Cancer Reseach

eric green:well, good afternoon, everyone. i’m eric green, director of the national huma genomeresearch institute, and i want to welcome all of you to this annual lectureship, thejeffrey m. trent lecture in cancer research. i’m going to give a general introductionto this lecture series, and then i’ll turn over the podium to our deputy scientific directorpaul liu, who will introduce today’s speaker, but i have the pleasure of saying a few remarksabout this lectureship and the person that this lectureship is named after, and that’sjeff trent. the story goes as follows. if you go backin time, the intramural program of nhgri has not been around forever, by any means, andin fact 23 years ago it really only existed

as an idea, but jeff trent was brought hereto essentially build the intramural program. he was brought here jointly when francis collinscame here as the director -- then the director of the institute and was asked to create anintramural program. that is not an easy thing to do when it doesn’t exist at all. he hadto start from scratch recruiting a number of people, getting space, getting staff, gettingeverything organized, and what i would say is he did it in a spectacular way. i had thegood fortune of joining the effort not in 1993 when jeff arrived and some of the firstinvestigators arrived, some of which are actually in the audience. i arrived a year later butwas very fortunate to be able to join what was already a growing and incredible excitingplace to be, an investigator doing research

in, and a lot of the credit for that reallygoes to jeff. he was able to not only recruit good people but to put together an organizationin the style for intramural program that, i think, has served it extremely well as ithas progressed through the years. now, actually in its third director, the intramuralprogram -- so, jeff was the -- was the scientific director from about 1993 to 2002. at thattime, he departed and became the founding president and research director of the translationalgenomics research institute, or tgen, in phoenix, arizona with -- and then had a similar opportunityto build science spectacular out there in the desert, and i’ll leave it to othersto talk, and you can read more about tgen, and there were some great things that havehappened since then.

what i would say is i feel very fortunateto have watched jeff lead the intramural program, and i think i learned a lot, as did all ofus who were research investigators in the intramural program during his stewardshipof the program, and then i was fortunate enough to then be appointed to be his successor,and i served as the scientific director of nhgri from 2002 until just about six yearsago, when i got this new job, but what i would say is one of very things i did back in 2003,shortly after becoming the scientific director was to honor what i felt was a very valuablelegacy that jeff had left behind in terms of his mark on the intramural program andbeing the founding scientific director, and we established this lectureship in his name,and so just the -- and meanwhile we have had

no problem getting truly outstanding scientiststo be honored by coming and giving the annual jeff trent lecture. just to give you a flavor, people like janetroley, lee hartwell, harold varmus, mike stratton, eric lander, brian druker, carol greider,charles sawyer, chris amos, bert vogelstein, stephen chanock. among that list are threenobel laureates and multiple members of the national academy. needless to say, i thinkpeople in the community also have deep admiration for jeff and for what he did here and areabsolutely will to say yes when we invite them to give this annual lecture, and sometimeswhat’s also great about this lecture is that it gives an excuse for jeff to come backand visit us for a day and spend time with

us, and indeed this is one of the years he’sbeen able to do it. so, jeff is here, along with his wife, dee, and it’s a pleasureto have you here, and thanks so much. [applause] so, that’s the history of the lectureship,and i’d like to turn this over to our deputy scientific director, paul liu, who’s goingto have the pleasure of introducing this year’s jeff trent lecture. paul. paul liu:it is truly my great pleasure to introduce john carpten. first, i will say dan kastner,who is the current scientific director, sends his regrets that he couldn’t be here. he’sactually in australia, you know. he’s 13

hours ahead of us, and so he couldn’t givethis introduction. john probably doesn’t need introduction to many in this audiencesince he was one of the first recruits by jeff into his lab as a post-doc fellow andthen a senior researcher fellow when john graduated from ohio state university withhis phd, and he came here, worked in the jeff lab for about six years, and then he becamea tenure track independent investigator in 2000. then, he left nih, went to tgen, becamea professor in relation -- of translation of genomics, and he -- that division of integratedcancer genomics and also became the deputy director of basic research in 2012. only recently,he moved to university of southern california, became the chair of department of translationalgenomics, and also the newly formed institute

of translation genomics. john has made many seminal contributions tocancer genetics and genomics, and he has published over 150 papers, many of them highly cited.i just checked this morning. one of them was cited more than 1,000 times, and -- whichis quite impressive. he has several patents, and he is a very busy person, so -- he’son so many committees, and he’s the -- and he’s the nih, you know, on scientist sectionsand other, you know, advisory roles, and it’s a really great pleasure to welcome him back.thank you, john. john carpten:thanks. thanks, paul. eric, i didn’t know about that list. that’s kind of scary. wow.well, i think i can say honestly that, you

know, no matter how far my career goes, idon’t’ think there could be a more honorable place that i’d like to be than right here,standing at the podium giving this lecture today. i have so many friends and colleaguesin the audience and a number of them who’ve, you know, been incredibly supportive, andthey have played very critical roles in some of the scientific discoveries that we’vemade through the years, and i won’t call names because i don’t want to miss any names,but to all of you, i just want to say thanks again. it’s been a fun ride. i remember one of the harry potter moviesthey were driving through london in a double-decker bus, and it was going all over the place,and the driver said in a very caribbean way,

