Cancer Pancreas Symptoms

[ silence ] >> and our third and final speaker for thismorning's session is eric collissen, who i think qualifies as a homegrown physician scientist.and, in the clinical realm, eric is a gastrointestinal oncologist focusing mostly on pancreatic cancer,but it's pretty clear he must have had a great experience training in martin's lab, becausehe is still extremely interested in lung cancer. and, i think that's what he's going to talkto us about today. thank you. [ pause ] >> okay. i wanted to thank martin and theother organizers for inviting me. it's really a great group of people and an honor to bea part of it. i got the invitation a little

while ago that this session was going to focuson nf1, which hasn't really so far. but, my talk today [laughter] is going to touch onsome, i think, evolving concepts in solid tumors about the role the nf1 gene plays,especially as we sequence more tumors, we accidentally or serendipitously discover mutationsin nf1, and we try to interpret those through the lens of human solid tumor biology andmouse models. to today, i'm going to talk to you about our efforts in lung adenocarcinomathrough the cancer genome atlas to discover new driver genes, using over 400 human exomesderived from primary lung cancers. i'm going to talk about some functional experimentsin our lab, dealing with double digits of mice, but not quite hundreds, and then finishwith a story about a single patient of mine

that might summarize my thoughts on wherewe are with nf1 and solid tumors. so, for me, unlike these kind of single hit leukemiasand other genes, and other diseases, the central question that i'll address is, does nf1 lossby mutation or copy number phenocopy, kras, or ras activation in solid tumors, particularlyin lung and pancreatic cancer, where we know a lot about kras and its ability to initiate,propagate and maintain those tumors. or the genomicist in me has to acknowledge the nowalternative hypothesis is that this is a big gene, nf1, and lung cancers and other solidtumors have a lot of mutations. so, if you're big, you're at risk of being mutated as passengerevent, and the bigger you are, the higher that risk is. so, we have to differentiatethe bystander effect of n1, which will happen

at a predictable and not rare rate, becauseit's a big gene versus the true driver ability of the event and solid tumors, which is muchless trivial than say a leukemia with one or two mutations in the cancer genome. okay,so the cancer genome atlas, tcga is probably known to all of you. it's profiled quite afew tumor types. the goal is to get to 500 of the most common deadly tumors afflictingamericans. about eight or nine projects have already reported. we've completed the lungadenocarcinoma project, and that paper will be out soon, but this a large consortium.this is how it works. a tissue sample comes in the front door. it is somewhere in bethesda,i believe, and has a pathology quality control. nucleic acids and protein are isolated, andthen standard sequences, rna and dna assays

are performed at the gnome level and wholetranscriptum level for rna and dna, and the data goes to centralized storage facilitiesthat are now available to you. cghub is available to anyone who logs on and registers. but thenit's the task of the integrative working group of which i had the honor of being a part totry to put these data together; not report a gene list, but actually tell you what itmight mean. and our group had a little practice, excuse me. we were lucky enough to work togetherin 2011 and 2012 on the squamous cell lung cancer project in which we characterized about180 squamous cell lung cancers. so, we kind of had a cleanly well-running machine, andwe hit the following lung adenocarcinoma project running. i want to reiterate a point dr. jacksonmade. lung adenocarcinoma kills a lot of people,

but what kills people is really this; advancedstage disease, with which most patients present. what we studies here was not that. we studiedexclusively early stage tumors; an important caveat to the results i'm going to give you.we feel pretty good about our quality control. we started with 290 samples that made it throughthose dna and other pathology qc's. then we had a real expert, bill travis at memorialsloan kettering and others look at every single sample. he kicked out around 60 samples dueto misclassification, something that happens quite frequently in the clinic and in tcga,leaving us with 230 free samples. the histopathologic subtypes of lung adenocarcinoma were all represented.micropapillary, i'm not going to read them to you, but the acinar and the solid subtypewere the most common in our dataset, as they

