together is better than just one.

if you put three separate independent co-variant together such as dna methylation, is it better? the answer is question. if you put four in, is that better than three? we're testing that right now. in an independent study also in

breast cancer, so this is ewy and giang, who was an oxford fellow, m.d./ph.d. who just finished training at duke for medical degree. and so they looked at adenocarcinoma of the esophagus and squamos cell. for adenocarcinoma, the

inflammatory risk is different than colon cancer, and a decrease in mir-375, tumor suppressor gene, are associated. in squamos cell carcinoma of the lung, adenois primarily due to barrett's and acid reflux, squamos cell due to smoking prime army.

so different cytokines here and different micrornas but you still see a common result, combinations of independent co-variants is poorer, i should say is better than a single independent biomarker. oh, i did something bad. there we go.

so very briefly, lung cancer and combining noninflammatory genes but other related genes which i'll describe in just a moment, and the combination of mir-21. and this is stage 1 lung cancer. and the four genes, protein coding genes, xpo1, brca1 is involved in combination, and

mutations, increased risk of breast cancer, and hip-1 alpha. other metabolites are involved. this is a japanese cohort, primarily stage 1a, 3 centimeters or less. and the four-gene signature prediction is here, and mir-21

is here. if you put the two together, the combination is worse here. and in a population from the u.s. and norway, we found again these are primarily stage 1b patients, no evidence of metastasis but the tumors were 5 centimeters in size, and again

there was association with poor prognosis with the four-gene signature or high levels of mir-21, poor prognosis, intermediate prognosis if they are both high. let's say something about the micro biome because that's a hot topic right now, and there's a

very strong inflammatory aspect of this. so in our group, leigh greathouse is investigating this, asking the question about the microbes that populate our body surfaces internal and external body surfaces, and their role in disease.

and this is, as i said, quite a hot and interesting topic. it's been long known certain microbes are associated with increased cancer risk, 25% worldwide are associated with infectious agents, on the previous table that i showed. examples of specific microbes

are hpv, in head and neck cancer, liver cancer, hepatitis b virus and hepatitisc, and in gastric h-pylori. now, whenever there was a break in the mucosa barrier, due to some sort of injury, free radicals or tumor starting to arise at this plastic lesion,

the microbes that are in that body space, that interior space, whether it's the stomach, or the intestine, or the respiratory tract, they invade into the mucosa. and this leads to an inflammatory response, including interleukin 23 that i mentioned

before, interleukin 17, which i mentioned before, and stat3, that's activated. so are there communities or specific microbes that are associated with, in this case, colon cancer? and a few years ago now sear yum bacterium was found in colon

cancer but not adjacent tumor and quite prevalent in colon cancer in -- this is a tumor. and this is suggesting an early row in tumor genesis, cell leads to genesis in a colon cancer model mouse, and that it drives myeloid infiltration into

intestinal tumors, and is associated with pro inflammatory signature, both in mouse and humans. i did mention that there are ten to a hundred times more microbes that we carry around than we have actual cells. so we're talking trillions of

microbes. so micro biome and cancer, the core human microbe biome geography makes a difference, so host environment factors. diet, smoking, obesity makes a difference. various single nucleotide polymorphism and mutations make

a difference. and receptors in cytokines, and one of the more amusing things just published is kissing makes so when you kiss someone, you transfer about 8 million if it's a juicy kiss. but only about 800,000 or less if it's a very light peck.

so people who kiss frequently, say you and your spouse, develop the micro biome in the mouth in a very similar way. now, this is the very interesting cartoon, which i assume gets your attention. this is published in the "new york times" a year ago.

and a little more than a year ago. and it was about a study that was published at this, and it makes a point of fecal transfers, from one person to the other. you might say, well, this is a pretty crazy idea.

but homeopathic medicine has been doing this for a long time. and in some cases, it's being done for treatment of inflammatory bowel disease, or other kinds. so in that case, the microbes are isolated and put into capsules.

so this is not so foreign that it's actually happening and has been happening for a long time, but now on a more scientific basis. the last thing i'll talk about is the metabolome. that's a reflection of tumor metabolisms.

and majda and ewy have done these studies, discovering biomarkers associated with risk, diagnosis, prognosis and therapeutic outcome of lung cancer, a wonderful collaboration with frank, one of the world's experts on metabolomics.