“it’s going to be a bumpy ride,” andi wish people had told me that before i started because it has been a bumpy ride, but i wouldn’ttrade it for anything. i have had the opportunity, again, to work with a lot of amazing scientistsand just want to say before i get started that i am -- i am presenting this on behalfof a lot of great people. i don’t take singular credit for a single thing that i’ve done.it’s just not my approach to getting things done. i was sort of raised in the cultureof team science by folks like jeff trent and francis collins. by bringing together, youknow, enough smart brains, you can get a lot of incredible things done. and so throughmy career i’ve done a lot of work, both on the germline genetic side, the cell biologyside, as well as cancer genomics and tumor

profiling, and i could talk about a lot ofdifferent things, but i really want to, you know, bring to bear some of the work thatwe’ve done and some of the concepts that we’re seeing moving forward and how aspectsof population and tumor heterogeneity have significant impacts on cancer genome scienceand clinical phenotypes. so, i’m going to spend a lot of time reviewingcancer, and in the new medical paradigm of precision medicine, here again focusing mainlyon the work we’ve done in the oncology space, and i affiliate both with the university ofsouthern california as chair of the new department of translational genomics but also maintaina relatively strong relationship with jeff and the team at tgen as an adjunct facultymember there.

i know it’s not acr, but i always like todo this, and, you know, there are some financial disclosures. i’m one of the founders ofashion analytics, which is a clinical sequencing laboratory, and i’ll discuss off-label useof several fda-approved drugs, including ponatinib, pazopanib, lorlatinib, and pertuzumab. so, getting right to business, i think, youknow, the -- my presentation will be sort of sectioned based on, you know, looking athistorical studies and some of the work that we’ve done and looking at population heterogeneityas a function of cancer genome science, and then i’ll go into some aspects of tumorheterogeneity and how that can influence our interpretation of data that’s being usedboth in the research space as well as in the

clinical space through precision medicine.so, in thinking about population heterogeneity, i think we want to focus on phenotypic differencesthat we see and looking at sort of global cancer statistics for incidence and deathrates in men and women and seeing the common cast of characters for incidence being prostate,and breast, and lung, and colon, and some other diseases that we don’t look at verycommonly because they’re a little bit more rare, like bile duct cancer, cholangiocarcinoma,and then looking at death rates again, lung cancer still being a very significant contributorto cancer deaths even with the advancements in smoking cessation, prostate and breast,of course, and colon. and then we start seeing tumor types thatactually come off as being kind of rare, like

pancreatic and, again, the bile duct cancer.i’m going to talk a little bit about the work we’re doing in that space. but thenwhen we focus on and look at cancer incidence and death rates in african americans or blacks,we see a similar picture where we see prostate and breast and colon, but then we see thingslike bile duct cancer actually float up a little higher to the top of the list, andlooking at -- looking at men -- african-american men, and it actually happens to be the fourthmost common cancer in that group. it’s not that far down the female list, either, butagain, you know, just thinking about -- you know, there are some commonalities when welook broadly at cancer statistics, but there are some differences, and these things thatwe tend to look at as cancer-held disparities

where diseases like prostate cancer are abouttwice as commonly diagnosed in african american men, and although breast cancer is more commonin white women, the cancers can be detected a bit earlier and have more aggressive biologiesin african-american women. and i’ll start off, you know, talking about,you know, these concept of race and ancestry, and i see vence bonham -- at least he was-- he was sitting in the -- in the -- in the audience there, and vence has really spearheadeda ton of work and really trying to educate the population on these differences and thatthese two things are not necessarily the same, and there’s a really complex interplay betweenthe two of these things, where race is being a social construct, and i think francis hascalled it a proxy of sorts, and is much more

related to social and societal-related factorswhere we have -- and then we have ancestry, which is real genetics and biology and isassociated with ancestral genetic material that’s passed to individuals, and again,there’s an interplay here were undoubtedly, you know, individuals of -- you know, thathave common ancestral backgrounds can definitely have societal clustering and higher interactionsand can have more ethnic similarities, but then you can also look at ancestry. we canhave individuals who have very similar genetic ancestries but have very different culturallifestyles, and so these two things -- there’s a complex interplay there, but i really wantto focus on, you know, how we look at these things and primarily, again, thinking aboutgenetic ancestry, and principle components

analyses, and individuals like rick kittlesand others have really spearheaded a lot of this work and understanding that there arevery specific genetic alleles that have very high -- very different frequencies in differentpopulations. a gene -- an allele in duffy is one of the more -- sort of the poster childof these ancestry-informative markers that we can use to actually bin individuals basedon true genetic ancestry, rather than sort of racial, cultural, or ethnic similarities. but again, undoubtedly there’s an interplayhere between race and society and constructs and ancestry and how they impact phenotypesand diseases, and, you know, several questions can be asked. can genetic ancestry be associatedwith biology and phenotype? and the answer

is undoubtedly an astounding yes. it thinkwe can think of a condition or disease such as sickle-cell anemia, and i -- and i, youknow, take a moment to -- you know, make the fact that -- we have to be careful, also,about characterizing phenotypes. everything’s not a disease. some things are just conditions,and it’s a condition of an individual in one environment versus another one. again,sickle-cell is a perfect example where individuals who live in a highly malaria-infested area-- carrying that allele actually have -- it’s been official because those individuals canlive to ages to reproduce. individuals who have a normal allele will be affected andcan die from malarial infections. you take those individuals out of that environmentand put them in an environment where there’s

no malaria. all of a sudden, they have a disease,so we have to be careful about that, and salt retention and hypertension is another classicexample, and we have like diseases and truly deleterious phenotypes like cancer that canbe definitely related more to genetic ancestry rather than race, and -- but race of coursewithout a doubt, and i know there’s social scientists in the audience -- i am not thebiologist who believes that these differences we see in outcome or health disparities arespecifically linked to biology. some of these things are associated with socioeconomic factors,such as poverty, access to health care, environments, and some behaviors that might be associatedwith diets or lifestyles and then this complex -- we get this complex interplay between geneticancestry and race.