are most common in the clinic. demographicallyspeaking, excuse me, a lot of patients with lung cancer are or have been smokers, butnot all. we had a subset of never smokers in the dataset; about equally balanced betweenmen and women, and had pretty good followup, but didn't use a lot of the followup clinically.reiterating here by stage. stage iiia is resectable, stage iiib and iv is not, so the majorityare early stage resectable lung cancers. that's why they were resected, and studied in ourefforts. let's start with somatic mutations. methodologically, we strove for and achieved100x coverage of the germline, adjacent lung, or a tumor dna in each sample. we were ableto pool the exomes we generated with about 180 previously-generated exomes from the brodeinstitute to get to over 400 samples for statistical

calling of mutated genes. and, we used thesemethods to detect mutations, and then tell us their significance. not that's important.mutsigcv2 is one of these algorithms. it takes not only mutation frequency but also hot spotness,and gene size, and location in the gnome, and replication time into consideration whencalling a gene significantly mutated. not all big genes with frequent mutations aresignificantly mutated. okay, here are the mutations we found. most were already known.there's a new gene i won't talk anymore about, called mga. inactivating mutations in 10%of lung cancer, negative effecter of the mic pathway, very interesting finding. we founda new gtpa, it's called rit1, a mutant in 2% of lung cancers; i'll tell you somethingabout briefly, and recurrent mutations in

nf1 in over 10% of samples that we'll talkmore about. since we had these whole exomes on all these tumors, we were actually ableto study not only what genes were mutated, but genomoid patterns of mutation. you'llremember your biochemistry that transversion mutations are one type of mutations seen inthe dna. those are associated with tobacco smoke quite heavily. on the other hand, transitionmutations are not associated with tobacco smoke; that's what most mutations and mostcancer gnomes are. and we could measure the relative frequency of transversions to transitionsin each exome, and couple that with some of with some of the analyses. so, doing that,we see if you rank transversions from low to high, you can see a transversion low subsethere that's kind of a molecular index of tobacco

exposure. it goes from about 20 to 80%. thenonsmoking females are all down here, as you'd expect. they have different mutations thanthose in the transversion high group. in fact, of the significantly mutated genes in lungcancer, which are here, only a subset are mutated in both subtypes statistically. infact, egfr is not even a statistically significant gene in the transversion high subset. krasis not statistically significant in the transversion low. so, these different tumor demographics,different patient demographics, tell you are priority, what kind of mutations you're goingto find in that exome. pi3 kinase not even a significant gene in lung cancer overall,but when you look at this subset, it becomes significant. okay, we've seen this pie chartof lung cancer. we're chipping away at the

driver genes in this disease; egfr and krasbeing the most prominent. we wanted to contribute to that fight, and tried to do so. are sampleset was unique, because we had the unique ability to definitively call fusion transcripts,which in lung cancer are also kind of an also ran. they're clinically very important fusionsbetween ras1, alk, ret, and other kinases. are always surmised to comprise some subsetof the "oncogene negative", but since they're not measures with exome sequencing and oftenmissed, we don't know what that fraction is. we did, because we did rna sequencing on allof these, and can definitively say that all of these tumors, left to right, have an oncogene,whether it be a fusion or a mutation. that's 62% of our samples had an oncogene, and thatmakes 38% did not. we call that subpopulation

the oncogene negative subset of lung adenocarcinoma.we went along first to look at copy number. if you're red, you're oncogene positive. thisis your gistic profile, looking at statistical significance of increase copy number acrossthe gnome from chromosome 1 to 22. if you're blue, you occur exclusively in the oncogenenegative subset. so focal amplifications in met and her2 were seen rarely, but exclusivelyin adenocarcinomas that otherwise lacked activated oncogenes. that allowed us to extend the oncogenepositive subset by a little bit. and this mutual exclusivity, you can see, is extremelypowerful. that if we can do our statistics on one population without noise of competingoncogenes, our power of discover is greatly increased. okay, so what about mutations?what genes are not amplified, but actually