this picture goes back to 1653. and this is my third quotation from hippocrates, he tasted urine in the diagnosis of disease of his patients. and this was quite common. when i went to medical school, fortunately we didn't have to taste urine.

we had other ways of doing it. but it's been a source of getting information from patients, diabetes, it goes back centuries. well, a physiologist named armstrong some time ago wrote, from a liquid window through which physicians felt they could

view the body's inner workings, urine led to the beginnings of laboratory medicine. so one of the biofluids we study is in fact urine. and we study, the metabolites 1500 molecular weight and less. and we're quite interested in lung cancer, but other cancers

too. afro-american versus european-american. and found that the metabolites in afro-americans can the different than in european-americans, but there's a prominent group, and we selected ten and asked the

question whether or not they were associated with diagnosis and these are called block curves, specificity curves, and this is the line here. but the area under the curve, this part right here, when it's about .7 to or 9 is highly significant, and so this is in a

training set, and in a test set a sizeable number of cases could throw a similar result. this is below .8, this is above. this is a strategy of taking a large cohort and dividing it in half in a random way and figuring out, more valid than that, doing multiple cohorts,

and we're involved with doing those kinds of studies too. another thing you could do is combine the microbe -- excuse me, the metabolites, and out of that 10, 4 of these were utilized, and these are kaplan meyer plots, 100%, 0%, five-year survival, these are lung

cancers. all four of them together were much more predictive than three or two or one in terms of predicting cancer prognosis. and this is the multi-variant analysis down here, ratios showing three is by far the best, or four, all four is even

better. and two of these creatinine riboside and n-acetyl neurominc, there's a positive correlation between the amount in the tumor and amount in the urine. so this biomarker was reflecting what is found in the tumor, a

tumor metabolite that you could measure in urine. let's end up with a current problem that terry and i and others who are involved in lung cancer research are facing. and that is the screening of small people, large numbers of people, ages 55 to 70, and this

has led to medicare, probably approving low dose ct for screening. and this was published at new england journal of medicine and compared to x-rays, known to be lousy, those a 20% reduced mortality with the low dose ct. however, there are lots of very

small lesions that were discovered, and the vast majority are not cancer. 96% false positive. the ones that prove to be cancer are very early stage cancers, stage 1, and stage 1a particularly. so there's going to be thousands

of people, tens of thousands, maybe hundreds of thousands of people going through this kind of screening and coming up with some lung cancer in about 4 or 5%, but many of them, we don't know what they have. why are we going to deal with that?

there's several strategies that we won't go into now, if anyone is curious about that. and i mentioned that even after curative -- presumed curative surgery, 25% of stage 1 lung cancers die of their disease. so we would be very interested in identifying those that are

low risk versus high risk, and then developing appropriate therapies, because right now you just wait till they have a recurrence. and how to do that? precision medicine. the objective is decreased false positive rate, decreased

financial cost, improved care and studies so one can take high risk individuals, put them in randomized trials with very early stage cancers, which they don't get treatment now and ask the question whether adjuvant immunotherapy or chemotherapy will have some benefit.

so this is a summary of what i've gone through in 15 minutes. i talked about precision medicine, cancer biomarkers of risk, diagnosis, prognosis and therapeutic outcome. i focused -- could have focus on many but i focused on this one.

combination of validated biomarkers, such as protein coding, inflammatory change or even noninflammatory genes, microrna, the better prognosis classifier. serum inflammatory cytokines, the micro biome may modulate carcinogenesis and could offer

as being a therapeutic target. if you have a specific microbe that is associated with cancer risk, or diagnosis, having antibiotics might have some benefits. so using a vaccine to human papillomavirus has great benefit.

using antibiotics for or heliobacter has benefits. and there's several therapies, including one that uses micrornas as a target for hepatitis c. metabolome analysis didn't help, is it a biomarker of risk? the answer is yes, and i

mentioned a really urgent clinical need for improving the diagnosis of lung cancer, and early stage prediction of those who have an early recurrence. so this goes back to the precision medicine paradigm, and this is some of the examples that i gave in last year's, i

guess this year's, extramural report focuses the attention on precision medicine, and this is the cover of it. and which i provided some of the dialogue. there was a lot of interest in the intramural and extramural community about precision

and these are my co-workers and collaborators, people who are either in the lab now or have been in the lab, and these are a list of some of the people that you have the privilege of collaborating with, and one of the things i enjoy most is interaction with postdocs and

fellows, and they teach me something every day. hopefully i teach them something. they teach me something every day. and so it's quite a joy. so thank you for your attention. [applause]