but i want to walk through a few exampleswhere, you know, we can start looking at ancestry or race and how it can influence disease riskand ancestry or race and how it might influence disease outcomes, and the first example willbe work that was done in lymphoma by a group out of the acs, where they published a paperlooking at disparities in the adoption of the use of rituximab in patients with diffuselarge b-cell lymphoma, where in the early 2000s it was shown -- it was shown that addingthis anti-cd20 antibody rituximab to chop therapy significantly improved both completeresponses and overall response in patients with lymphoma, but they then went on to showthat there was a disparity in the adoption and use, showing that individuals who wereafrican american were less likely -- much

less likely to get rituximab added to theirchemotherapeutic regimen, and then they were on -- able to go on and show that individualswho had insurance versus those who were uninsured where less likely or more likely to get accessto this blockbuster drug. so, in this case, one can actually almostsay that there’s really no genetic ancestry associated with this disparity and outcome-- these disparities and outcome or access to these drugs. it’s purely access to care,and another example which i get really excited about is work that was done in pediatric leukemia,jun yang and mary relling and the group out of st. jude’s, and this is to me one ofthe real, strong studies that really sort of made a true link between ancestry and biologyin a disease, and i think it should be always

known as one of the first, seminal examplesof this. so, we know that all five-year survival ratesare really high. most of these kids do pretty well, but not all of them do, and there aredefinitely some ethnic or racial differences in survival and outcome, with poorer overallsurvival seen among african-american kids and kids of -- at the time, it’s -- youknow, we’re -- you know, hispanic ethnicity. we can use various phrases, latino -- comparedto european americans or asians, and so jun performed a genome wide associations studyon about 2,500 kids who had been diagnosed with all, and it’s important to know thatthe treatment regimen tends to include initial induction therapy, which can be quite intense,and then consolidation or intensification

to really try to drive the cancer into remission,and then after remission, they’re then given maintenance, and during maintenance in somecases what they’ll do is they’ll do a real intense infusion for a month or two,and this is called delayed intensification that occurs at the start of maintenance therapy,and this will be important as i -- as i talk about -- summarize the rest of the study. so, he performed principle components analysisfrom amassing 500,000 data on these kids, and if you look at the kids -- the sort ofgenetic contribution the individuals with the red -- this is european genetic contribution,africans are in grey, the african chromosomal material or genetic ancestry, green is asian,and then the blue is native american, and

this is where the analysis got really interesting.he was actually able to show that kids that had greater than 10 percent native americanancestry based on their pca had a much higher probability of relapse after therapy, andthis was relapse after induction -- i mean, after maintenance, and so you can -- you canlook here and see even kids who self-reported as white but had greater than 10 percent nativeamerican ancestry also had a much greater probability of relapse.. and then when he looked at outcomes basedon therapy -- when he looked at kids who did not get delayed intensification during maintenance,there was a significant increase in probability, especially when you look at kids with greaterthan 10 percent native american ancestry,

but when you looked at the kids who all gota delayed intensification, you can almost completely eliminate the disparity. so, ifyou have a kid who comes in -- greater than 10 percent native american ancestry, and perhapsthat child should always get delayed intensification as part of their maintenance therapy, so againthere’s this real cool association between genetic ancestry and a clinical outcome, butthrough more of a social economic approach in making sure that they get the appropriatetherapy, you can completely eliminate that disparity. so, i think this is a great exampleof how we can think about ancestry and population heterogeneity and its influence on diseaserisk and outcome. and then i want to switch gears and talk alittle bit about -- a little bit about the

work that we’ve done in multi myeloma inmy lab and collaboration with, you know, scientists from various institutions. i’ll walk throughthat. so, myeloma being a disease of plasma cells with a very well-known molecular pathogenesisand from normal b-cell to monoclonal gammopathy of undetermined significance or premalignantb-cell disease through full-blown myeloma with a primary genetic lesion being definedby translocations of the igh locus on chromosome 14 to a series of oncogenes and downstreamprogress events, including deletion of chromosome 13 or chromosome 13 monosomy, somatic mutationsin ras, ras genes and fgfr and other secondary events such as amplification of cdna. it tends to be a disease of aging, meaningthat the median age of diagnosis is around

70. about 30,000 newly diagnosed cases, about10, 11,000 deaths per year. we’ve seen this incredible development of blockbuster drugs,particular the immunomodulatory inhibitors or imids such as lenalidomide and proteasomeinhibitors such as bortezomib and more recently carfilzomib. the five-year overall survivalrate before the use of these drugs was only about 37 percent, and it’s increased toalmost 50 percent because of the development of these incredible drugs, but also of importanceis the multiple myeloma actually happens to represent on the cancer health disparities.looking at incidence rates in males, it actually has the second highest rate ratio -- raterace ratio between african americans and europeans, both in males and females, and in lookingat death rates, the rate ratios is also among

the top killers for both african americanmen and women. so, it’s been a very -- a disease of quitea bit of interest for my research program, where we’ve got some hypotheses that we’retrying to test, one being although mortality disparities have decreased on the outcomeside, there’s still a consistent disparity in incidence rate, and that along with thishistorical difference in mortality could suggest a possible genetic role or biological rolefor these disparities, and we’ve set out to determine if somatic events and tumorsthat are associated -- have been associated with poor outcome in myeloma through otherstudies might be enriched in tumors from african-american patients.