somatically mutated at a greater rate in thosesamples who otherwise lack an oncogene. it seemed like a worthwhile question. fisher'sexact test is a good way to do it. that's what we did, and found p53 indeed is mutatedcommonly in both, but more commonly in the oncogene negative, as is keap1, a well-knowntumor suppressor on the nerve 2 axis. but, two other events strongly stood out becauseof their biology and biochemistry. nf1 was greatly increased in the oncogene negativesubset. in rit1, although rare, it occurred exclusively in the oncogene negative subset,so two new driver events possibly on the map kinase pathway here. and there they are. so,you can see nf1 is not totally mutually exclusive, but pretty darn close, and rit1 is entirelymutually exclusive. and there are samples

over here that we could not figure out, butare still trying. what is rit1, just a slide or two on that? it's a gtpa, so it looks alot like ras, actually a lot like eras structurally. the mutations cluster and the homologous regionto q61 in the ras oncoproteins. alice berger has published a paper earlier this year onwhat these do. they transform free t3 cells with varying efficiency; the tumor-associatedmutations. they activate map kinase and pi3 kinase, and look to be bona fide a loss offunction gtpas mutations. so, at the end of the day with the exomes, we felt pretty goodabout our ability to bring a little bit more to the table in the previously oncogene negativesubset. we identified met amplification, her2 amplification, rit1 mutation, and nf1 inactivationas potential drivers in a pretty big subset

of these cancers. but, you'll notice one ofthese things is not the same here. every single one of these, as dr. mccormick has told mein his office many times, can transform 3t3 cells. these are gain of function events.nf1 loss is not. so, while i think we could plausibly understand how nf1 loss might takethe brakes off the pathway, it remains a question, i think, if that's enough to turn on the gas.to summarize the map kinase pathway alterations in the disease, we're now up to over 76% ofthe cases. we can explain some activating event on the pathway with rit1 and nf1 beingrelatively new contributions to this. we had some proteomic evidence as well with the tcga.rppa was done on all these samples through md anderson, and since the map kinase pathway,either through nf1 loss or activation down

the pathway is rich with phosphorylation events,we were able to build and then validate a phosphoproteomic signature in these tumors.and so that was using phosphomap kinase, phosphomac, etc. etc. everyone got a score, from, youknow, negative 1 to 1, and if you supervise by kras mutant versus kras wild type tumors,you indeed. there it is. do see that there's statistically more map kinase pathway proteomicactivation in the kras mutant tumors. none of the other lesions be it in raf, nf1, oregfr met this criterion of statistically increased phosphor-map kinase activity. so, i thinkthe jury's still out whether all these events are biochemically functional in the same tumors.we performed whole transcriptum profiling as others have done previously in this disease,and found the same transcriptional subtypes

that others have found. we renamed them, mostlyto be annoying and more reflective of what subtypes actually represent. and, we then,since we had exome data and transcriptional data on the same sample, which remember, ispretty rare, we were able to look at each transcriptional subtype. let's just call themred, green, and black for the genomic events that were enriched therein. so, whereas everysubtype activated the map kinase pathways in some way, they tended to do it in differentlevels, so that this tru subtype; this is the well-differentiated egfr tumors, nonsmokers,have all the egfr mutations, or at least most of them. have all the kinase fusions, andexpress a lot of ttf1, a marker of differentiation. on the other hand, this inflammatory subunitalmost has to lose p53, so instead of 50%,

it's more like 70% mutant in this cohort.and these are the guys that really lose nf1 quite frequently. finally, the last transcriptionalsubtype is in low with this stk11 keap1 tumor suppressor locus on chromosome 9, chromosome18, that is lost frequently in lung adenocarcinoma, and has a high degree of kras mutations, butthose are actually not significantly different between the subtypes. so to summarize, krasis promiscuous between the three subtypes. but, egfr, nf1, and other drivers on the pathwayare specific. so, we have new data on 230 new samples. we think it's pretty comprehensive.it's available to you. we've now identified map kinase pathway "events" in over 3/4 ofthe disease. we're going for 100, but we're not there and never will be for reason wecan talk about. and, i think pretty convincingly

