Colon Cancer Cell Lines

[ silence ] >> so first speaker of today's from ucsf,kevan shokat. he's a chair of cellular and molecular pharmacology. he's also affiliationwith department of chemistry at berkeley. he's a member of the national academy of sciences,the institute of medicine, the american academy of arts and sciences and i think all of thetypicals, there are two things that are two things that actually show can kevan shokatis in this type of research, he's a passionate researcher. i think he will never give upuntil it's done. utilizing chemistry synthesis and protein engineering to investigate signaltransduction pathways into to really understand how do they communicate with each other, buthow also can you utilize them to design new

very intelligent drugs. and in recently ithink some started to shed some light how this can be done. but he's also an entrepreneurand bringing his discoveries to clinical market which i think is important because at theend we only use it to treat patient. so the floors to you for your talk on small moleculeinhibitor of kras g12c. >> thank you, [inaudible], thank you verymuch. well, great. well, welcome. that's for--thanks for laura, frank, kevin and martin for puttingtogether the meeting. and before i start i want to make sure that sort of bring out onething frank said that it'd be great if during the meeting people interrupted with any questionsand we sort of the next day and a half we really dig in to ras i think in a way thatfrank really envision so don't hesitate. you

don't have to raise your hand, just startspeaking over me and interrupt me. i would love that. i would love that. and i also wantsomebody to keep track of how many times frank says his favorite word. i think you all knowwhat that word is so, all right. we're going to have a counter up here. so we will havea world cup scores and then frank's tabulation. all right. well, i want to build on many ofthe points that frank made and i think we all aware of how important mutations are indriving cancer and how important of a signal that is for guiding toward discovery. we knowfrom much of the success of targeted therapies that once a kinase is activated in cancersuch as bcr-abl that provides a fantastic direction from the distal chemistry and drugdiscovery, creating a drug like gleevec which

shuts down the pathway that is aberrantlyactivated in the tumor cells and then by shutting that off because the tumor cell is dependentupon that signaling pathway, the tumor cells, apoptos and we have a fantastic therapeuticindex. i think we all know how this is a much, much oversimplified picture of how lucky weare that drugs work in this way and one point that i think we often forget about is thatwe don't necessarily understand the true nature of oncogene addiction and why this cell istruly dependent upon the activity of this mutant oncogene and cannot survive when thatoncogene is inhibited, and i think julian downward will tell us more about oncogeneaddiction and we'll understand how much more we have to understand there. from a chemicalstandpoint one thing that this slide sort

of glosses over is that the inhibitors thatblocked the mutant proteins, also bind to proto-oncogenes. so other thing this--successof this kind of strategy relies on is that the inhibition of the non-mutated kinase andthe rest of the body has to be tolerated at a dose that affectively kills the tumor cellsand shrinks the tumor. so because small molecule chemistry in the kinase area has not beenable to be successfully developed just the mutant protein and leave behind them non-mutantprotein we have to rely on sort of oncogene addiction and the ability to tolerate inhibitionof the pathway. so that's a chemical challenge that still has not been--been met in the kinasefield. the other feature is, is that we know these are proteins that operate in pathwaysand if we take a pathway view, then we should

be able to just inhibit anything in the pathwayto block the aberrant activation of that pathway. think of statins targeting hmg coa reductaseand cholesterol. that's not the point of the pathway that's mutated and on for examplebut it works very, very well but signaling cascades are much more intricate and complicatedand as i'll show you, and as we'll see in many of the talks here how targeting evenone step below the oncogene but right in the same pathway is very, very much differentthan targeting the exact oncogene. so, there are much--many more intricacies of how thepathway works and how the oncogene collapsed the pathway that we have to take into accountwhen we think of targeting the drug. but the two points i want to make here is that thesimplest and most direct way would be to make

a drug that only inhibited the mutant oncogeneand targeted that mutant oncogene level of the pathway rather than somewhere upstreamor downstream. other strategies might work but we will see that much more complicated.so kinases are mutated but also as frank mentioned we're here about ras. how are these two nucleotidedependent enzymes working? at a very basic level they both co-op and use phosphate asa switch. in the kinase area, increasing the kinase activity is what mutants--mutationsthat drive cancer do because activating the kinase is what transfer the phosphate to substratewhich go on to signal downstream. so this is a situation where in the unactivated case,kinase has have many ways of keeping themselves off and then the oncogene sort of releasethat off state and push towards the on-state.

