Cancer Research Prostate Cancer

>> good afternoon. it's that week where the traffic gets much worse in bethesda and its surrounding areas. it's the week where kids go back to school, sometimes smiling and sometimes complaining and sometimes both. it's the week where the first

nfl football game occurs this evening. i won't even talk about politics. and it's also the week where we kick off the wednesday afternoon lecture series. and it is wonderful to see you all here and to welcome you as

well as those who are watching by videocast, probably in great numbers. that tends to be the case. and i hope you have seen the line up of wals lectures for this academic year because it's truly stellar. and i would encourage you all

who haven't done so, to mark your calendars for 3:00 on wednesday afternoons to try to protect that time so it is going to be an amazing practice raid of remarkable scientific ltd.ers covering a wide virally of topics of interest to a lot of us.

a lot of effort goes into identifying the speakers for this series with lots of input from all the institute and we are quite successful in lining up the people we met want to have come and speak because this is seen as a very prestigious place to speak about your

science. i would encourage everyone to try to be sure to take advantage of that during the coming months. you will learn a lot and you will go away enriched and and inspired by the nature of the talks that you hear about

cutting-edge search and biomedicine. today's kickoff is a great example of that and we are delighted to have as our speaker today, dr. a gargarhe is an md and ph.d. -- levi garraway and had most of his professional career - i would say just about

all of it, in the city of boston. having received his undergraduate degree at harvard, his medical degree and his ph.d. from harvard, he followed on being a resident of internal medicine with the mass general a key resident at mgh.

he followed that in the clinical fellowships in medical oncology at dana fasciaer and brigham and women's and now an associate professor at harvard medical school and in the department of medical oncology at the dana-farber cancer institute, as well as serving as associate

position at brigham and women's and senior associate member at the borough institute keeping himself connected with all these amazing institutions up there in the city of boston. he has received a number of important honors, i'll mention one because it's sort of a nice

ninety-five connection. he was was one of one -- nih connection. new innovator's award. an award specifically designed for individuals woat that time had not previously served as a pi on an nih grand and had a particularly innovative creative

approach a problem that was considered groundbreaking and to get one of those was to be extremely peltive process. so levi must have written a bang-up proposal to achieve this. it was on defining melanoma therapeutic avenues with

functional genomics. also a member of the asci and his research, as you will hear, has been on leading edge of precision medicine for cancer. and specifically, in the talk he is about to discuss today, going beyond that first level of identifying what kind of

actionable mutations might occur in a particular cancer and also identifying which cancers are going to be responsive and which are not to target the cancer therapeutics. he has made many contributions to that field and published more papers than i can read off to

you. just in the last year or two. and it is a great pleasure and privilege to have him as our kickoff speaker and i would ask you to give a warm welcome to dr. levi garraway. [ applause ] >> thank you dr. collins for

that wonderful introduction. my nih new innovative award is just ending and i'm already missing it. that's what grant writing is all about. so i'm very pleased to be here and give this lecture and have a chance to describe to you at

least from my perspective, and the perspective of a lot of investigators in the field of cancer, the excitement and the promise of the emerging era of genomics-driven cancer medicine. and by genomics-driven cancer medicine, one can think of that as a sub set of precision

medicine but in the first instance of precision medicine will be using genetic dna-based alterations and exploring the extent to improve the care of cancer patients and the other point i think is that in moving into this era, it's really about testing a hypotheses that has

budget brewing really for decades and that hypotheses is shown on this slide which is that the use of cancer genomic information to guide treatment choice may offer a categorical means to improve the care of cancer patients. that suggests that many, if not

most patients with cancer, would benefit from some type of systematic genomic profile and the use of that result to guide treatment choices. and of course the rational for testing this hypotheses is rooted in several key observation that is have been

present, many of which have been present for a long time but others are more recently relevant. the first is that molecular pathways involved in tumor survival and progression are enacted by genetic alterations. we have known this for a long

time. the cancer genome atlas are reinforcing this every day but the second and third points are recent lehmann fest. anticancer agents targeting many pathways entered clinical trials. this at the scale and scope it

is now true has been the case for the past two or three years. for the first time in history, we can say that the major clinical or the major signaling path ways we have known about, map kinase, many receptor tyrosine kinases and apoptose and metabolism are being

targeted by multiple drugs. that's a critical ingredient for testing this high pot assist. the final point is the genomics technologies have grown to the point where they enable robust -- robust tumor profile the in the clinical arena. we have known tat cancer genome

is in from for quite sometime now we have a repertoire of components in place to allow us to test the guiding hypotheses of the cancer genome era from a clinical perspective. and the testing of that also inspires a framework, which admittedly in most cancer

centres, is still largely aspirational but has some definition to it in terms of what process, what infrastructure is needed to push this hypotheses. and it begins of course, in a patient-centered fashion, the notion that we encounter

patients either in a focused way prioritizing particular characteristics of the patients or in some cases perhaps in some cancer centres in enterprise-wide way but ultimately we hope that we would use tissue ideally fresh biopsy material and generate profiling

using either existing or increasingly emerging technologies that will then generate data that we now apply new algorithms and flaps some cases a corlat of lab testing to interpret the data and with that, we will be equipped to make management decisions.