show that nf1 loss is recurrent, and statisticallyenriched in samples that have no other business being a lung cancer. they have no other obviousway of activating the map kinase pathway. and i want to acknowledge everyone here withoutacknowledging anyone, because it would take forever. big group. so, let's keep our eyeson the prize and stay with the question. so, the question we started with, can nf1 lossphenocopy ras activation in solid tumors? i think, i'm trying to keep the question simple,because that helps me stay focused. and so, we went to the mouse. lucky enough to havea lab here with a little bit of mouse space. and we went into the mouse with dr. shannonand dr. mcmann and others mentoring, and basically used the kras mouse idea as a control to testthe concept of sufficiency, of nf1 loss in

adenocarcinoma initiation. this is how thehighly complex experiment was set up. we took kras mouse or nf1 mouse, and put cre in themwith adeno-cre weighted. so, that was the experimental design. the kras here acts asa positive control. to be clear, these do not have anything else. they just have nf1flox aleals [assumed spelling], either heterozygous or homozygous. so, this is a sufficiency question.we would be able; you've seen all this. this model recapitulates lots of lung adenocarcinoma.i won't belabor that. the kras mouse behaves as billed. adenomas form; adenocarcinomasform, and then for the nf1 result, i'm going to ask you to hold your applause [laughter],but no result whatsoever. so, heterozygous or homozygous deletion of nf1 at six monthsis insufficient to initiate lung carcinoma

in this model. mice don't smoke cigarettes.they're pretty young. there's lots of caveats, but in this system, nf1 is not enough. butwe went back and looked at our tumor data, and as i already told you, nf1 almost always,or very frequent co-occurs with p53. so, the next obvious experiment, and see the [inaudible]is available, breed the nf1 and p53 mice together and see what happens. nature is colludingagainst us. nf1 and p53 are unfortunately on the same chromosome; in the human and inthe mouse, pictured here in the human and here in the mouse. but crossover events dooccur. so we finally establish a crossover event. actually, dr. riley at the mci hasdone so, and shared the mice with us. we call this the p53 nf1 cyst mouse, and we're gearingup to do that experiment. have some breeders,

but don't have any results yet. but to summarizethat second part of the talk is that i think pretty definitively, nf1 is insufficient,at least in this model. i guess that means that it has to be a passenger event in thetumor, if we believe that the mouse is a perfect model of the human disease. or, it could bea modifier, and it requires other events, whether they be epigenetic or genetic to trulyto get the ras pathway cooking in an otherwise wild type tumor. so, that was where i leftthings about two months ago. i was kind of demoralized, because i thought this wouldbe a great experiment. i still do, but it didn't work out perfectly from the beginning.and, i think it got me thinking that this is probably a mutagenic event, where otherinjuries in the lung cancer genome collaborate

with nf1, but nf1 itself is not sufficient.but then i went to the clinic one day. i have a gi oncology clinic, and we get a lot ofreferrals from other places. so, these patients will often have been treated elsewhere forawhile. it's gone well, it hasn't gone well. but, we'll get a big stack of files, medicalchart to read before i even see the patient. so, i did my homework. this patient had presentedabout a year ago almost with a 20 pound weight loss, anorexia; quite a young patient here,38 years old. that's not very frequent in the pancreas cancer clinic. jaundice and anorexia,and was found to have an obstructing mass in the pancreatic head, okay? so, this isoften why people turn yellow with pancreas cancer. they occlude their biliary drainagesystem. the bile backs up. they turn yellow.