nucleotide dependent enzymes like the gtpaseshowever, operate very, very differently, almost in an opposite sense. they are nucleotidestate dependence switches so in the gtp state , they're held down in that very tight confirmationthat frank alluded to serving as a switch to produce a surface of protein that is complementaryto effector proteins that are downstream. and in this case then the job of the non-mutatedprotein and normal signaling is to activate the gtpase activity, releasing the phosphateand then relaxing the protein so it's unable to bind. so in one case, we're going to findmutants that drive activity, catalytic activity and in one case mutations that cost a lostof catalytic activity. ok, now if you understand how a gtpase can act upstream and turn ona kinase what you can then also readily appreciate

as frank mentioned is just by following twoamino acids, really two proteins in one signaling cascade the ras/raf/mek/erk pathway, you canappreciate almost 30--i understand how 30 percent of cancer is produced. so in the krascase, we mutate up the glycine 12 or 13 or 61 position that leaves ras predominantlyin the gtp state producing the surface that binds and recruits and the raf kinase whichhas a ras binding, effector domain localizing it to the membrane and activating it and thisis much, much more simple, i will see many, many more examples of how complicated thisin fact, is and frank already alluded to how there's many pieces of the puzzle we stilldon't understand. in a mutually exclusive way, even without a mutation, at the ras level,raf can be mutated at the valine 600 position,

immediately downstream to activate the pathwayat that level and that explains another large fraction of cancers. so with that understandingof how ras activates the pathway and how raf has been activated independently and we knowhow druggable kinases are--we've all seen in the last 15-20 years of drug discovery,you might think that we should go to the list of approved kinase inhibitors, find the rafinhibitor because raf is below ras, and pick out vemurafenib or now dabrafenib and usethat to treat ras mutant tumors. it works in the pathway, it's immediately downstreami think by all of the early logic we have even five years ago, ten years ago. it shouldhave been a fantastic approach. well, you need a fantastic drug luckily flexicon andearlier many, many mek inhibitors were developed

so we had excellent kinase inhibitors downstreamto target the kinases that are activated by mutant ras. and not only is a good moleculein vitro but the early clinical success of vemurafenib in patient that had v600e mutationswas astounding. so, from the flexicon data we knew that about 81 percent response ratein phase i, in metastatic melanoma patients that have v600e mutations, pet scans beforeand after therapy have fantastic responses. so, it's a fantastic drug, clinically veryactive, inhibits the pathway exactly like we would expect. so everything is lookingvery optimistic and my point here is what happens when we take this into a ras mutantsetting? well, interestingly, unexpectedly in a very surprising way, neal rosen at thetime of the phase i clinical trial learned

and told me that in phase i there were patientsthat were benefiting their metastatic melanoma tumors were regressing but about 31 percentof patient in phase i were developing keratoacanthomas another skin tumor and this were in sitesthat didn't seem to have--have tumors previously. so this is completely paradoxical, how wouldthat work? neal's lab in his postdoc at that point glucose developed a cell line modelwhich mimic this potentially--this surprising paradoxical activation of the ras/raf pathwayby the raf inhibitor. by--i showed you this side of the gel where cells that have a rafmutation treated with plx-4720 block mek and block erk, that as we would expect but thesurprising effect is if the compound is added to a cell that a ras mutation rather thana raf mutation. so only wild type raf. you