i had a large series of about 250 tumors thatwe profiled in collaboration with the broad institute, who performed genome sequencing.my lab did -- and in collaboration with jeff, we did a bunch of work in developing geneexpression information and copy number data from these tumors using ray technologies.all the data’s available publically through a portal. within our -- interestingly, withinour cohort, there are about 15, 16 african-american patients, about 180 european-american patients,and what we want to do is to look at these regions of the genome that have been previouslyassociated with poor outcome or high-risk disease to see if there were any differencesin the frequency of these events and tumors derived from african-american and european-americanpapers, and angie baker was a staff scientist

in my lab, was the lead author on this paper,which was a seminal study that was the first to report biological and genomic analysisof tumors from african-american patients and comparing to tumors from european-americanpatients. and just looking at the results of the copynumber, i won’t go through everything in detail, but one thing that jumped out at uswas one q gain, which is highly associated with a high-risk disease actually happenedto be more frequent in tumors from european-american patients than african-american patients. itreached statistical significance, and after correction it went away, but it was stillpretty close, and that was an indication that -- are we thinking about it the wrong way?could african-american patients actually have

tumors that are more associated with favorableoutcome? and if they got the right drugs, maybe they’d actually end up doing better,and looking at, you know, some other -- like the arkansas high-risk gene expression profile,we didn’t see much of a difference there. we looked at 14 q breakpoints. we also sawhere that tumors from european-american patients were more likely to have breakpoints of 14q than african-american patients, again suggesting that perhaps african-american patients havetumors that are more associated with favorable outcome. and in talking to lots of -- hemonc -- hematologists,oncologists who treat myeloma patients, they all tend to say their african-american patientsdo better when they get the imids and the

proteasome inhibitors, so in this case, webelieve that african-americans may have tumors that are more associated with favorable outcome,and if they get the appropriate drugs, then perhaps they might actually do as well orbetter, but there’s still this historical difference in incidence, which could suggestthat there might be an inherited factor involved in more frequent development of multiple myelomain these patients. we’ve now since begun to apply deeper whole-genomesequencing technologies to analyze these tumors, and of course they’re awesome because wecan do very comprehensive genomic interrogation -- point mutations across the exome or genome,copy -- we can deduce copy number changes, identify gross rearrangements, and we haverna, of course. we can do all sorts of cool

transcriptional analyses, and just a few slidesabout the work we’re doing. another study through the multiple myeloma research foundation,which i think is -- it could be a model for cancer genome science going forward. this is a longitudinal study. it’s not taking-- it’s not performing genomic analysis and taking a snapshot of a single sample froma single tumor -- single tumor. in this study, we’re actually recruiting patients at diagnosis,pre-treatment, and we’ve recruited 1,000 patients, and we’re profiling all of thesetumors, normal tumors pairs. tumors have been enriched for a whole genome, whole exome,and rna sequencing. all of the patients will go on one of three treatment regimens, andthen at relapse we’re collecting tumors

and profiling there, as well, so we can reallyget a sense of the natural history of these tumors during the course of therapy, and thisstudy is really being driven by a really talented scientist at tgen named jonathan keats, whoi share the pi-ship with, and winnie liang, who runs the collaborative sequencing centerat tgen, and comparing this study to all of the other tcj enlarged cancer genome studies,we’ve generated more data than all except the nextgen sequencing data. it’s -- forall except the breast cancer studies, and just looking at, you know, the distributionof rna exome and whole genome, again this being compass and this being breast, so we’vegenerated quite a bit of data for this project already, and these data will be made publicallyavailable. one of the postdocs in my labs,

zarko manojlovic has been -- has taken thisdata and run mutsig analysis to look at the mutational status of myeloma tumors and lookingat the landscape. and again, we’ve published some paper previouslyon the thematic landscape of myeloma in a smaller data set, and -- but overall the mutationfrequencies are pretty similar, and we know what the most commonly mutated genes are,but what this study empowered us to do is to be able to look at over 100 african-americantumors, making it the largest study to date and over 500 or 600 or so tumors from europeanamerican patients, and what we’ve down instead of just using self-identified race, we’vebegun to extract genetic ancestry information from this end. so, this was a principle componentsanalysis done by zarko where we took -- first

anchored our pca using 1,000 genome populations,focusing on the yoruba of benin and nigeria, the african-american population from southwest,the sef [spelled phonetically] -- or the ceu from utah, the mexicans from la, and the hongchinese, and so here are the different groups. here’s the yoruba in the brown, the sef,the mexicans from los angeles, and the hong, and the african americans here, you can seethis sort of distribution or continuum through the -- sort of the -- we know the admixturebetween the african and european chromosomes, and then we can overlay our myeloma patientshere in blue, or we have our european self-identified patients and those that self-identify as africanamerican, and then patients that kind of fall in between different areas of the principlecomponent.