and the input into the decision may come from review of the data by a committee by having frameworks in place, pathway-like frameworks, clinical pathway-like frameworks to guide decision making and ultimately in the near term being applied in

hypotheses-driven phase i trials or other mechanism-based clinical studies. and then with that decision, we will then look to see whether or not all of this input increased the prevalence of clinical responses and there are a variety of ways to do.

that the other hope would be that often we can gain another biopsy at this point so we can understand whether or not our clinical experiment was accomplished therapeutically. did we inhibit the target the way we hoped? and then finally, if we see

responses or whether we do or don't, we need to understand mechanisms of drug resistence because certainly particularly if these are single agent studies, we recognize and even if we are fortunate to get responses, these are short-lived in advanced cancer.

so we get another biopsy at the point of relapse could be crucial and discerning for us mechanism resistance therefore can inform a salvage therapy or perhaps a novel therapeutic combination that can subsequently be tested at the beginning of this process.

now i think it's naive to assume that every cancer center could put in every component of this engine but yet it's a nice ark-type for thinking about the types of activities and processes and technologies and algorithms that need to be pursued in parallel to bring

this notion of precision medicine forward. so clearly at the will be occupying us greatly for the next decade or more. so for this talk i'd like to get a relatively high level touch on aspects that our lab has been pursuing that really get to

three underlying activities that can be informed by various components of this process. the first of which sort of at the front end where one is generating and interpreting data, speaks to the need to understand salient driver mutations, particularly those

that could be actionable using therapeutic armamentarium. i'll give you a vignette there. the second is the clinical testing, how are we going to carry out clinical tests of the precision medicine hypotheses? and the third, leveraging information at the back end both

tissue-based studies and coordinated systematic experimental studies to understand mechanism resistence and incorporate the knowledge of both resistence and the spectrum of dependency into the future development of novel combinations, the gel of which

would be durable control of specific cancers that are driven by cardinal genetic alterations. so to begin, i'd like to tell you a story of how moving from the catalog of alterations that are being generated by efforts such as tcga to, knowledge of salient driver events was a

problem we needed to solve in a cancer that has been a major focus in our lab, which is melanoma we as many others, including your dennis samuals, who is here, have been sequencing melanoma genomes trying to understand what biology and new drivers there

may be, and this is a figure from a recent report of whole genome sequencing in melanoma. one of the most prominent features that leaps out is the mutation rate in melanoma is rather high. so, this is showing mutations per megabase, the melanomas on

the right are more typical of what you expect for a lot of epithelial malignancies but a vast excess of mutational load and that is almost completely attributable to the mutational signature that is characteristic of uv light. so uv causes major damage in

melanoma we anyhow that but this gives us a clear view of this. and this phenomenon posed a huge problem from the standpoint of discerning driver versus passenger mutations. and there are three components of that problem that i'll just highlight briefly for you.

but just to illustrate the breath of the problem. if you look at the mutational rate of melanoma, you consider the vital stats of air recent whole exome sequency. we saw nearly 87,000 independent coding mutations which meant that there were 14,000 genes

mutated at least once, two-thirds of the genome had at least one somatic mutation and there were 5upon 15 genes mutated in at least 10% of the samples. now if you discover a gene that is mutated in 10% of a cancer, that is a respectable mutation

frequency. those kinds of genes could have a substantial clinical impact if they were actionable. but this seemed like an awfully large number of genes to really be biologically relevant. and when one started to apply statistical algorithms that

predicted what to expect boy chance, there was clearly inflation. so there were 7genes deemed statistical significant by theal go rhythmms state-of-the-art at the time. that seemed too high. they tended to be absent or very

minimally expressed in melanoma. any given gene might show lowics r. expression for a number of reasons. to see the whole category of significantly mutated jeans poorly expressed was a red flag and perhaps the most deeply concerning aspect of this

problem was that when we considered the silent mutation rate, meaning the mutation that is didn't give rise to ammono acid substitutions in the respective genes, one found that in general, the statistically significantly mutated genes tended to have a very high ratio

of silent mutations. so this was a dope concern that despite our heretofore attempts to apply robust controls for the size of a gene and the samples specific mutation rates, we were still leaking in terms of lots of passenger mutations coming in with the drivers.