[low audio] >> well, as terry is implying, the exosomes are transport mechanisms. not only for micrornas but for messages and protein, small micro vesicle, they fuse with other cells which transfer the content.

soths a lot of interest in exosome at this point. they have specific proteins on their cell surface, so one could in fact, maybe using antibodies, but they can be transport mechanisms for therapy. so you can make micro vesicles and insert into them drugs or in

this case micrornas, and then give those micro vesicles to people. so this is a new transport mechanism for drugs and for small molecules. a number of companies are working on it right now. i was giving a talk in san diego

at the aaps, one of the federations, and there were a number of companies making antimirs, putting micrornas into vesicles, and there's even a phase 2 study ongoing with treatment of colon cancer patients with a tumor suppressor micron.

it's a brave new world out there. who is going to take some fecal transplant? any volunteers? >> thank you, curt. >> all right. >> our next talk will be given by jun s. wei, ph.d. at baylor

college of medicine, subsequently he joined the nci working with paul meltzer, in cancer genetics, currently he's working with chavat kahn. >> thank you, terry, for asking me to give this lecture. it's really my pleasure to give

you an overview how to use genomics research, to do the research, also hopefully i will show you some examples to use the genomic research in the clinical setting. oops. okay. this is the outline for today's

seminar. i'm going to give you some of the background on the success and challenges of treating pediatric cancers, because our lab is interested in the pediatric cancer research. and i'm also going to tell you the application of genomics to

discover the biomarkers, drivers and therapeutic targets in the pediatric cancer. and some concept in the genomics and this is two examples to study the level of fgfr4 in the rhabdomyosarcoma and translation of genomics and the personalled therapy, and i realize dr.

harris just gave you the lecture certainly of the precision medicine, so i think it's going to be brave on those points. hopefully we won't have too much overlap. this slide shows the pediatric cancer cure rate is improved a lot.

you can see the disease, the survival rate in the '60s, by the '90s you can see a lot of pediatric cancer is curable now. however, in the last about 20 years, the success for this is leveled off. and this is other cancer mortality rate, it keeps on

going down. this is lymphoma. recent about two decades or so, you see this mortality rate levels off. so the progress in the lab for 20 years is not the same as before.

this slide shows the same kind of things. the cure rate for the rates of disease, the cure rate is higher, but for some cancer the cure rate is the same in the '60s and '90s. and this is the same thing in the ewing's sarcoma, for some

localized disease the cure rate is fairly high, but if it's a metastatic, the cure rate is low, this is the same with other tumors, metastatic disease, cure rate is still pretty low. so that's the challenge of the pediatriccancer now. so the metastatic disease.

the cure rate is still pretty low. so we want to find what is the new target and the new therapies for this disease. by the way, those are all -- those are improved rates, basically due to the advance in chemotherapy and also the module

therapy, combination of chemotherapy and also radiation therapy, bone marrow transplants, it's very aggressive therapy. here i give you some definition of the biomarker. the biomarker is basically the characteristic that can be

objectively measured and evaluated to use for clinical use. either diagnosis, prognosis, and to -- for the pathological process, or the pharmacologic response to therapeutic intervention. another concept i need to define

is the driver. so the driver is basically the genomic alteration that causally is implicated in onco-genesis or tumor survival, positively selected, and shows a recurrent pattern within or across tumor types. this is the opposite of the

passengerrer events, which arise from the background mutation rate and do not contribute to the cancer process. so this is the driver concept. is there another concept using the target, therapeutic target? so there's two categories of therapeutic targets.

one is a molecule of protein that is differentially expressed in a tumor which can be used to home in the lethal therapy. for example, you can use congregated antibodies against the specific tumor to deliver toxic therapy. that falls in this category.

there's another category, if the target of the molecule, you can use this small molecule or some other agent to inhibit its function, or downstream molecules that can lead the tumor growth suppression, or even better growth suppression or regression.