and then begin to look at mutation differencesbased on the pca, so african americans that have higher degrees of african ancestry versusthose more centrally distributed in the continuum, and then look at the differences, and so thisis just looking at, you know, the genetic alterations across a series of commonly mutatedgenes and some things -- interesting things have popped out. for instance, looking at,you know, p53 mutations are much more common in tumors from european american patientsthan from african-american patients, and tp53 -- lots of mutations are significantly associatedwith poor outcome in myeloma. so, again, leaning -- these data suggesting that african americansmight have tumors that have -- are associated with more favorable outcome, and then allthese mutations that are more common in tumors

derived from individuals with african ancestryversus european ancestry, and some of the genes of interest we’re looking at is ptchd3.not a well-characterized gene, but with a very significant difference, and i think thisgroup is really interesting because even though we have a lower number of african -- of tumorsfrom individuals with african ancestry, we still see these significant differences. we’vegot almost over 600 tumors here, and so these frequencies should be pretty stable over time. so, again, for the first time, really understandingsomatic differences in tumors from individuals based on ancestry, and again, you know, lookingat the -- zarko looking at the pathways that might be more differentially regulated inindividuals and tumors from individuals from

these different ancestral populations. so,again, being among the first to really look at this and in depth at the tumor biologyand relationship to genetic ancestry, and our data again would suggest that these african-americanpatients may have tumors with features associated with favorable outcomes, so if they get theright therapies, they actually may have better outcomes, but again, we’ve not addressedthe issues related to the incidence and are hoping to do so through genome-wide associationstudies in tumors from african-american patients, using our whole exome data from compass orperforming high-density snp array experiments and gwas analysis to see if we can identifygenetic variants that might be associated with this increase in incidence in myeloma.

so, hopefully, you know, this -- you know,this review period is sort of -- walks you through some aspects of how ancestry can eitherinfluence disease risk or outcomes, and it could be race or ancestry. two examples whereaccess is clearly driving the disparities where there is diffuse b-cell lymphoma, multiplemyeloma, and then one example where i think ancestry and biology actually is driving outcomedisparity in all, but you can eliminate that disparity by making sure kids get a very specifictreatment regimen as part of their clinical management. so, now i want to shift gears a little bitand talk about some of the work we’ve done in precision medicine, precision oncology.again, there’s a lot i can talk about. we’ve

run a number of clinical trials, a melanomastand up to cancer study with jeff, and studying glioblastoma with prados and the group -- mikeprados and the group at ucsf and -- or the study -- triple-negative breast cancer. we’vedone it with joyce o’shaughnessy at baylor, but, again -- yet again, i really want tofocus on how population and tumor heterogeneity can influence our interpretation of data relatedto precision medicine, and, you know, the slide just showing, you know, the fact that,you know, targeted therapeutics are here to stay, and the more we understand about thegenomic landscape of tumors, we can provide novel targets to drug developers who can thendevelop drugs that specifically -- that hit very specific genomic alterations in thesetumors, and i think one of the things that,

you know, really is awesome is the fact thatthese drugs don’t have to be, you know, indicated for a single disease type. we knowimatinib works both in bcr-able, positive cml, as well as kit and pegfr mutated gist,and, you know, we’ve seen activity with egfr inhibitors for non-small cell lung cancerand colon cancer and more recently we’ve seen then development of parp inhibitors andwork that we did with roisin ine [spelled phonetically] and ken pienta showing the brca2homozygous deletions occur at frequencies higher than we would have once thought incastration-resistant prostate cancer, and these tumors respond to parp inhibitors. and it was fast-track, expedited approvalof olaparib for the treatment of a subset

of castration-resistant prostate cancers,and then, you know, i’d be remiss to not say that this new sort of -- it’s reallynot new, but a newly adopted approach of immune checkpoint inhibitors treatment modality andtargets like pd1 or pdl1 -- on the tumors, pd1 or ctcle4. in the t cells and the incredibleoutcomes that we’re seeing with these immune checkpoint inhibitors. it doesn’t work foreverybody, but in those tumors where it does work and we have some biomarkers where itmight work, like hypermutation, expression of these biomarkers, or the expression ofneoantigens. so, this growing laundry list of genomic alterations and targeted therapieshas led us to this new revolution of precision medicine, and where we can profile a tumor,identify the appropriate alterations, and

then perhaps select a targeted therapy forthat patient’s cancer. these events, of course, or these genes aremutated by various mechanisms. we know the oncogenes in many cases can be amplified ormutated or overexpressed, tumor suppressors deleted, mutated, or hypermethylated, andwe also know that some of these genes are altered by breakpoints, translocations leadingto the degeneration of oncogenic fusions that can also be targeted. so, you know, throughcollaborations, a lot of folks at tgen and partnership primarily with my good friendand colleague, david craig, just, you know -- i mean, god, i can’t say enough awesomethings about david and the work that he’s done to help build this out, but buildingout this algorithm -- and this was -- and

i have to give credit to, you know, the standup to cancer team that helped us derive this and submit an exemption for device to thefda in support of our -- jeff’s stand up to cancer study, but, you know, starting withpatient and doc and consent and biopsy collection and -- of specimens and sending this specimensthrough quality assessment of the anolytes and sequencing and bioinformatics analysis,and then essentially merging the somatic information to the drug space, and then generating thesereports -- holding molecular tumor boards to vet that information and then provide whatwe feel is the most appropriate therapy for that patient based on the molecular profile. part of this was building out an incrediblystandardized and validated platform, part

of which was the creation of a controlled-accessclinical portal, again, it was david craig, my just tremendous friend and colleague, whereeach patient can, you know -- a clinical research nurse can go in and create a new patient for-- fill out a form with a controlled lexicon and vocabulary, standardized clinical annotations,and then from this be able to generate -- auto-generate a report that can be used as part of the moleculartumor boards and the sequencing. i don’t get any money from illumina, but i have theirmachine here. just want to make sure that that’s said, and then we have, again, david’steam building out just incredible bioinformatics framework to support all of -- everythingfrom data collection, data management, data analysis, secondary analysis, generation andannotation of the information, creating the