this is what his ct scan looked like withthe distended gallbladder and some biliary dilatation. so outside, he had a resection.he didn't have metastatic disease, so the surgeons thought they could cut all this out,and tried to cure him of his pancreas cancer in august 2013. he unfortunately had a positivemargin. he recurred very quickly after that with abdominal pain in the er, and the emergencyroom doctor said, i'm sorry your pancreas cancer is back, and it's widely metastatic.tough news to get. he came to us soon thereafter for management of what to do now. so now acure is no longer an option. this is palliative chemotherapy. in pancreatic cancer, that therapyis often pretty short and pretty tough, but we were going to see what we could do. i hadn'tmet the gentleman. but i did meet him, and

examined him. i have his permission and consentto show you these pictures. he had some stigmata. this is kind of a milky coffee-colored stainon his leg, and you'll notice that the lower part of the leg is missing. he had a sarcomaresected about nine years prior in his 20s. he had other findings, palmar freckling andother cafã©-au-late spots that had led other clinicians to think that he had clinicallyhad nf1 disease, although it had not been genetically confirmed. and i'm going to takea brief detour into the ras opathies of which nf1 is one. these are hereditary syndromeswhere defects in various genes, nf1 or other ras family members, including the ras proteinsthemselves are mutated in the germline, but they're not all the same phenotype. just likeour lung cancers i showed you with different

ras pathway mutations don't have the samephenotypes, these kids do not have the same phenotype either. for example, if you clinicalneurofibromatosis, you have a slightly different clinical phenotype than if you have noonansyndrome. the genes involved are different. the clinical presentation is different. predispositionto cancer is common to the two. so, whereas in non-small cell lung cancer, we had, itwas kind of a pick your, kind of dim sum multiple choice menu here of however you wanted toactivate the ras pathway seemed good enough. in pancreas cancer, that's clearly not thecase. you have to mutate ras. that's the end of the story. 95% of pancreatic adenocarcinomasharbor ras mutations. they're all a codon 12. they're mostly g12d. there's really nonegotiating around this. you have to do this

to become a pancreas cancer, or do you? so,we had a protocol on cell free dna going, and we asked the patient to participate; thinkingwell this will be easy. he's got disseminated disease. we should be able to detect the krasmutation. we did detect the 10% of a circulating dna was tumor derived. none of it harboredkras mutations. we performed target exome sequences in the following genes. we foundmutations in brca-2, inactivating mutations in cdk and 2a, which are common. stk11 mutation.we sequenced kras many times, and found no mutations in the kras gene. we found two differentmutations in the cancer gnome in the nf1 gene. this is the patient's copy number profilefor your review. there are no other high level gains that we can detect that would otherwiseexplain activation of the [inaudible] kinase

pathway, other than homozygous. where it is,there it is. loss of the nf1 gene by bilelelic [phonetic spelling], one event of which isgermline. it's a funny thing, clinically you can get a tumor exome sequence in about 21days, but it's taken over nine weeks to get one germline gene sequence, so that is stillcooking on his nf1 diagnosis. but i would be willing to be one of these is the germlinealeal. so the patient thought all this was interesting, but wanted to kind of get onwith things and get to treatment. so, around february, we considered the brca-2 mutation,and started our best first option for metastatic pancreatic cancer was folfurinox. this isa mouthful of all the chemotherapy we have in the gi suite, mixed together, and givenas first line therapy. it contains oxaliplatin,

which is a platinum, and we were hoping wassensitizing to his brca-2 mutation in his genome. it unfortunately was not. he progressedhere, as evidence by a logarithmic run. >> it's kind a myth that oxaliplatin is selectivefor brca mutation. it's assist in carboplatin. >> right. >> and oxali gives a different kind of crosslink,and they're not very sensitive. >> i think you're right. look at that! [laughter]because, it didn't seem to work here. >> can you repeat that comment, just so thatwe hear? >> yeah, the comment was that all platinumsare not equal either in their cellular penetration nor in their mode of dna damage, and oxaliplatinhas not been formally studies, as far as i

know in brca-2 mutations where a cyst, andi think carboplatin have been nr sensitizing, as i recall, as shown by dr. ashworth's work.okay, what's going on with this guy? he's in trouble, right? his ca19-9 is rising onthis therapy pretty quickly. we decide to stop that, do something else. gemabraxane,gemcitabine and now paclitaxol is the other approved therapy in pancreatic cancer, sowe try that, and that, thankfully, as of a week ago, is declining. it has him respondingclinically, but the question is what to do next, and i think we have a couple optionswith respect to the nf1 mutation. could be a mek sensitizing event, or try a parp inhibitortrial with respect to the brca-2 mutation. now so to summarize the final part, whereasi kind of gave you doom and gloom from the