these are two categories of in the perfect scenario is if we can find the molecule that can be used as a biomarker, very important for cancer is the driver that we can target, the therapeutic target. that's the perfect scenario. but we know life is always not

this simple. usually a biomarker sometimes can be just a surrogate. for example, this tumor has service molecule that can be very specific. for example, the cd99 or the mic gene in ewing's sarcoma can be specific to the tumor but if you

target it, it doesn't do anything. and another category is drivers may not be easy to target. for example, transcription factor, like mycn, and the pax3-foxo 1a, this is the fusion gene that's generated by the translocation of the chromosome,

and those molecules are normal transcription factors, they drive the transcription of the genes, on the other hand, the kinase, tyrosine kinase or other kinase is easier to use small molecules to inhibit this function. so this slide shows you the

central dogma. the biological information flows from genomic, dna, to rna, and translates into protein. and this line was divided area of genomic and proteomic. and you know, this is a system that's very complicated. it's not a single direction

information flow. and those loops, the protein will regulate the dna, the translation -- transcription process and translation process, and on the top of that there's a microrna, and there's -- this is a fairly new discovery, category of molecule that's very

important to regulate in the code genome, is functional. also about 80% of the genome is functional. in the past we always think it's only a very small portion of genome is useful, the rest is junk dna. but the nccor study shows 80% of

genome is transcribed, so a lot of those molecules made to regulate, when and where and how the genes are functioning in a particular cell. so this is very complicated system. and in the very early days, this is almost like 20 years ago,

that when the time we started doing this kind of study, this is the early days with the srbct marker, this is a study we did compared to nowadays standard is very small rate, only 4000 probes, one dna array, compared nowadays the whole genome array you can have tens of thousands

of probes. in this study we try to use different dna microarrays to do the diagnosis of cancer, because in this category of pediatric cancer, small the salmonella -- small round blue cell tumors, this kind of cell, okay?

with the staining, they look very alike. so at that time we say our hypothesis, can we use gene expression or array to distinguish them, okay? that's the idea. and the reason we wanted to do this is because doses are

highly -- these are all highly malignant solid tumor, coming from different tissue. but they look very similar in their histology, routine histology, but they have very different treatments and outcome. so the di

important for how to treat those and this is the study, published in nature med sip, we used a moderate array, and artificial network. this is four different cancers, ewing's sarcoma, and we detect this marker, no markers that would know specifically

expressed in particular tumor, in this case these are the molecules again, mic-2, i pointed out earlier, called cd-99, very specific expressed in the ewing's sarcoma. the red means high expression, where the green means low expression.

so you can see this particular molecule is very specifically expressed in ewing's sarcoma. this is the marker they use to diagnose ewing's in the clinic. another example is this expression of fgfr4 and 2 molecules in the rhabdomyosarcoma.

this molecule i'm going to target a little later. so from this study we identified 41 genes not previously reported to express in this category of tumors. and so we can use those markers as novel diagnostic markers, and lately we developed this

diagnostic assay with this company, althea diagnostic. and also the expression of these markers in the cancers also implicate their biology of the cancer. and we can also use this highly expressed gene, highly specifically expressed genes,

candidate therapeutic targets. so here i want to show yous one of the uses of this diagnostic assay. so at that time, there's a 5 year little girl seeking opinions from the pediatric oncology branch, this german patient from an old colleague.

she had a history of weight loss and reduced appetite, fever, abdominal pain. on examination shows there's a big left side abdominal mass. and the ct scan shows a big mass on the top of the left kidney, and also there's a mass in the insuperior vena cava.

we cannot do biopsy on this patient because the risk of bleeding. and urine catecholamine is negative, and the mibg scan is negative. so this is the other test for neuroblastoma. so that's the reason at the time

this girl was diagnosed with wilms tumor and they started chemotherapy on her. but lately, she had tumor excision, so after the histology is done, they discovered it's not a wilms tumor. instead it looks like an undifferentiated neuroblastoma,

that's the reason they contacted pob, asking for a second opinion. so we used this diagnosis assay to perform gene expression on this test sample, this is unsupervised class array, ways the biomarker was discovered from the previous study.

you can see the clustering of these samples, neuroblastoma, not a wilmt tumor. we can see it is a neuroblastoma. this is a microarray we examined, using a commercially available array. it's a much bigger array now.

and confirm the diagnosis of the neuroblastoma, and not wilms. this is a pc analysis, this is the loop, this is the test sample. it's clustered with this group but not with the other one. this is the wilms here. it's quite a distance.