reports that can then be vetted by the tumorboard, and an incredible establishment of relationship with dell computing, a lot ofwork -- hard work that jeff did that has supplied us with an incredible high-performance computingenvironment to support this work. and then, you know, the process of comparingeach patient’s somatic tumor profile to a relational database developed by jeff kieferand maurice and sara byron and using heuristic programs and algorithms to match the mutationprofile with the drug-gene relation database, and we can sort of modify the -- or tune thepharmacopoeia however we want based on the study to generate reports that’ll have thedrug, other information, like is there an oral formulation for the drug. sometimes it’sreally important. the alteration driving that

relationship -- is it a positive or a negativeindication -- and whether or not there are clinical trials available. we built a clialab, fully validated, so it’s clia certified, cap accredited. i have to give credit to jeaninelabello [spelled phonetically] and lisa baumbach--reardon, who were the lab directors, mary ellen ahern[spelled phonetically], who was the lab manager who helped us build this clinical laboratory.and so -- and jeff and pat had to submit this document to the fda in support of their clinicaltrial, and i have a picture of their faces before and after this process, and i thinkthere were a number of babies born, some divorces, and a few people quit after that -- this documentwas submitted, but, boy, the overall benefits of having this in place have been tremendousfor us because, you know, we feel like we’re

thought leaders in this space, and dna willvalidate a platform such that it meets the rigorous and stringent guidelines for clia,cap, and the fda. a number of papers have been published onmethodologies as well as the results of some of these trials, the triple-negative breastcancer study being the first and was the most cited paper -- molecular cancer therapeuticsin 2014, amazing study led by sara byron with mike prados in glioblastoma. even on the pediatricside, i work with lenny sender at the -- at children’s hospital orange county, led bymy postdoc, troy mckecrin [spelled phonetically], who was recently promoted to assistant professor,and i’m going to talk a lot about the studies we’ve done in cholangiocarcinoma and ofcourse the work that jeff and pat are leading

in looking at non-braf v600 mutated melanomas. and even some papers related methods and forvalidation. we just had this in “nature reviews,” and just sort of reviewing thetranslation of rna sequencing into the clinic. we have this paper where we have a set ofsynthetic oligos that mimic oncogenic fusions that can be used -- spiked into a sample ina clinical lab and used to validate the detection of oncogenic fusions from rna-seq data, andthen just a week or two ago, this came out in collaboration with illumina as well asmarco marra’s group in vancouver, where we have a normal tumor cell line pair, colo829, where we’ve all sequenced this in a clinical laboratory setting and actually developeda set of truth variants that can be used or

compared to in any clinical laboratory toput this cell line pair for -- as a somatic reference standard for clinical sequencing. and so, now moving onto population heterogeneity.i think one of the issues we face in precision medicine today is the fact that, you know,in many cases we want to just be able to use a tumor sample because it -- in using thismethod, we, you know, we really don’t interfere with the clinical standards and approachesand pathology labs, right? we’re not asking them to give us a piece of fresh tissue andsort of alter their day-to-day work flows, and -- but some feel that requesting a samplefor constitutional dna is not always feasible and complicates or slows the testing, and-- but, you know, there may be high risk of

false positives, and this is kind of wherei want to focus this talk. several -- and this is very important, i think,from the -- from the population standpoint as some populations have higher genetic variation,so much like -- much more likely to have private variants in the genome, and those privatevariants can sometimes trigger germlines, and we’ve seen it many times, and just someexamples. i’m not going to name labs. i’m just going to say lab e, lab a, and lab b,this lab doing somatic analysis on the same tumor or the same dna that was used for tumor-onlyanalysis where this lab had normal tumor, knew or was able to state specifically thatthat was a really -- a germline brca mutation, and looking at the variance of undeterminedsignificance on this list, almost 10 to 12

of the variants actually were germline variants.they were -- had nothing to do with -- these were not tumor-acquired somatic mutations. another case -- colorectal cancer, same situation.another patient with metastatic triple-negative breast cancer where our lab detected a frameshiftmutation at pic3r1, which is therapeutically actionable. it happened to be missed by theother lab. we won’t go into that very much, but there were two other mutations on thefront page of that report, one of which triggered a drug rule, and interestingly we were ableto show -- because we did normal tumor pair -- that that actually was a -- those two weregermline variants. one of these genes -- nach -- even thoughit didn’t trigger drug rule here, it later

could have, would have -- which would havebeen the use of -- again the secretase inhibitors, for which the -- one of the side effects isincredible diarrhea, and so we want to make sure that we’re doing this appropriatelyand that individuals, again, from populations that have higher degrees of genetic variation,where data may not necessarily be in the public databases for filtering, can have more ofthese false positives on their reports. and so it was work done by rebecca halpurn[spelled phonetically] and david’s lab and zarko in my lab, and we took a series of patients,and we had african americans, european americans, and performed principle components analysisand looked at the number of mutations that would have been -- sort of from a tumor-onlyanalysis, if you just use filtering, using

publically available snp data, showing thatyou would have had a much higher number of false positives in individuals who have alot higher degrees of african ancestry versus individuals of european ancestry, and lookingat the number of tumor -- of mutations, again, the average being 242 here versus 125, andthen looking at european versus african american. this table’s not right, but essentiallywhat we’re -- what we’re seeing here is that if you were african american, there’sbe an average of 240 variants on average versus 125 if you were european american, and twoand a half or three would have triggered a drug rule in an african american versus 1.5of those false-positive drug-rule triggering variants would have been on your report if-- in the case of european americans.