see a huge activation of mek phosphorylationand erk phosphorylation. so how would that work? how does an inhibitor of a kinase activatethe downstream pathway? and the genentech group and neal rosen and our group collaboratedindependently to reveal a mechanism that might explain how a kinase inhibitor that targetswild type raf in the setting of a ras mutation could activate the pathway. and the way webelieve this works is that as i told you there are intricate mechanisms for kinases to maintainthemselves in the off-state and as frank mentioned there are many important steps we still don'tappreciate but one thing we basically agree on is that wild type raf in the off stateis in some state of a monomer state and then it becomes activated when ras is activatedand in that dimeric state, further activates

and mek and erk. we'll hear from debra morrison[assumed spelling] more about how much more complicate this can be. but what we realizedwas that kinases are not just simple catalysts that chug out protein--chug out their productall the time that they're in this two state, the off state and the on state. and what wereason that that kinase inhibitor although would inhibit the activity, it could alsochange the confirmation when it bound to the kinase. and in that new confirmation it mightpromote the dimerization, and then at certain intermediate concentrations, intermediateconcentrations that were achieved clinically you might be half occupying one of the protomerand activating the other unoccupied protomer which would lead to hyper activation. so this,i think illustrates two points. the first

is is that with ras mutations, there is alittle bit of a trickle of activity sometimes and an inhibitor of a downstream kinase canfurther accelerate that so that illustrates the point that going even one step away fromthe oncogene can be very detrimental, in this case completely paradoxical, not just--notwork but it can actually amplified the pathway. so that's that--the sort of bad thing. i'dsay the silver lining however, is that small molecules can do surprising things in complicatedsignaling networks because these enzymes are regulated at so many levels. so if somebodyhad told you, "hey, in this disease, i want to turn on the activity of a kinase, i don'twant to turn it off. let's say, ok, that seems tough. but then they told you, the only thingyou can use is an inhibitor of the enzyme

then you would say, "ok, that's really crazy,i can't even begin to understand i'll do that" yet that's exactly how this works. so i thinkthe point i like to make is that is with careful structural biology, careful cellular biochemistry,we can really understand new opportunities for chemical transformation of these proteinsto elicit the kind of therapeutic beneficial affects we want. so even though this is causinganother cancer, and largely i think that's mitigated now by combination with mek inhibitorsas you might already imagine, it really tells us that the amazing sort of gymnastics thatsmall molecule inhibitors can elicit in these pathways might give us new opportunities.i think we need completely new ways to think about chemistry where it attack some of theseproblems. yeah?

[ inaudible remark ] i think that people believe it has a--theyhave a ras mutation in them and that they go into in senescence state and then the inhibitorpushes them over but maybe martin-- >> yes, about 60, it reported 60 percent ofthe keratoacanthoma have an hras mutation. >> hras preexisting, great. >> that is how hpv-- >> either way, they--oh ok. and do those--thehpv ones become also driven by that raf inhibitors? >> yes, they do. >> ok, so they'll multiple anything that willtrickle to give a little bit of signal, the

raf inhibitor will push. all right. ok. so,so, a little bit back on ras and the oncogenes, this is just a timeline of sort of the discovery,fantastic, early investigation of tumor viruses and the isolation of ras and sark is primarilyoncogene. and what i think is amazing from the work back then is that although we haveno reason to expect that those would be the prime oncogenes that we're driving so muchcancer, they did turn out to be this sort of the prime drivers. and with focusing onthose early oncogenes we've been able to understand immediately one binds gtp, one binds atp,one is a kinase, one is gtpase, then all the different regulators, structures were solvedand this is amazing that ras is identified very early then turned out to be the mostfrequently mutated oncogene in cancer. so