how do we solve that problem? eran was an associate combinational biologist in the lab, had the notion we might need to define the definition of background mutation rates. typically we thought as genome uniform. so more or less it would occur

randomly within the genome. maybe a given sample may have a higher mutation rate than others. at a genome level it would be random. maybe that was wrong. we than there are heterogeneous aspects of the background

mutation rate, transcribed genes are mutated at a lower frequency than untranscribed genes. there are variations depending on whether it's coding or noncoding. parts of the genome intron that is are transcribed versus those that are not transcribed.

so we know that this can vary and perhaps particularly when you have an extra layer of mutational insult, this was posing a problem. so the notion can be shown additionally pictoraly here which shows the hypothetical example.

these are exons, this is intronnic sequence. these are exon that is have equivalent mutation ratios. if you look at the exon mutation ratio. however, you could imagine a scenario where there was one locus that has a high background

mutational rate in which case you see high intronnic mutations and another locus that has a low intronnic mutation rate. in this case, you might expect these are mostly passenger events and it just so happens that this locus has hydrogeneration and low repair

but this might have undergone evolutionary selection. society question was, how could you enrich for loci that actually underwent evolutionary selection in cancer? it turns out that when you generate whole exome sequencing data, even though you're

targeting pulling down the sequence, you get a whole bunch of intronnic sequence as well. the reason for this is because the fragments you pull down with the baits that are targeting the genome, tend to be more or less randomly generated so you get a distribution of sort of adjacent

intronnic material that comes for the ride and that allows us in the regions flanking the exons to calculate the mutation rate even though we typically only care about the exonic mutations that will allow us to develop a algorythms that tallies the mutational burden at

each locus and allows a per mutation of the mutations in a locus specific way across the sample set that gives us a background mutational distribution and if the mutation burden is considered to fall outside of the that mutation, we could say that there is evidence

of fossil selection whereas if it falls within, there would be background even if there are multiple recurrent exonic mutations of that was the notion and we could apply additional algorithms to annotate these mutations as plausibly flongsal or damaging and now we can sort

of see how did we do compared to the prior approaches? and indeed, this algorithm if one considers the criteria of selection where there was a low expression and prior significant genes suddenly now you're in a situation where in general, the mutations, the genes being

called significant are in fact expressed in melanoma and now when you look at this qq plot where there was a vast inflation earlier, we now see a much more of a distribution in line what have you expect boy chance until you goat a few genes that are clear out liers and these 11

genes caught our attention. there were several very interesting aspects of genes. one of which maybe the most notable was a gene called rac 1, a member of the rasgtpase super family. there was a mutation at codon 29 and other members of this family

that had analogous mutations and functional studies i can't describe in great detail suggested that in fact, this mutation, p29s did activate rac as evidenced by a gtp loading assay. so this was a functional mutation and now very

interesting to think about the relevance of this gtpase in melanoma biology but there were earl other interesting mutation that is caught our attention ha had not been prefersly known. for example, this protein, foos days, which has not been described as a cancer gene but

it's known to interact with psyche clin d and negatively regulate cyclin d ask they were loss of function events and there were several stop codon in this panel. so this is a novel tumor suppressor gene. this is another gene sorting

snx31 a gene reported to interact with activated hras so this may be a ras effector protein and here is another tacc1, which is again reported to stimulate both ras signaling and pi3k signaling to pathways and it also interaxe with aurora kinase as does ppp6c.

so a number of interesting genes that are clearly showing evidence of positive selection during melanoma evolution and have or point a finger to novel biological mechanisms that we did not previously recognize in this malignancy. now you can flip this around and

ask not just about activating mutations or functional mutational burdens but focus on loss of function events and when one does that, we see some interesting hits emerge for example, arid2 is homologue to genes such as arid1a and b, members of the switch sniff

chromatin modifying complex, this is a complex component of which now have been shown to be mutated in many cancers so we can add melanoma to the list of cancer that have significantly recurrent mutations and chrome tine modifying enzymes. this was a usesful stud they can

focus studies of chromatin bile nemelanoma. and just to kind of make a brief point of a subset of the mew talks we saw by this analysis, we have we of course know the large fraction have oncogenic bras mutations and nras mutations.

but a substantial subcelt of 10 or 15 or so percent ever mel gnome that is lack either of these mutations and one of the interesting observations which was confirmed by a stud they came out by ruth at yale that was many of these bras wildtype melanomas have nonsince nf1

mutations, arrases gap so it's normal function is to inhibit ras signaling so loss of function in nf1 deregulating ras would be another mechanism of map kinase pathway activation in melanoma and we can assign this as a cancer gene in that sub celt of melanoma so all

together, we can put or take the output of this algorithm to look for positive selection together with bayesian mining of the list of known cancer genes and define a landscape of melanoma driver genes which we wouldn't argue to be complete but we have certainly largely eliminate the

the problem of massive passenger dilution of the signal. so therefore, getting back to this issue of salient, understanding salient driver genes from which we can now look to understand actionable therapeutic manipulations, this was a necessary component of