so, again, and this is a statistically significantdifference. so, there’s definitely an ancestral bias to this, and we’ve been playing aroundwith optimizing approaches using probabilistic approaches, and essentially what we’ve doneis we’ve sequenced a series of tumors -- really deep, done some dilution series, and cameto the conclusion that if you have more normal contaminated stroma in the specimen, you canuse the germline variants because they -- whether it’s tumor or germline, germline variantsshould have, you know, an allele frequency of about 50/50 in a polyp if it’s a heterozygousman, but if it’s a tumor variant and there’s normal contaminating stroma, those heterozygoussomatic variants will have an allele shift, and you can sue that information to reducethe number of false positives, and if you

sequence higher, meaning if the -- if youhave a really quality sample that’s like 90 percent tumor, you have to sequence upwardsof 1,600 to 3,200 x to be able to identify these allele shifts, and so we’ve been able-- we’ve played -- david and rebecca and the team essentially came up with some algorithmswhere if this was a triple-negative breast cancer from a ghana african patient, these-- the -- you know, you have, you know, almost 350 false positives if you use filtering only,but if you use our approach, you can significantly decrease the number of false positives thatwould end up on an approach. and here ghana and african -- european american triple negativesand african american triple negatives and a patient with glioblastoma that had latinoancestry.

so, i’m going to use the final few minutesof my presentation to talk about tumor heterogeneity and its impact on precision medicine. i thinkmany of us in cancer have followed the work of charlie swanton and the group over at cambridgeand in the uk -- the sanger center, where they took a sample from a kidney cancer patientand chopped it up into various pieces and then looked at the regional profile of eachof these specimens, and they also had a chest wall metastasis that they looked at, as well.they were able to show that one of those specific regions -- region r4, which was this region,had mutations that were also available in the mets, but none of the other lesions fromthat tumor did, suggesting that this tumor sort of set in the middle between the primaryand the mets, and so they were able to draw

these phylogenetic trees. just amazing workthat showed at the gene expression. a level r4 also mimicked the mets more than any ofthe other from the primary renal lesions. and so we’ve worked out some algorithms,so this is one of our compass multiple myeloma patients, and jonathan keats built out bioinformaticstools that will allow us to look at clonal variance in the tumors. so, in this particularpatient, we picked up three different clones. this is looking at the -- sort of the bimodalor tri-modal distribution of heterozygous variants, which indicate various sub-clones,and so in this particular patient, looking at time point one and time point two, thedistribution of these clones is pretty similar, meaning this clone remained the dominant clonethroughout the course of this patient’s

clinical management, yet another patient,you see very significant clonal dynamics occurring, where at diagnosis you have this red and blue-- these red and blue clones that are the primary and this low-level, low-lying greenclone at -- was also low at this time point but became among the dominant clone, and theblue clone goes away over various courses of treatment. so, this tumor heterogeneityand this shifting and flowing of various clones under various therapeutic conditions is goingto provide us with information on understanding what mechanisms are associated with responseor resistance to the various therapies. and then finally, work we’ve in cholangiocarcinomaor bile duct cancer, tumors of the bile duct. they can occur sporadically or be associatedwith liver fluke infection. we’re primarily

focusing on the sporadics, and it can be verydifficult to treat. we published this paper with alan bryce and mitesh borad at the mayoclinic at scottsdale, where we showed the integrated genomic characterization and clinicalsequencing identified therapeutically actionable contexts around fgfr and egfr pathways andtreatment. one patient with an errfi1 homozygous deletion -- errfi1 is a very obscure gene,was not on many of the cancer panels that people were using in clinical laboratories,but it’s critically important because its role is to prevent dimerization of erbb receptors.so -- and it can be among the -- it has to be one of the most important tumor suppressors,but understanding this patient had loss, hypothesizing that it was -- it was leading to hyperactivationof the erbb signaling, egfr signaling -- tumor

before erlotinib. tumor after erlotinib. thisis looking at metabolic activity. just a dramatic response because we selected erlotinib becauseof the hypothesis of egfr activation. we also identified tumors that had fgfr2 oncogenicfusions and translocations -- treating those patients with fgfr inhibitors, looking at-- this is the metric scans before ponatinib, which is a select fgfr inhibitor and tumorafter -- the tumors just melted away in these patients. so, moving forward and thinkingmore about the concept of heterogeneity in a different context. patient comes in withadvanced cholangiocarcinoma. sample sent for clinical sequencing -- looking at the copynumber data, we saw this even on chromosome eight, which was characterized as leadingto an oncogenic fusion that encompassed nrg1.

nrg1 happens to be the ligand for erbb receptorsleading to increased phosphorylation activation. again, more evidence that in this tumor type,there’s a context around activated erbb receptor signaling and erbb fusions have beenseen in other tumors, not specifically this one, but what they showed was that these fusionslead into activation of hr2 and hr3. we also identified yet another errfi1, anactivating event in this -- in this tumor, so we had activation of the erbb receptorpathway on both sides, both hyperactivation through nrg and loss of mig-sigs errfi1. thediscussion of the tumor board focusing around erbb receptor signaling, not wanting to targetany of -- any of the specific receptors but thinking about targeting multiple receptorsby using a drug that would prevent dimerization,

that drug being pertuzumab. this is livercancer before, liver cancer after, just an incredible response to this, but it was atransient response. there was a progression after several months on therapy. we were actuallyable to collect a biopsy of progression, and looking at the copy number, the event wentaway, and this blazing amplification of egfr pops up that wasn’t there in the initialtumor. right, so treat with pertuzumab. we get rid of the nrg1 clone. this new clonehas arisen that has this incredible amplification of egfr. significant response to lorlatinib,but again, it was transient, and now we get a third biopsy, right? so, what happens next, right? you’ve beenon pertuzumab, had a response. you then go