i think this--as frank said prompted muchearly you know, efforts on targeting ras and what's also surprising is that the kinasemutants have been much, much more successfully targeted. we now have over 20 different kinaseinhibitors that target oncogenic kinases that have been approved. and although the farnesyltransferaseinhibitors and the early kras focus turned out not to be as beneficial, i think we'llhear from herbert boldman [assumed spelling] that the idea of mislocalizing ras by relyingon even new aspect of how ras farnesylation traffics in the cell is going to be a veryexciting new direction as frank mentioned that maybe a focus on hras by this farnesyltransferaseinhibitors could be useful. there's a lot of excitement still on the farnesylation,just not as simple as we thought. so what

i'd like to focus on thought, what i talkedabout in the first two, first slide is that we really need a kind of chemical approachthat focuses just on the mutant oncogene if we could achieve that we would have the chancefor a huge therapeutic index because we wouldn't be targeting ras in the rest of the body.we would target just the mutant that is produced in the tumor cell. so that's one challenge.the second is i focused on a lot is that because of the nature of the difference between agtpase and a kinase and how the enzymatic cycle works, the competition for the nucleotidegtp is much, much difficult in the gtpase's case than it is in the kinase case, will muchmore difficult? is it really that big of a problem? well, if we look at the comparisonof kinases and gtpases, the real key number

here is what is the difference in kd betweenatp for kinases and gtp for kras and it's about a million fold. so that means we wouldhave to be a million fold better in our chemistry than we have been in all of the kinase fieldto come up with the molecule what competes for gtp. that's possible, maybe, very, very,very, difficult. i think probably the only strategy that has a chance at that is onethat nathaniel grays' lab that dana farber is pursuing which relies on a covalent bondto compete out the nucleotide. that is the one i think window and so it's great thatpeople are pursuing that but that still is a very, very high bar but i think that isthe largely the reason for the low. yeah, that's a great question.

>> dr. shokat can you repeat a question? >> oh sure. >> we couldn't hear it [inaudible]. >> sure, sure. so, frank sort of said, well,why can't we turn that around on itself? maybe a drug that looked like gtp would have alsopicomolar affinity and then it would compete. the problem is is that, we have not--phosphatecontaining drugs are very poor and they're from a single properties getting across thecell, prodrugs [phonetic] don't work great so there is that--that problem that. and theother problem is nobody has every found a good enough surrogate substitute for phosphate.there's something about two negative charges,

tetrahedral and it's very, very hard to getthat to go. even a single phosphate like in sh2 domains, very, very hard. so to thinkabout three phosphates is tough. >> --like a hundred thousand peptides, [inaudible]came out, it was about 60 picomolar. [inaudible] was the best thing you get when you modifygtp is something we can achieve. >> yeah, yeah. >> could you repeat [inaudible]. >> so herbert said, years ago, they trieda chemistry around gtp, you started with gtp and added peptide to the base or the phosphateand you basically came within five folds, six folds of this but could never go betterthan that, yeah, yeah.

>> hey, man, if you could just follow up alittle more on this. even if you imagine the thought experiment of an extremely high affinity,deep you know, small molecule. >> yeah. >> whereas the therapeutic index for all ofthe normal proteins in the cell, [inaudible]. >> --cycle [inaudible] like drug and justwipe out all kinds of-- >> right, yeah, so kevin brings up the pointthat a gtp drug would just block all the gtpases and you didn't have much of a therapeuticindex and that goes to this question. i think that's a little let. it's a sort of mitigatedmaybe a little bit by the kinase experience. i mean kinases we have an even larger problemthat we have atp mimetic drugs. we have 500

kinases, surprisingly once you find that mimicyou can eke out selectively even if every amino acid that contacts the atp is the same.so it's a good. i mean it's something to worry about but i sort of feel, it's better to put.it's better not to close off any road. it's good if somebody have a good idea to replacethree phosphates, start there then probably by decorating, you might be able to get selectivity.so get potency first then selectivity. >> in phosphates is what you mean. >> two, two should be enough, sorry, yeah,let's make it simpler. take one away. yeah, because a three would keep on. so exactlywe want it off. we want it off very, good, very good. so, now we'll get into the rascycle a little bit. i'm sure we'll hear about