mouse experiment where this wasn't sufficient,i think pretty clearly in this one patient, nf1 loss homozygous is the only map kinasepathway event we can find. we could look harder. we could do a whole genome. we're not thereyet, but i think in certain patients, this loss of function in nf1 can drive the pathway.the question is does it in most times? so, is it going to be just a question of put thetumor through the sequence or find an nf1 and off you go, or do we need a contextualfilter through which to view these events in human tumors with attention to sensitizationto particular therapies like mek inhibitors? and, that's all i have. i want to particularlyacknowledge john greer for the mouse work, anna and elizabeth for their excellent clinicalcare of the patient, and my scientific collaborators

who have helped out, and i'd be happy to takequestions if you have them. thanks. [ applause ] >> sir, that was extremely interesting. >> thanks. >> but, you're well-aware of the fact thatpatients with nf1 disease have one defect, in nf1, nf1 aleals. they don't get lung cancer,as far as i know. >> so you could have predicted in humans,initiation of lung cancer by loss of a second aleal is not a frequent event, whereas othertumors are initiated in frequent events. you've got a comment?

>> true, i do have a comment. i think, thatwas, i think the no hypothesis. they're not supposed to get pancreas cancer either. iread the chapter on nf1, and they're not supposed to get that, and they usually don't. thisguy did, and people with nf1, people, i'm sorry, i want to be exact. people with neurofibromatosismight have bad luck like everyone else, and could get pancreas cancer with a kras mutationlike everyone else with pancreas cancer, but this guy doesn't. so, i think geneticallythe onus was on us, and i think we're trying to get there, is proving that there is noother smoking gun in this one gnome; that doesn't mean every nf1 genome, but this onethat could explain his map kinaseness. because i think it's safe to assume you've got todo it to be a pancreas cancer in some way.

it's always kras, unless it's not, in whichcase it could be nf1. so, with an nf1, i think we're trying to make the case with this guy.yes. >> i have question here before you. oh, laura,yes. >> do you think that with this brca-2 mutationthat was found in the circulating dna that maybe these unfortunate circumstance, thisperson had to brca-2 germline mutation as well, and that actually caused the tumorsyou are seeing and therefore, could help in the treatment. >> yeah, the question, well, you guys heardthe question. we don't know, because the company that does the circulating dna was scared offby the patent, and do not test brca-2 for

specifically that reason, so there's yourlegal system protecting you from information, but it would be pretty unlucky to have a doublenf1 and brca-2, but it could happen, and we're checking that. that's on the germline panelthat's been cooking for nine weeks. >> one other thing, maybe looking back throughall of this, i know that for childhood cancers, there are [inaudible] i have to admit. thereare clinics now that if children have cancer, they actually are also immediately checkedfor other malformations or subtle changes that could for instance show you that somebodyis nf1. maybe that's something we should think about that for all sort of patients beforethe age of 30 that we should consider, because this person could have been helped much earlierif that would have been followed up by a much

more thorough genetic evaluation. >> i couldn't agree more. >> that's a gift. >> i mean, this patient lives in the bay areawith us, and has been treated for nine years in a bay area system, and has never been testedfor nf1 despite having an invasive sarcoma resulting in an amputation at 29, and havingthe physical stigmata i showed you. he has two children, four and five, who also havesome of the same stigmata. so, i mean, i agree that i think an emerging theme, leaving thenf1 theme is, you know, what is the job of the oncologist when the smoking gun is smokingbeyond just the patient, and i think that's

a big factor as we get these somatic sequencespanels back, i think we got a double thing. what a brca-2 mutation could mean. not everyp52 mutation is a lead ramini family, but some are, especially with the rarer ones. >> brca-2 and pancreas carcinoma is very likely,actually. >> absolutely. yes. >> so, with nf1 loss maybe just being a modifier,but you have this patient with nf1 mutation, so i was wondering, what about crossing withlike the cdk n2a loss or one of those other genetic alterations you found, and that there'smaybe some cooperation then between those two genes, and you do get a full blown cancer.