on lorlatinib, have a response. cancer comesback. so, what happens? brrrrr, right? is this going to change? i think i ran out of-- will this work? oh, oh, acknowledgement. what did i do? you guys can see my talk backwardsto forwards. okay, here we go. one more. can you do one more? no, i don’t. one more.go forward. it should -- i am. i like pressed it like 80 times in both directions. don’tpress the button. oh, guess what happens? the first clone comesback. so, you’ve got these clonal tides. so, pertuzumab kills this clone. lorlatinibkills this clone, but the first clone came back, but what happens in clinical oncologyat progression? you never give the -- a previous treatment because in the oncologist’s mind,the patient has failed that treatment, right,

and so we have to take these aspects of tumorheterogeneity and -- i can’t get this to move -- into context of precision medicine,and i think it’s complicating our ability to see improved outcomes using, you know,this incredible clinical approach. so, in summary, population tumor, heterogeneity,and genome science, which leads into oncology. it’s this. there is race and ancestry, andthey can be related but are not necessarily the same. i think i’ve shown that. theremay be an enrichment of certain biological factors that can be associated with biologyand phenotypes and outcomes. research into uncovering these factors will help us betterunderstand a disease etiology within various populations that could provide informationon how to better -- how to best approach clinical

development of tools and methodologies canhelp to improve the performance of precision medicine, which can have critical impact onclinical interpretation results. i think i’ve shown that, particularly in ancestral populationsthat have not undergone significant bottlenecks, where there are lots of genetic variants.and tumor heterogeneity, of course, exists and is a confounder in current precision medicineapproaches, and better tools and methods will allow us to better sample the entirety ofa patient’s cancer instead of taking snapshots of individual biologies to uncover these clonalaspects of these tumors that are primed to induce relapse. so, in conclusion -- more. there you go. shouldi touch the button? who cares. acknowledgements.

boy, this thing was full of people’s names. [laughter] it was all me [laughs], but, you know, just,you know, specifically, of course, david craig and all the work that he’s done. he’sbeen an incredible collaborator and partner in a lot of this work. jonathan keats andwinnie liang in developing a lot of our sequence and infrastructure there at tgen. can’tsay enough about the support that jeff has given me through the career to have the freedomto approach this high-risk studies that have, i believe, allowed me to apply, you know,knowledge and science and genomics to improve clinical care for patients, and then our clinicalpartners, mitesh and alan at mayo clinic,

pat lorusso at yale, dan von hoff and thegroup at scottsdale health, and dan has been sort of the godfather of precision medicineat tgen, and then all of the funding agencies, mmrf, the prostate cancer foundation, komenfor the cure, the ben and catherine ivy foundation, the nih, and the nci. so, i’ll stop there.thank you. i don’t know if there’s -- paul liu:a couple of quick questions. john carpten:if there are any. i’d love to engage the audience. eric green:i think on the --

john carpten:thomas. male speaker:this new -- and actually one -- this team actually localizes at the site of a commoncopy number variant, and we and others have shown that these have sites where there comesome breaks, and this is correlated to either translocations or sites of copy number changesif you do see the change. john carpten:absolutely. no, absolutely. you know, there’s, you know, a really interesting observationthat we’ve made in triple-negative breast cancer. i don’t want to give it away yetbecause i think it’s going to be seminal, but we’ve seen breakpoints in a very specificregion of the genome encompassing a very important

tumor suppressor. only tumors from african-americanwomen so far, and my hypothesis is that there’s some genetic haplotype or some -- that’smore frequent in individuals of african ancestry that’s leading to these breakpoints occurringat the very specific region of the genome, so, you know, it’s part of our ongoing studies.our hope is to better understand that to be able to, you know, utilize publicly availabledata like the data you guys are generating to look at where these breakpoints are occurringand to understand more about the genetic contribution and haplotypes and how they relate to specificalterations occurring in specific somatic -- specific somatic events occurring at specificregions of the genome. i think that it’s been a lost opportunity in the cancer genomespace, that there really hasn’t been enough

integration between the germline communityand the somatic community to really understand how the germline influences somatic alterationsin cancer, and i think that that could be one of the more exciting opportunities forus going forward and understanding cancer initiation and progression and outcomes indifferent -- in different groups of people. paul liu:maybe i can ask you a question. john carpten:sure. paul liu:so, for the multiple myeloma story you mentioned african americans -- they actually have agood favorable prognosis signature, but their prognosis, you know, is actually worse, sohave you looked at the treatment, you know,

they have received? john carpten:yeah. paul liu:well, is there a correlation or something [inaudible] john carpten:that’s the amazing thing about the compass study is that it’s longitudinal, and weknow treatment upfront, and we know treatment along the patient’s clinical management,so we don’t yet have that information, but we will by the end of the study because we’llhave arguably the only data set that will allow us to do that, that’s powered enoughto make those correlations to be able to say,

if you have these events and you have favorableoutcome but you didn’t get, you know these three specific regimens, you’re likely todo worse, and you’re likely to do better. so, we’ll be able to make those correlationsbecause of the design of that study. paul liu:if no further question, then it’s like thank you very much for this wonderful lecture. john carpten:thank you. paul liu:there will be a reception right now in the -- in the library, and as a token of appreciation,we’re going to give this to john. it’s a, you know, aerial photo of nih, and you’rewelcome to sign this along, you know, the

frames. you know, this will be in the library,you know, and -- john carpten:where is building nine? that’s what i want to know. paul liu:building nine. it’s somewhere right there. john carpten:there’s history in a basement of building nine. i need to go take a picture of it beforethey tear it down. thanks, paul. thanks, everybody. [end of transcript}