this much more, more detail but what appreciateand i told you about so far is that ras is in the gdp state which is off and it is sortof in the relaxed state and switch 1 and switch 2 are in the open confirmation when it bindsgtp, the switch 1, switch 2 are tied down. now, other ras effectors that helped convert,and are convert from the gdp to the gtp state are enzymes called guanine exchange factorsthat bind to open up the nucleotide pockets so that gdp can fall out and then by virtueby the ten fold higher concentration of gtp then gdp in the cell, gtp will become loadedand the cycle will go in the forward direction. once it's in the gtp state they'll be theintrinsic gtp's activity as well as the gdp is activating protein that introduces an arginineto complement the transition state to greatly

accelerate the conversion back. then the two,classes or examples of effectors that i've showed you before, the raf interaction here,with the gtp state as well as pi3 kinase. so many, many people have looked at druggingevery aspect of this. drug companies have put the entire cycle together, run the entiredeck of compounds through and it's been very, very difficult to come up with molecules thatcan interrupt this cycle but we'll see opportunities from steve later in this portion how someof these complexes provide other pockets that can manipulate the activity of the proteinwhich are very exciting directions. so how do the mutations cause the transformation?if we put the localized--the g12, g13, glutamine 61, on the surface of the inactive form ofthe protein and the active form you see that

in the inactive state the residues are farapart but they cluster very close to one another in the act of state suggesting that theseresidues are important in actually disrupting this active state of the protein. how doesthe 61 position cause transformation. so, in this close up you can see that we havea crystal structure of gdp in it bound with aluminum fluoride mimicking the gamma phosphateand in gray and light blue is ras and then the gap in the lighter blue color is the gapputting in the arginine finger. so glutamine-61 is aiding the catalytic step by activatingwater for attack on the mimic here of the gamma phosphate so mutating this catalyticresidue is going to destroy catalytic activity dramatically, and carla mattis [assumed spelling]will--can tell us much more details about

this. if we flip the protein around and highlightwhere the glycine 12 and glycine 13 are, you see that they'd lined the side chain of thearginine from the arginine finger so named by frank and harry bourne [assumed spelling]here at ucsf is complementing catalysis and what's important to note is that the sidechains are absence in the glycine and so any side chain, any mutation to any residue largelywill clash with the arginine, misaligning the arginine from the gap, greatly decreasingthe gap, facilitated hydrolysis of gtp. so that's the basic mechanism that neutralizedthe transformation, the biochemical inactivation of the protein. so what--what we decided tolook at is that since any amino acid can sit at the twelve or thirteen position, and disruptthis catalytic mechanism, what kind of features

are accessible chemically by the differentresidues that are put there? so here are some of the pie charts sort of like what frankshowed before about the dominance of aspartate, the most frequent glycine 12 mutation, thenext most is valine but the cysteine is a significant fraction of all kras mutationsand we if looked in lung cancer it's actually the most frequent residue that is replacedfor glycine 12. and in this lung cancer situation we know that a lot of that preference forcysteine or valine is driven by smoking-induced mutations because in current smokers, thecysteine and valine are the predominant mutations and the never smokers that's in the aspartateso there's some amount of carcinogen induced specific allele specificity but as frank mentioned,there's a lot, lot to understand about why,

why glycine gets mutated to what it does getmutated in various cancers. for our purposes we saw the cysteine as a chemical handle thatis basically to a biological chemist, cysteine is like, you know, honey to a bear. i meanthat is like the one thing that will save you. it's like water in a desert. i don'tknow if you know, if you know what the analogies are but you need that. so that cysteine was--goingto be the focus of our effort. as we solve the crystal structure of the glycine 12 tocysteine and the gtp state, we expected to show the cysteine was exposed to the surfaceof the protein. here's a switch 1 it was ordered in the inactive state, switch 2 was disorderedand in fact some of the residues cannot be defined as our many of the cases of gdp structuresof the gtpases. and then--so how do we find