this is confirmed. this is probably neuroblastoma. after the change of diagnosis the parrot was changed to high risk treatment including stem cell transplant, and this girl is doing well a year after the diagnosis. so here is the example to show

how we use genomics to discover the marker. so now i want to give you another example, how do we identify a potential therapeutic target in rhabdomyosarcoma? so this is rhabdomyosarcoma, arises from progenitor cells, the third most common in

children, it's about a 5% of all pediatric cancers. so incidence is 4.3 cases per million children, the united states is 350 new cases a year. it's a very rare disease. there's two kinds of histology peptides.

usually it's characteristic with alopecia. another is alveolar peptides. this alveola usually is lower than embryonal. the survival rate again is less than 30%. so as i showed earlier, that

we -- from this study, we discovered this fgfr4 is highly expressed in the rhabdomyosarcoma tumors. and again, what is the fgfr for? tyrosine kinase cell surface receptor, over expressed in rhabdomyosarcoma, and the later group demonstrated that this

molecule is target of pax3 forecast fusion gene. that's the reason it's very high, fusion positive, this molecule expressed during muscle development, normal muscle development. and it's induced in the regenerating of the muscle of

the injury, not expressed in mature muscles, suggesting possible role in my oh genic stem cells and possible oncogenic role in these tumors. so the specific aim for this project is first to determine if fgfr4 expression levels are associated with high stain and

survival in rhabdomyosarcoma. and also we want to sequence this molecule to identify if any mutation in this cancer. and subsequently we have this study, biological sequence of the overexpression of molecules and mutation of these molecules, what is the consequence.

and another important thing that we want to verify, if this is potential target. this is the kaplan meyer curve showing the expression of fgfr 4, this is high expression of fgfr4, and this is the low expression, you can see the signatures are different in

terms of survival. if you knock down this molecule using shrna under the doxycycline motor, you can see the knockdown of the molecule resulting in reduced growth in this model. also the knockdown of this molecule caused less metastasis

in the model also. so then we ask, is the fgfr4 mutated in rms? if it's mutated are they activating mutations? so this is a big experiment that would condense into one slide, we sequence about 100 samples. our rhabdomyosarcoma samples of

fgfr4 using the sanger sequencing, using the 1030 healthy controls. we found 7.5 have a mutation of tyrosine kinase. and we also sequenced the germline dna corresponding, we did not find the mutation in the germline of the patient.

so those are the somatic mutations. and there's two in the tyrosine kinase domain, k535 and vs about 55oe, we put these in a cell line to do the in vitro work. this can cause phosphorylation, with other ligands, it can self-activate.

and the other mutations caused faster growth in vitro and in vivo, this is the mouse study. and in addition, we are using this mouse model to see this is the metastatic model. we inject into the mouse and to observe if the lung mass is increased.

so indeed these two mutants caused highly increased metastatic nodules in the lungs in these mice. so this is a vector control in the wild-type of fgfr4. and then there's small molecules to inhibit the tyrosine kinase. so this ap24534 is developed by

this ariad pharmaceuticals. at the time, this is ro active kinase inhibitor. if you look at its chemical property, you can see this is all -- these are nanomolar inhibitor, so we thought we could use this to inhibit the function of the fgfr4.

and this is the expression of fgfr4 in multiple pediatric sarcomas. you can see this is another mass, alveolar form, also expressed in other cancers. this is the drct tumors, ewing's sarcoma. we think probably if we can

inhibit the function of this, it's not the only benefit to the patient with rhabdomyosarcoma, probably they can benefit to other cancers. and this slide shows sensitivity to this inhibitor is inverse of the fgfr4 expression. that means the higher

expression, the more sensitive the cells towards the inhibitor. so the future direction, several projects with this molecule, first we use sirna screening of multiple cell lines to identify synthetic lethality with combination of this inhibitor.

another way, we use -- we developed the antibody against this molecule. chimeric antigen receptor. we also tried to use this inhibitor to do the precision therapy for the rhabdomyosarcoma with this fgfr4 activation. either you have a mutation

activation or amplification of the molecule using this inhibitor. so in summary, fgfr4 is a diagnostic marker in the rhabdomyosarcoma, highly expressed is associated with adverse outcome. and the inhibition of the

wild-type in the reduced growth of the tumor and also reduced lung metastasis. the first report, this is the first report of the mutation in the receptor tyrosine kinase domain alpha receptor, in the rms, and 7.5% of the tumor having this kind of mutation in

this tumor. so potentially those are the patients that they are -- they could benefit in transformation of these molecules. the mutation regarding the transformation, increased both, serum free survival, and the cell harbors mutation showing