melanoma biology and was enabled by this kind of study. so, and there are many examples of studies like this taking the next step going from list of mutations to the salient subset for a clinical use that are ongoing in many different cancer types of the so now i'd like to

turn my attention to another effort that has been going on on the opposite end of the position medicine process which is to dissect using both sequencingsequencing and preclinical studies and mechanisms of resistance and to use that to come up with novel frameworks for therapeutic

combinations and here the notion is that single drugs are unlikely to cure or durably control most metastatic melanoma. we need to understand what types of combinations, possibly 3-4 drug combinations might be given in various cocktails to patients

with particular genetic alterations and it certainly has been shown in human malignancies that cocktails are needed to achieve durable control or cure in cancer. so understanding mechanism resistence might help with us that particularly in melanoma

when there are variety of therapies that are fda approved such as ras inhibitors or advanced clinical development against the map kinase cascade activated mutationally by mutations in b ras particularly at codon 600 or b ras. so this is an area where

understanding resist sense incredibly important because despite the increase in survival by using ras inhibitors in b ras melanoma, resistence is quite prevalent and this is prevalent in two ways. son if you look at a water fall plot where the fraction of

patients that have clinical benefit are shown below the zero line, if we apply a formal criteria for clinical response, a clinical partial response, you see that really only a subset of these patients actually achieved a partial response. a large fraction of patients

although they are benefiting, there is a lot of tumor around. intrinsic resist sense a major issue. the median survival is about 6 months with the ras inhibitor. if you at a mec to a ras you can push this out some. so it is a pervasive problem of

the so how are we going to solve this problem? well, first of all we can benefit from the fact that we know a fair bit about themes of resist tones kinase inhibitors. so b ras mutation which activates melanoma is a kinase-driven, genetically and

kinase driven cancer. if you take a kinase inhibitor and you get a response and you relapse, it can fall into three major categories, typically. not universally but typically. one major cacategory sereactivation of the target. often this occurs by acquiring a

secondary genetic alteration but this can also happen by activating up stream effectors so to put the ar get into overdrive and the drug is not as effective at a given dose and alternative approach is that a bipas mechanism gets activated. so essentially you work around

the effect of the kinase inhibitor and engage a parallel cascade and feed into the downstream oncogenic output in a way that leads to resistence. the third major category seone can activate downstream effectors. so members of the pathway that

act downstream of the target oncoprotein, if you turn them back on, you can achieve resistence. an example of a by pas effector we have known about for several years is the activation of the met tyrosine kinase in lung cancers that are driven by

mutations in the egfr, and treated with inhibitors of that receptor. an example of an activation of a downstream effector is recently example kras mutations in colon cancers treated with inhibitors of egfr. so these are three major

categories of resist tones kinase base therapeutics in cancer. the typical output, the typical effect of these, is to reactivate the downstream pathway and therefore lead to disease progression. in general, we haven't or don't

have a great understanding of mechanism that can go all the way around the original pathway. and cause disease progression. most of the time, you turn the pathway book downstream by some mechanism and that leads to so with this framework, we can set out to begin to understand

mechanism resist tones ras inhibition in melanoma. and the approach we have been taking is to blend systematic experimental studies, preclinical studies, which are aimed at systematic functional screens that define the universe of resistence mechanism and i'll

describe these shortly. and the idea is to blend these with the results of deep omic sequencing, deep clinical characterization of samples acquired prior to treatment and following relapse and the notion is that integration can allow us to see clinically relevant both

the spectrum of candidates that can cause resistance and a filter using clinical data to see which do cause resistence and the hope is we can leverage this knowledge to speed the design of rational therapeutic combinations and i'll show you how we are thinking about this

in melanoma. now one of the first patients in underwent sequencing and it wasn't whole exome sequencing. it was targeted. it was this patient who had widespread melanoma refractary to a variety of conventional therapeutics who had a b ras

mutation, had a dramatic response but unfortunately after really only several weeks, three months or so, there was widespread tumor relapse. now this spectrum of pictures raises many questions, one is did every tumor have the same mechanism resistence and of

course these are not answerable without autopsy-based studies. we only could sequence a specimen from one tumor but that specimen was highly informative because what came out of a mutation in mek. so mek is the kinase immediately downstream of b rav and the

mutation was conferred robust pharmacologic resistence to a rav inhibitor as well as to a mek inhibitors which had not been given but if it had been given it wouldn't have worked this this setting seen pharmacologically and recapitulated measuring

phosphorylation and these western blots here. so, in parallel, our lab has been conducted random mutagenesis screens taking cdna from mek kinase, mutagenizing that and truthing it into sensitive melanoma cell lines that have the b raf mutation and

asking are there alleles, secondary alleles to promote drug resistence in-vitro? we did wi a mek inhibitor or a raf inhibitor as the selective agent and this allowed us to get a robust distribution of individual mutations within mek itself that could confer

resistance and these fell into a couple of different categories. i'm not going to go through this in detail but what i'm showing you here is the distribution within the mek 1cdna of high frequency or robust mutations that were associated with the mek inhibitor or raf inhibitor.