a drug that uniquely relies on attachmentto the cysteine? very, very luckily for me, my colleague here jim wills [assumed spelling]for many years developed a very, very elegant chemical strategy for focusing on moleculesthat attached to cysteine and identified pockets neighboring that are adjacent to the cysteine.and so what this relies on is a clever chemical trick which is an inner conversion betweenseveral different disulfides. so we start with the protein with a cysteine that is eitheris a natural one or is oncogenetically induced or is introduced into the protein. and inthe presence of beta-mercaptoethanol there's an inner conversion between the cysteine andthe disulfide of the beta-mercaptoethanol. and then through synthesis of a collectionof molecules that have different functional

groups, little different fragment, piecesof drugs, small molecular weight that have a sulfide that are in disulfide with beta-mercaptoethanol,there's an exchange. now if one of the features in this molecule are complementary to a shapethat is adjacent to the cysteine on the protein then these disulfides will become relativelyresistant to reduction and therefore they will build up in concentration and beta-mercaptoethanolwill be out-competed by a specific drug. what's great about this technology is that the earlyleads in drug discovery in the fragment case which are small molecular weight moleculesof 100 or 200 molecular weight don't have enough intrinsic affinity for you to pickthem up very easily by noncovalent mechanism. steve fesik's group developed elegant biophysicalmethods using sar by nmr to identify pockets

using noncovalent chemistry but that is--imean he enabled it when he was at abbott and that is a herculean task. now it's been unableto [inaudible] in many places but this technique is a very nice method because you know immediatelythat the molecule in the collection that bound is what you think it is because you measurethe intact protein mass of the entire complex. so using that approach we collaborated withjim's lab and identified several hits from his collection that bound to the kras in theg12c form of the protein but the wild type form which has several other cysteines didnot bind at all. so these are two features, one in dark grey, one in light grey. if weintroduce the cysteine into hras we can also get binding even though we didn't screen itagainst hras. so that says this pocket is

somewhat conserved. if we take g12c ras andput it into gtp state and we ask if these molecules can form the disulfide only minimallyif at all. so that was actually very, very depressing because we have worked so hardto get a molecule that bound to the cysteine, that bound very selectively to just the cysteinewe wanted but when we put the protein in the active state where we think the state of theprotein is in the tumor we see really, really no binding. no worries, let's just keep goinghoping that some little bit of unexpected luck will come our way and also we said immediately,we screened against the g12c in this gdp state, we screen the collection so we said immediatelylet's go back and screen in the gtp state. we did that screen, we came up with zero.so that's when we really said we have no chance--we

couldn't abandon these molecules, we haveto keep going forward. so we then solve the crystal structure of this molecule bound toras and at this point it could have bound in the nucleotide pocket. this covalent binderit could compete out with picomolar gtp or it could bind somewhere else and very, verybeautifully it bound in a pocket adjacent to the nucleotide so we don't have to competewith the picomolar affinity and it looks like it has a pocket that usually was unappreciatedbefore because of the flexibility of the switch 2 in the gdp state. so although we knew thatgdp--the switch 2 has to fold in to this pocket, our drug is binding effectively what is theswitch 2 pocket in the gtp state but not drug--without the drug there this pocket is really not fullyformed. so with that as a guide we started

to optimize the chemistry and think how thismolecule would disrupt ras function. and this is a little animation of that inhibitor boundstructure, switch 2, switch 1 and the conformational changes that would happen in the absence ofthe drug. in the absence of the drug switch 2 would fold into that pocket producing thesurface for ras binding and activation. so our first inclination was that we had a switch2 pocket binder and we would disrupt any interaction with ras that require switch 2. so that seemedfine to us but we're still thinking about the gdp state problem, you know. is this goingto be a killer for us and so we actually had a very nice guide, it's always nice to seethat nature discovers something before you did because then maybe you have a chance ofhaving yours to work and there is a natural