increased sensitivity to pharmacological inhibition of these molecules. other studies indicated this inhibitor of this molecule may be benefitting for the patient with rhabdomyosarcoma. so i'm going to shift gears to tell you about this technology,

massively parallel sequenced, next-generation sequencing. this is the latest technology to replace microarray. so this slide shows how this technology works. here is the genomic dna, or rna, you can just use -- using chemical or mechanical

fragmentation to make it to small pieces, and are a size selection and adaptor, and then you can use that adapter to put on the sequencer to sequence all the things. and after the sequencing, you can use computer to align those sequences to the reference

genome. then you will reconstruct original molecules of the sequence. that's the general principle of the next generation of sequencing. the concept is this same as in the early days.

you just chop up the dna randomly and sequence each pieces and put them together. so this is platform of next generation sequencing out there. so what is the computation and natural selection? so nowadays i think most of the people, it's a favor of illumina

platform, high seq 3500 for the next generation sequencing experiments. and also labtech has another technology called the semi conductor sequencing, so they don't need complicated optics or dyes to sequence those molecules.

instead, they use this semi conductor wafer that separates the molecule, these two are the most common platforms. those platforms now shrinking the sequence time into a very short time. the original first human genome sequencing human project spent

13 years, one genome, currently if you use this kind of technology you can sequence a whole genome for $2000 now, and two days you can finish a whole genome, the projection is cost and time will be reduced to about $100 a genome. in less than an hour, these are

powerful technologies. and these slides show you image from the illumina sequencer, each dot is equivalent to one piece of sanger sequencing. this is a very tiny field of the whole slide. you can imagine on these slides, you can sequence billions and

billions of little dna fragments, so the throughput is very high. what kind of information do you get from the sequencing experiment? this slide shows you can detect a single point of mutation, after you align it with the

reference, if you don't have anything aligned this is the detail. the other, you can have the homologous solution, so reduced coverage of your sequences in your sample. or you can detect the dnas, and in this case you can detect

abnormal junction, which in the normal genome you won't have a piece of dna coming from the chromosome but if you have a translocation happening, you detect abnormal junctions in a week, and another application is you can discover nonhuman sequences that indicate a

pathogen, maybe a viral sequence. and this is rna sequencing. you can get all this information. you can get the gene expression information, just like also you can know what is the alternative splicing events,

what kind of transfer in your specific. and you can detect fusion transcripts, also rna editing bank. and novel transcript, and the current hot area in noncoding rna, the rna that has a very important regulatory function

genome. so the power of this technology is you have all these different samples, okay? whether you can use the same platform to do all the experiments, i gave you all the information i just mentioned. so one platform you can get all

different kind of explanations with different samples. that's the power of this technology. and here i want to use the very first sequencing study to introduce you, how do we use this next generation sequencing to identify novel targets in a

patient? so this is a patient of high risk neuroblastoma, and when she was diagnosed, she was already 18 -- 19 years old. so this is very not typical usually the patient is diagnosed with cancer within the first five years of their life.

very young kid, or even infant born with this. but this patient is 19 years old. and the diagnosis, we acquire bone marrow biopsy, that's called the met1 bone marrow of the after four months, very intense therapy, usually very

toxic, okay? they use multiple chemo agents to try to cure as many as possible. and but unfortunately she didn't respond to the treatment well. so they had to take out the tumor from they are adrenal glands because the tumor grew

big. it starts to compress her vital organs. this is a second piece of tumors we got from this patient. and she went on with three years more therapy, in a very toxic therapy, all different kinds of therapy.

eventually she died of disease. at autopsy we acquired liver mass from this patient. so with this patient, we have -- we will have tumor patients from different stages of her disease. so first the question we want to ask at that time is can we use next generation sequencing

technology to identify a therapeutic target in a patient? another question is was this kind of sample, we probably can see if the tumor will be the same or be different. we combined the whole genome sequencing to answer the question.

here is the experiment. we used this met2, the autopsy sample, as index case and we performed the whole genome sequencing with the germline dna, with the skin biopsy from the patient, and we identified somatic variants, and then we go back to the primary tumor, this

is a tumor that's quite big, and we take a different section from different parts of the tumor to see if we can see the same kind of variant in the original primary tumor. and then we found actually we have some common variants we can identify, but interestingly we

found a lot of unique variants just for this liver mass. i would tell you the detail later. this is the overview of the whole genome sequencing of this liver mass, and this is called the circle plot, which has a different track to summarize the

whole genome sequencing information in this one. so this is the other track it notes all the somatic mutations in this tumor. and here is the copy number of the chromosome. you can see there's a lot of copy number changes in this

patient. this track shows the variant allele frequency of each variant here is the track that shows the you can see chromosome 3, the last copy, you compare to adjacent adjacent chromosome, the last copy.