and a couple of themes that emerged is there were mutations in the end terminis which contains inhibitory alpha helix, a helix, such as this q56p mutation as well as several others that were associated with the resist tones one or another inhibitor of the there were also

a whole cluster of mutations in the c helix which is component of the mek kinase has to adopt the fully active confirmation and has to become a closed confirmation. the whole series of mutations in this were able to confer resist tones one another and the skep

121 i showed you before falls in the middle of this clustner this same region. and then finally, we saw mutations in the kinase domains such as around between codons 203 and 211 that were able to confer. so they fell into functional

categories and very recently, what was quite satisfying is that a study boy a clinical team of melanoma oncologists that was presented in abstract form at as co, just this past june, has shown that if one looks at melanomas of progression compared to baseline, one sees

several instances of mek 1 mutations present at progression and absent at relapse that were identified boy the random mutagenesis screens and published several yours ago. this is quite satisfying. some of these are straightforward and easy to

interpret. one, which is the mutation and the codon 124, a leucin or syrian substitution is confusing because it can arise in association with the resistence but can be present dinovo and some cases patients can still respond.

so it's complicate whad is happening although i'm going show you data to speak to that shortly. clearly, the several of these mutations were known to cause so one of the studies we had done several years ago had been to take a patient who had been

treated with a mek inhibitor, had a b raf mutation and respond the to the mek inhibitor and we found the patient responded but whether they developed relapse were able to culture cells from this patient. they were strongly resistence to the mek inhibitor.

they this element. when you reintroduce that allele you can shift the gi50 but the magnitude of shift was less than what about seen in the short-term. now similarly if you look instead of a mek inhibitor, you look at a raf, you found a more

profound resistant affect in the short-term culture. it was quite dramatic. but if you looked at the effect of p124l, you could get a shift but it was rather marginal. now here is 256p. this is seen clinically. this was came out of a random

screen but clearly a robust shift. so p124l was less so but i'll show you more data shortly that explains the dichotomy. we found a mutation at position 203 when we did our random mutagenesis screens with the mek inhibitor but used the raf as

selective agent. each of the mutations emerged clinically were shown preclinically to be relevant to resistinence a satisfying way. we have recent data where if you look at this p124s, not l, and look tat now in the context of inducible system, so ra3

inducible pattern as opposed to steady state. there could be a means by which one 24 causes we assistance. a blend of target-based systematic preclinical studies together with deep sequencing of clinical specimens allowed us to clearly nominate a distribution

of allele that is confer resist tones these inhibitors. several of which can be shown to be relevant clinically. now the other approach has been to move away from focusing solely on a particular target but to now go to a near genome scale and to carry this out, the

notion has been to leverage resources that's been developed at the broad institute in which the large majority of genes from the human genome that we know of have been cloned to the lente viral or libraries and so the notion is to take these libraries, introduce them in

arrayed format into sensitive melanoma cells that harbor the mutation and do a phenotypic rescue screen if the presence of inhibitory concentrations of a raf inhibitor. the initial study we published a couple years ago was a pilot. and the output of that was quite

revealing for several reasons. one of which is that the major one is that there were several kinase that is scored as resistence effectors that essentially at least direct mechanistic similarities to what subsequently have been shown to be clinically relevant

there were several receptor tyrosine kinases that were safes to confer resistance and validated in independent cell lines and it's clear that receptor tyrosine kinases comprise a resistence. there are a variety of that have been described.

i won't be able to go through all of them but this captured that category quite nicely. we also found that c raf confers it is a sister to b raf and fur put it into over drive there are interesting mechanisms involving doimerrizations by which leads to resistance and i'll say a

little more about that shortly. the big novel observation in the study was a gene called cot, a kinase is a cousin to b raf. not a raf family member but phosphorylates and activates mek so this whole spectrum of kinases converged on to several mechanisms that made sense.

each of which as shown bite western blot looking at phosphorylation compared to the wildtype setting or the control setting, led to sustained erk activation. so if you aggregate the results of this, before i get to that, let me make one more point.

coming back it this p124l mutation showing this short-term culture that had discordance between what you see if you just put in mek p124l rather than the resistence phenotype. it ounce ut the drug resistence culture had whopping cod expression in addition to mek.