>> sure, yes. that's possible times, you know,19,900 other genes in the gnome, and i think you can make some educated guesses. p53'sour first, but, this is, excuse me, an example where maybe 400 lung cancer exomes is actuallyinsufficient. you need more like 4000 to look not for only mono-occurrence, but co-occurrenceof certain gene pairs. and, believe me; people are keeping the sequences busy looking forthat. >> just a quick comment about when you showthe overlap of the different map kinase pathway genes, it was striking that nf1 kind of turnsup as being sometimes co-mutated with another pathway member, and i just wonder if thismay be part of a process that tumors use as they accumulate enough genetic events thatthey'll tolerate a very high level of ras

signally, that they'll then lose their nf1aleal to sort of amp up the gate in the system. so, i certainly agree with you that it's probablydoing something on its own because of the mutual exclusivity most of the time. but,i'm not sure that it wouldn't be a pretty good tumor to. >> fine tune. >> to sort of fine tune higher as the mostaggressive clones start to grow out. >> and i would also agree with laura's comment.i mean, even for somebody with nf1, this guy's been very, very unlucky. and, you know thefrequency of the nf1 mutation is relatively high in the population, and the frequencyof brca-2 isn't all that low, and you just

wonder if this is that rare individual who'sgot a double hit. >> of the disease. >> yeah. and with respect to the first comment. >> please use the microphone. >> oh, i'm sorry. regarding the first commentabout fine tuning the ras pathway, i think we have to keep open mindedness about whetherthat second event or that driver event, that nf1 loss modulates could just be wild typeeffr signaling for example, or some inflammatory ongoing process that smolders, and when youlose nf1, that's enough in the absence of a somatic event to kick you off, and thatmight not present in the mouse. who knows?

>> i wouldn't rule out nf1 loss as driver,based on the experience with nf1 disease itself, which has incredibly variable penetrationas you know. this mutation in a parent can just give cafã©-au-late spots. >> but the child with the same mutation canbe, you know, severely affected. >> so, we know there are very strong modifiersof every aspect of the nf1 disease, which haven't yet been mapped. >> right, and they seem to be, from what mylittle knowledge of it, they seem to be germline. they do not appear to be somatic. >> they're germline. identical twin studieshave studies have shown that each of the measurable

phenotypes of nf1 have a different modifiergene probably. >> but, this, you know these patients couldhave, you know, just that particular set of modifier genes, which enable the nf1 to becomepenetrent enough to become to be a driver. >> totally, yes. and, it's a fascinating intersectionof like really shoulders of giants, because there's been so much nf1 germline work, thatnow the somatic cells, genomics of cancer is kind of barging in, and the rules are notas simple as codon ras mutation, and we're done. i think we've got to think about this. >> eric? >> yes, so frank will correct me when i'mwrong here, but when luis parada was trying

to model nf1 in the mouse, i recall in fact,that it was important that mass cells, as well as the developing tumor cells also hadthe nf1 deficiency. and, so although your experiment with nf1 deletion in response toadeno cre is a fairly clean demonstration on its own, it doesn't do much, it's alsopossible in that in the context of a germline mutation, perhaps there could be a contributionfrom the tumor micro environment that would actually accentuate that in some fashion. >> absolutely, and with respect to the germlinein the tcga, lung cancer, recall that we did sequence the exome of every peripheral bloodcell in the patient. so, none of those patients had somatic nf1. the parada work, and others,i believe, underscored the importance of the

interplay between the "stromal cells withhaplo insufficiency" and the "tumor cells with loh". kevin will correct me if i'm wrong,but i think in the lung cancer, we've ruled out germline as a driver. okay, thanks, everyone.