this denoted normal junctions. so you can see there's the junction across the -- intrachromosome. so it's a very complicated change in the cancer genome. and also we detect the complicated rearrangement within a short piece of dna.

so this is on chromosome 4 and 13, and it's like 2 or 3 megabase, you can see the model of normal junction happening in the literature this is the type of phenomenon, shuttering of the genome, and so they have in that case -- the cancer genome tried to repair

themselves, regenerates in a defined stretch of dna. so if this shows how easy you can detect something that is in the past is very difficult to detect in tumor sample. this shows just the simple coverage of the lead from the whole genome sequencing, on this

x chromosome reagent. this has been reported as mutated, deletion mutation in 40% of adolescents. and indeed we detect we can see this is normal coverage, but in this stretch it's about 15,000 phase, you have reduced coverage.

that is implicated that there's deletion of the chromosome, which is not easy to detect in the past, with this kind of -- because of the resolution we can detect it. we designed the genomic primer to verify this position. you can see the deletion of

normal junction can be only detected in this sample, also in this primary tumor at the origin of tumor, not in normal liver or normal skin. and we sequence to verify this. so this table, this gives you a sense, each line is one somatic variant detected in this

patient, and other variants are the shared variants in other tumors, there are unique variants in the liver mass. we can see there is a shared variant. it's actually a minority. two-thirds of the variants is only unique to this liver mass,

that implicated this rapid accumulation of de novo somatic mutation. only about three years treatment, this tumor accumulated 2/3 of more patient. so we combine with the expression of the tumor to see which gene -- which variants

express at high level. here this is all the shared and this is all the unique and this indicates how much is the gene expressed in the tumor, and here we only are using the -- we can detect the variant allele fraction in this rna-seq, so how much variant

expressed in the rna samples. and then we detect this molecule is highly expressed in the tumor, also the allele is highly expressed in the tumor too. so this molecule, the receptor is on top of the pathway, this is important in neuron development, also important for

cell migration and cell differentiation and cell survival. this mutation is exactly at the second intracellular domain, predicted to be deleterious. we performed the microarray experiments to see what kind much pathway is upregulated, and

here we can see this analysis shows that direction of cell motility genes is upregulated, one hour stimulation with lpar1 ligand, as well as with the serum itself. and another category is the role pathway is upregulated, so that will make us think this mutation

probably is important in the cell -- in the regulation of cell activity. so next we look at the cell, 3t3 cell behavior in the culture. this is cell growth assay. you can see with different kind of serum, wild-type and mutant, they grow in exactly the same

fashion. there's no difference in growth, in terms of growth. if we put the cell in both chambers, which is measured, we can see immediately that mutants has elevated, either in the chemotherapy -- either in the sds or lpa.

and then this activation of this molecule, if we put lpa, which is ligand of this molecule, into the system, you can see that this role kinase activated low molecules, upregulated in as short as five minutes. so that means the rho pathway is activated by this mutant.

it's much more than this wild-type. we cloned sre, which is the serum responsive element, downstream of the rho pathway, we cloned this luciferase, and you can see this response, with the dose response of fashion, if you increase the ligand, you can

see more. so from this experiment we see this variants is really signaling through the rho-rock pathway. if we use our rho kinase inhibitor, y27632, we can really stop the wound healing, so this is the media, this would be the

vehicle control, this is you can see the process. we think this migration is mutant receptor mediator migrator phenotype is the rho pathway. so again another publication with this whole genome sequencing of the high risk

neuroblastoma, in this study they also found that this rho pathway is mutated, another molecule we discovered, but it's downstream of this rho pathway, so those kind much mutations will result in rho signaling, which tips the balance of the rho signaling, will cause the

collapse of the neurons, which is prevented differentiation of the neurons. so this study is really support what we found in our study. so in summary i just show you the whole genome sequencing, we showed a massive chromosomal walt ration together with a