this suggests that multiple mechanisms resistence can exist within the same tumor within the same cancer cell. now, if we come back to this notion of mechanisms of resistence and kinase-driven therapy in cancer. and you start to overlay what we

have now learned thus far with raf inhibition in melanoma, what we can say is that although nobody has yet seen secondary mutations directly within bra, clearly raf dysregulation has been observed as a mechanism. roger low's group described mechanisms of mutations in n ras

and essentially put raf into over drive. there are b raf splicing variants that can dimerize and activate c raf and our group, i don't have time to talk about it, has identified again using random mutagenesis mutations in c l.a. that can put raf into

overdrive and there was a recent paper suggesting that some melanomas may have b raf amplification which would put into overdrive. these are target-based mechanisms and all the inhibitors in clinical use also hit c rav.

so these are target-based reck films of row assistance. the boy pas sharm explained by receptor tyrosine kinases and the doyne stream effector is sort of a mutation in mek 1. we can neatly populate these bins of resistence mechanisms by multiple individual mek that is

have been described boy us and others in the field. now, the problem here at one level is that there is no evidence right now that knowledge of individual row assistance mutations is saturating. we have no notion that we have

identified anywhere close to the full spectrum of resistences and on the one hand that seems like a problem. it seems like if we're going to have dozens of independent resistence mechanisms, distant thwart our ability to apply precision medicine, have maybe

3-4 drugs as a cocktail? and the answer could be yes, but the more optimistic view is that if you look at this more carefully, one sees that although there are multiple individual mechanisms they tend to converge on these themes and in fact, one of the themes is

reactivating map kinase and this ends up working out to be at the level of erk. there may be parsimonnuous convergences of individual mechanisms on to limiting cellular effector nosed that could be exploited -- exploited by a smaller number of clinical

combinations even though the number of individual mechanisms might go into the dozens orreds. now that notion can be tested preclinical by expanding the systematic functional screening approach out of the chinome and now to the near genome and this is what we completed.

we have used the full brode orf collection to screen a quarter of a million individual instances again using our workhorse a375 melanoma cell lineup and it is sensitive tow map kinase path way inhibitors. it was done it a near-genome scale with not just a raf

inhibitor but also a mek inhibitor or erk inhibitor or the combination of the raf and mek. the notion here is that these instances, these conditions phenotypically mimic the clinical trial list that are ongoing in the field right now.

so as many know, there are clinical trials that are ongoing testing the combination of a raf and a mek inhibitor which look promising in terms of improving the scenario. so here the goal is two fold. one is to begin to anticipate the spectrum of resistence

mechanisms that are likely to emerge even with drugs that are only just now in clinical trials where the clinical studies haven't caught up. and number 2 is to test this hypotheses that there may be a convergence on to limiting cellular nodes that even though

we might see many individual hits. so when you look at what come out of this screen thus far, we can see with this heat map that most of the genes that have scored and there are about 170 or so, are validating and you can see this because blue means

that there is no activity. these are a bunch of controls. but if the cells or the picks ils are white or red, that means that there was a percent rescued that was at a z score, essentially coefficient of variation that was high in the screen as a whole.

you can see that most, not all, but most of these are validated and in fact, many of them a substantial fraction actually are pan resistant. that means no matter where in the path way you hit, you get we can see that. we can also begin to bend these

individual hits into categories and notions signal transduction factors such as kinases which we expected to see, we are seeing very interesting additional families such as gtp exchange factors. receptors and very prominent set of transcription factors and so

already you can see based on these high level views that we we can begin to populate novel by pass mechanisms and novel downstream effectors that below erk and begin to annotate what the spectrum of resistence might be. i'm just going to briefly point

out we have also done the converse experiment not with over expression of orfs but by knocking down genes with rna i and out puts of that screen has been nf1 emerging as a prominent loss of functionect, ifor and that makes perfect sense from the genetic data.

we know that nf1 encodes a gtps gap and its loss would disreg lat signaling so it makesstancy loss would confer resistence to raf inhibition. what we like to do with this preclinical spectrum is to integrate the results of the preclinical data with the

results of deep clinical characterization. what i will tell you is that we still are not at the point in the field where there is a robust collection of many dozens of pretreatment post relapse melanoma specimen that is have been subjected to whole exome

and transcriptome sequencing. we and others are working on that but we don't have it yet. it turns out that even if you intersect these results with our preclinical data, the oaks ohm sequencing project i told you about, one learns interesting things and rather than walking

you through all of the steps of how we did the interigration, i'm going show you a picture that is emerging and you'll have to wait until we submit our paper to believe the experiments that support this. if you look at this picture, this is the kind of kerf map

kinase cascade. these are inhibitors here currently used or tested in the clinic. if you have one asterisk that means you're a gene that was significantly mutated in melanoma if you have two, it means you have been implicated

in clinical resist tones raf inhibitions. so we have our core modules here. we have genes like cot and others and now you can see that nf1, which we had seen showing up en upon riched in wildtype melanoma, there are those that

have mutations showing up on our list and also you can see that genes and sort of effectors that are in pink are hits that have come out of our orf screen and we can now begin to nominate clear modules that look like they could function as bipas effectors and we know there are

functional interactions. so receptor based signaling activate certain transcription factors known in the literature. gtps mechanism of which raf one feds into this as does p rex 2 which i don't have time to describe but we recently published on this.