product that binds to the gtpase gq, very,very potent inhibitor, very, very selective. when its crystal structure was solved it wasshown to bind to the gdp state of this gtpase and bind under switch 1 so that's very, veryanalogous to our gdp binding switch 2 pocket binder. so that reassured us. the other featureis that right before we published our work, the genentech group and steve fesik's groupidentified using sar by nmr another pocket in ras that was behind switch 1 and switch2 and what we liked about our pocket is that it was closer to the catalytic machinery andso even though we have this gdp problem we at least appreciated that our pocket was closerto the catalytic machinery of the protein. so we push forward on that basis as well.so then the question is converting the disulfide

to a carbon-based electrophile that can workin cells so we converted and kept the same reversible binding piece and then convertedto electrophiles at either vinyl sulfonamides which are kind of too reactive for a realdrug or acrylamides that occur now relatively commonly in kinase inhibitors that hit non-conservedcysteines. so we start with reactive groups and we incubate in a sort of very standardizedway, 10 micromolar, 24-hour incubation, this molecule gives 18 percent modification butthrough subtle changes to that blue ring, we can increase right up to 80 percent andthen we can even go to less reactive electrophiles and then further basically combine, mix andmatch all of the pieces to get the optimal compounds. so this was about an effort fromthat disulfides to the optimized compounds

i'll show you of about 200 molecules thatwe synthesized. yeah. >> so in the absence of the electrophilicwarhead, do these molecules bind to a ras? >> not that we can appreciatively measure.they must bind without the electrophile but it's probably in the, you know, tens of millimolarrange too difficult for us or too big to do a sort--treat as a fragment and bind. i thinksteve one time i talked mentioned that things, using things bind in this pocket in sar bynmr screens. >> that's actually in my talk. i'll-- >> you will, ok. >> i'll mention it.

>> ok. sorry--yeah? >> yeah, with covalent inhibitors you havethis within traditional [inaudible] covalent inhibitors. you usually optimized both kiand kns >> in this situation your enzyme is essentiallydead so you can't--you--and we can't do the exact same experiments but yeah. >> right. >> [inaudible] are you optimizing reactivityor is there actually an affinity components that you're improving by making this one [inaudible]. >> can you repeat the question?

>> yeah, yeah. so she--this question is thatyou know in traditional drug discovery you would maybe start with a reversible bindingmolecule, get the molecule in there and then change the warhead and keep tuning that inorder to get faster and faster inactivation. we have a harder time in this case becauseit's--we're trying to optimize essentially both at the same time. so what we've doneis we've basically abandoned--we sort of kept with one electrophile so we kept with theacrylamide and then we've systematically investigated the reversible binding mode in the contextof the irreversible reaction. so we don't really have a good measure of the pure reversiblebinding yet. i think probably people at well spring are getting closer to that but we don'thave that yet. it's a great question.

>> may i interrupt for you a minute? >> oh sorry. >> so we have a handheld mic out because thepeople who are streaming and people who are in [inaudible] if you have a question pleaseinterrupt anytime during the talk or raise your hands so we can give you the handheldmic so that the people who are streaming can hear it. >> great. >> but we also has one in the back. >> you've got five to ten minutes.

>> what's that? >> i have five to ten minutes. no problem,i got that. that sounds like one minute. i could even do that. so we had a molecule,binds to ras, how does it disrupt the function? i told you disrupted switch 2 confirmationso immediately asked does it--should disrupt the gef, that ability to do the exchange reactionbecause it relies on contact with switch 2. so in a mant exchange reaction we have gdp-mant.we add either sos or edta as the chemical exchange mediator and we ask for loss of fluorescenceby competition with non-labeled gdp and so the g12c mutant without any added sos or edtahas a very slow exchange reaction. if we add sos, we have very accelerated exchange. ifwe add edta and yank out the magnesium it's