small set of so matically acquired liver mass, and the examination of this meta revealed rapid accumulation of de novo mutation during the therapy. and the parallel whole genome sequencing, the transcriptome sequencing identified cell

motility driver mutation in the lpar1 gene, suggesting such combination of this approach will be powerful to use the precision therapy to identify identify the driver mutations. and this study also shows the rho pathway activation may play in an important role in high

risk neuroblastoma. so where we go with this kind of study is to really using the genomic tools. this is the future of pediatric clinical trials. so we can, you know, as i said earlier, the problem of the pediatric cancer is this

metastatic disease, and now we can use this genomic biomarkers to tell which are the patients that it has a good signature to responsive to the standard and for those poorer signature type of patients, we can use the technology to identify their driver mutations and then put

them into the targeted individualized therapies. hopefully this approach we can improve the therapy. so those genomic, including amplification, translocation, overexpression. these two are currently active, the training protocols in our

branch. this is the comprehensive analysis of pediatric solid tumor, establishment of repository for related biological studies. the idea is we collect other patient materials found in the patient in the clinic, so that

we can first using the tumor bank, we can study what is the genomic chance for the treatment of patients. and these are the schematic things. we take the blood, frozen tumors, and other specimens, archive in the freezer for

research purposes, and found with blood we can extract germline dna and fresh tumors we extract dna, rna, perform perform next generation sequencing to identify the target. and this slide shows that currently, the ccr leadership, the branch, views this clinomic

core to use the sequencing technology, using -- to provide information in the clinic. so if the patient referred to the ccr in the protocol, we can obtain tissues from the patient and we can acquire the germline data, somatic data, also the research data, and all this data

will be kept in a database, so those data can provide the clinician with the information, they can use in the clinical to perform precision therapy, and at the same time we can use the database to do the genomic research in our research area. so this is ongoing project.

we just started from this summer of this year, so hopefully we can have the first sequencing out of this core by early next year. so in conclusion, the integrated analysis of cancer genome will identify the biologically relevant diagnostic, prognostic

biomarkers and novel targets for and these powerful emerging cools of next generation sequencing will determine the completed genomic portrait of the cancers at the base level resolution within the next 2 to 5 years. there's so many studies now

coming out with this -- using this technology, so our knowledge of the cancers is really rapidly improved. so this will lead to that identification of the key drivers, enabling development of future novel therapy. so this is a slide

acknowledging. this is -- because of this complexity of the system, also the technology, and truly a team science. this is all sections with a big group of biologists, biologists perform the well lab experiment, the mutations provided by

support, and also the fgfr4 study by jen and sylvia helps us with lpr-1, and steve and marshal are also contributing to the lpr-1 study. so i will stop here and see if you have any questions. >> yes, yes. other experiments show it drives

faster growth. >> not yet. currently this drug is in the clinical -- i think it's a phase 2 trial for some at least. but it's on the way. >> yeah. [low audio] >> well, that's a little bit

tricky number to give, because in the transcriptome it's different, because you have genes very highly expressed, and genes at a low level. if you want to use this technology for low level express, you need to sequence a lot, okay?

and then your sequence, most of the sequence is the highly -- >> yes, so it depends what kind of question you ask. yeah, usually. so let me give you some numbers probably to make you a more clear idea. currently we sequence about five

to ten gigabyte sequencing through the rna sequencing. and we have enough coverage for expression and detection of the fusion. okay? and detection of the transcript. and the ways highly expressed mutation, five to ten gig

sequencing, it's okay, usually okay to detect. but again if it's slowly express the gene, then it's hard. for example transcription factor, it doesn't express at a very high level. but they are very important, right?

so in that case, usually what we do is we look at the dna size, okay, we do the whole exome sequencing on the dna, if we detect the mutation we go through the rna and say we see this variant. yeah. what do you mean, genome

sequence? just the dna. >> usually if you want to call -- confidently call mutation, at least you need to have 30 to 40 x coverage, okay? in cancer, it's even worse, because cancer you have the hemipoietic.

sometimes you only have one sometimes you have seven copies, you know, then your mutation is only 1/3 or 1/4 of your whole genomic dna. so that's the difficulty is the sequencing of the cancer gene. so it all depends what kind of question you ask.

so in the field now, people think if you sequence the kind of genome, you need to get at least 100x coverage to detect the mutation. so you lower. >> yeah, exactly. that's another problem, contamination, right?

that's another thing. so it's all this balance you have to trade off. thank you very much.