so the nice thing about this is that indeed, it seems that we can readily construct a model where in the individual hits from our orf screens are not just scattered all over the genome, they are actually converging functionally or they readily fit models of signaling

cascades that we already know about that you could boy pass effectors we need to understand and indeed, could offer dependencies that are drugable in their own right. so we need to follow this up quite a bit more. but the blend of sequencing and

preclinical studies have already done are leading to to very exciting testable models. finally what we'd like to do is leverage this knowledge that we have gleaned from studies of resistence preclinical and clinically, to speed the design of rational combinationed.

if you go from this panel which is busy to a more schematic view which is shown here, we can already envision some tantalizing culminations that would be worth testing. so, the combination of raf and mek inhibitors in clinical trials it looks like it will

extend the survival but not by enough to say that we are done. but given what we know beg your pardon a number of mechanism resistence, not all, but many, it seems reasonable that testing a combination that would include a raf and an erk inhibitors with or without a mek might be of

interest. erk are in early clinical trials right now. our preclinical data tells us we can already anticipate new by pass pathway that is will either reactivate downstream transcriptional effectors or have additional mechanisms.

so in terms of the oncogenic transcriptional output, there are many transcription factors that can lab 38 output. perhaps we need to been something that can suppress there and it wouldn't be targeting transcription factors directly but perhaps something

that modifies chromatin, for example, an hdac inhibitor. those are certainly in clinical trials perhaps one should consider maybe a raf and an erk and hdac inhibitors as a triple combination. conversely we may have not nailed additional path ways

involving receptors or gtps signaling that could be considered as independent nosed. so already, from these studies, we feel like we are gravitating towards the kinds of combinations that if we could test them clinically, would be high on our priority list

besides what we have learned genetically and experimentally. let me skip this slide in the interest of time. in the last really three-4 minutes, because i want to leave time for questions, i want to come back to the ultimate test. we have talked about knowledge

of salient drivers and talked about leveraging understanding the resist tones come up with novel combinations. what about the here and now with the ingredients we have in place? this really gets down to clinical testing of the

precision medicine hypotheses and recently at the dana fasciaer and the brode, we have initiated a project which we call can seqt started with a grant that came from the acrl although it was the moan was to fund this cancer effort was supported boy nci soo are very

grateful for that. the folk us that grant is lung cancer and colorectal cancer and the notion is we will do perspective whole exome sequencing and add transcriptome on patients at the dana fasciaer and brigham and return actionable information to the

clinical care team. this is now being added and in addition we are launch something studies with eric in breast cancer and studies together with george in sarcoma. we have a 5-pronged approach to can seq thus far in which specimens are retrieved at the

dana farber and brigham and dna is extracted and in some cases rna as well and the report will be made available to the treating oncologist. so, briefly we have a series of algorithms that have been developed and we are reiterating and boy a collaboration with two

medical oncology fellows, ellie van ellen and nick, in the lab, and so there are a variety of approaches where we can both bend the samples based on actionability and investigate their relevance and also begin to look at variance among certain significants and map

those and we have developed a report that is online and actually we plan to make this available broadly quite soon because it's more or less ready to go. but we have ways in which the committee evaluates this rapidly click and see the content of

what is coming out of exomes. there have been a variety of interesting exome that is have emerged thus far. some of which have multiple drugable alterations and others that have variance of significants but they occur in drugable oncogenes and there are

still others where we see plausibly actionable alteration that is are off the beaten path such as bcl6 mutations which we didn't know about before. and this is a sobering remind they're as we throw the kitchen sink at samples genomically, we find still case where is there

is very the to do from a national standpoint. we see only people with mutations and nothing else or this striking case of a sarcoma that had become resist want to radiation. we saw no alterations that were actionable.

there is not going to be the end all but it's a great way to get us started. let me in the interest of time, close by coming back to this ark type but pieces of this ark type are clearly implementable now and being implemented in many cancer centres around the

country and around the world where we have begun with the patient and we are hoping to blend biopsies with profiling and novel approaches to data interpretation and ultimately measuring response and resistence in this approach. the knowledge that emerges can

be blended with preclinical studies, companion preclinical experimental studies to refine our understanding on the biological end and also to refine our thinking about high priority novel therapeutic combination that might have a great chance at achieving

durable control of many different cancer types. and i like tow close by thinking a number of people who have been involved. i'll highlight eran who led the melanoma landscape study together with eon watson and linda's lab and corey and ellie

and steven who has done the rnai screening work. and with that i'll take questions if there is time. >> you began your talk by saying there are a lot of mutations and probably you're not responsible for melanoma and ended by saying there are a whole variety of

different boy pass alterations that will go pass some initial growth mow meeting pathways and you got a lot of data based on transfection experiments and knockout experiments and so on. so obviously it would be great to know in that patient who had hundreds of melanomas and all

which became resistant, but lacking that information is the -- there are any indication that the genes that give you alternative pathways to boy pas resistence are in any way related to the genes that come up from the initial melanoma? any correspondes to that?