instantaneous. now we put our drug on andwe see that we blocked completely the sos-mediated exchange though we have no effect with edtabecause that is independent of switch 2 just pulling out the magnesium. so we learned thatwe could block the gef-mediated exchange. i think that's one area i'm very interestedin understanding is whether the mutant rases really rely on the gef very much. i thinkthat's one of the basic kind of questions frank highlighted. there are many, many unansweredquestions. i think that's a very important one to think about. the other thing we startedto realize when we looked at the crystal structures was that if we look at the textbook, a picturethat frank showed of the--wait, still on further? there we go. release the phosphate and theswitch 1 and switch 2 opens up. we realize

that our compounds sitting under switch 2is actually further disrupting the confirmation of switch 2 residues. and if we looked closerand asked well, what would happen when gtp would try to come in to the drug-bound state?we could see that glycine 60 in switch 2, threonine 35 and tyrosine 32 which have thehydroxyls that key into the gamma phosphate and make that confirmation change would bedisrupted by the presence of the drug. so they'll start to move in but if the drug ishere you'll see that the glycine 60 in particular would not be able to complement the phosphateposition in gtp. so this made the prediction that our molecule could disrupt gtp affinityin preference over gdp. and with similar exchange reaction we can titrate in gdp in wild typeg12c or two different drug-bound forms and

the gdp affinity is unaffected by the presenceof the drug. but if we compare now and titrate in gtp we see wild type and mutant have highaffinity. this is sort of limited by the concentration of protein we have so it's not the true affinitythat's why it doesn't look like picomolar. but then with the drug-bound we see a rightwardshift and it is lower affinity for gtp. this is fantastic because it kind of unraveledone of the puzzles that we had at the beginning is that how can we compete with the nucleotide17 picomolar? well, what this drug really does is it doesn't compete for the entirenucleotide, it just disrupts the gamma phospate which is the achilles' heel of ras and ifthat's true then the molecule should have some ability to kill g12c cells in preferenceover other alleles and so we took a number

of lung cancer cells in colons that had g12cmutations in kras versus serine, valine or aspartate. and although it's not perfect withthis early stage molecule you can see very nice dose-dependent killing and the g12c cellsare uniformly more sensitive than the non-g12c cells. and for example, this cell has onecopy of g12c, this copy has six copies of g12c so there's some--we think there's somereasonable hypothesis about why cells that have g12c are differentially sensitive althoughone of them like 23 is not particularly sensitive. so we still have a lot of understand aboutspecificity but importantly i think we always want to see a very, very good correlationbetween biochemical ability to bind to the protein and the cellular effects and so what'svery important is that these reactive molecules

could have very nonspecific effects so wewant to compare specific electrophiles like this compound which binds to g12c at 100 percentand compare it to a molecule that has the same warhead but doesn't bind at all by virtueof poor reversible binding and then there's an intermediate molecule here, number 10.and the rank order in cells should be 12, 10, 17 and that's 12, 10, 17 and completelyinactive. so that is all very, very gratifying in [inaudible] that we believe are on target.so now it's a game of optimizing the chemistry here for more and more potency and the like.so to summarize what we've identified as a molecule that keys in to just the oncogene,blocks the gef-mediated exchange and since it can't load gtp we think that we're usingthe cellular gdp as our inhibitor in keeping

the protein in this state and that will thenpreclude it from binding to factors. so the summary of the challenges that i set out wereinhibiting just the oncogene. i think inhibiting just the g12c versus the wild type solvesthat problem and we thought we would have to compete with the nucleotide but reallyall we needed to compete with was the gamma phosphate. and so with that, the people whodid this work were too fantastic, people in the lab, postdoc ult peters [assumed spelling]and graduate student john ostrom [assumed spelling] who've both gone on now and martinsos who we put on the project only because his name was one of the ras effectors andthen a new student danny gentile [assumed spelling], so a fantastic group and reallycouldn't have done this without jim wells

and a lot of help from this postdoc jeff sidowsky[assumed spelling]. we started a company with frank to pursue this in the clinic and we'repushing that forward now so thank you very much. [ applause ]