>> that's a great question. so the question is whether or not are the mutations associated with resistence different than the steady state or or overlap? there is a lot of overlap. nras mutations are seen at baseline in 15-20% of melanomas and n ras the same mutations can

arise as a mechanism of acquired mek one mutations are seen at probably about 5-10% frequency at the steady state but clearly associated with acquired resistence to melanoma thus far, what we have seen genetically overlaps strongly with the steady state which speaks to

what you're saying. undoubtedly a lot of heterogeneity we are not measuring. so what we see across many individuals can be active at various points during the biology. so i think an important point to

emphasize here is that we're not really throwing them out. we are just saying if you want to discover the drivers, you have to have a scenario where you're not being confounded by passengers. but it's not to say that all the genes that were thrown out are

at 121 samples, this is a statistical test so limited by power and the higher the mutation rate in the cancer type, the more samples you need to sequence to be equivalently powered. something that may be is a passenger at one point could be

acted on boy evolution later in the cancer. >> nice talk. ure mentioned that the -- [ indiscernible ] >> it's the analogous residue. so in rac 1 it's codon 29 and in cdc42 it's the analogous codon of so it's exactly functionally

the same mutation. >> sos are those mutations [ indiscernible ] >> so we have not seen the equivalent to codon 12 and codon 61 mutated in melanoma but people have made the mutations in raf one. one point i should make and i

haven't looked at the specific resdo yous but the l.a. one mutation is a cdc transitionition -- the raf 1. and so you can imagine that mutations that are cdc transitioned that are activating are more likely to be seen. i than codon 12 is not a cdc

transition it's a transversion. so it may be that some of these mutations which are activating in other context are just statistically les likely to occur in melanoma and given that rac one mutations are already low frequency events. just even though they may

theoretically occur, it is just below the frequency of detection we have 120. >> one more quick question. >> so when you take these melanoma cells with mutation in rac, can you test it whether or not the activity for >> great question.

we have a couple of that have raf 1 mutations. we don't have cd42 as a cell line. and the short answer is, you can detect raf activation but to say it's high compared to other cells is actually thus far maybe you were just our assays are not

sensitive enough. we can't say in the steady state it's greater than what you mute see without the mutation. inch dodginess can be tricky -- endogenous. >> [ indiscernible ] >> so i think now with the latest algorithm, we haven't

removed them because if you're comparing with within a sample exon to intron mutation rate it doesn't really -- it's a ratio so it doesn't really hurt you. but there have been other genome projects where one sees a certain distribution mutation rate and then outliers where rue

moving the outliers has been helpful because otherwise you see mutations all over the place. particularly when you're talking about low frequency events, you worried it's being skewed by one or two samples. >> so i want to get back to

michael's question as it relates to inherent to acquired. so you have all these uvb mutations and we see them in our mouse models as well. i don't know it causes all this this one shot or multiple times or whatever. the mutation thank you's see in

the recurrent disease, did they tend to be uvb mutations? are they things you think are lurking around in some cells and then they become more important in resist innocence. >> so, i think hopefully in the not too distant future we will have enough relapsing specimens

where i can give you a answer. but the short answer is, yes. the short answer is, we have not globally seen a different distribution and mutation. i think what really the key things is going to be when we hone in on the mutations that are different in the post

relapse compared to the pretreatment at a exome level. does the delta, is it still have a high cdt transition that i can't really answer yet. we haven't done that analysis. but very soon we should be able to answer that. i think it's a very important

question. >> i want to look at your problem in a very simple way. the lessons we have learned that if you have targeting upstream compliment of pathway which is relevant to cancer, we should expect downstream acquired mutation on some of

the -- discern disen hopefully we should be able to predict the mutation so is it appropriate to say that we have decided to target upstream compliment or gene that we should get ready for targeting downstream gene because in variably we will get that mutation?

>> i think it's an interesting general theme that one could lob up would be to say, if we have either predicted or experimentally that yes this downstream pathway is always a mechanism resistence, why not go in with an upfront combination that hits the index oncoprotein

and the downstream pathway? that's essentially what the raf mek combination attempts to do partially but maybe in lung cancer and other cancers one should try that as well. i think it's a very interesting and people are certainly -- >>

worth trying. i think we are out of time for questions if i read you correctly. >> let's thank dr. levi garraway for an exciting talk and helping us started our series this year so well we have a reception that is supported through the

foundation for the advanced education of science in the library and a chance to chat more informy. inform