Colon Cancer Picture

>> the good afternoon. we'll get started. i wanted to mention for those who have asked me, you know, if you attend 50% of the sessions during the course of the year and you take the final exam which is a multiple choice electronic exam that you can

take several times. it's nothing you should lose sleep over. but if you pass that or when you pass that, you'll get an electronic certificate to say that you took this course in demystifying medicine which may be of some, hopefully, some

value to you. at any rate, the final exam will be electronically available by the end of the month hopefully a little bit earlier. i'm trying to write an exam that i can pass. so today's subject, drug resistance in cancer certainly

needs no introduction. it's a beautiful example of the sort of theme of this whole course of the brooklyn bridge and connecting what was two different disciplines. and in this case, of course, the discipline of clinical oncology and the big problems of cancer

or biology, macro biology. okay. so our two speakers today are michael gottesman and antonio fojo. i was kind of thinking on walking over here that it was almost 30 years ago, michael and i had dinner in holland after a

meeting. i was at the albert einstein college of medicine. i did not know much about the nih except i admired him a great deal. he told me he was appointed as director of intramural research. i didn't have a clue what all

that meant but it sounded very important. it took a long time. this is not a job, this is the job. this is one of the most difficult jobs i would visualize would ever hope to be. it blends knowledge patients and

michael has been the director of the intramural research since 1976 and we are up deeply indebted to him for all of his efforts. now, as a physician scientist, michael graduated from harvard, took his medical training there, trained in molecular biology

here with dr. geller. went back to harvard, came back to nih and has been here ever since. but michael is the chief of the laboratory of cell biology in the nci. and that has been one of the certainly predominant preeminent

laboratories in the world for studying drug resistance, particularly in cancer and understanding more and more about mechanisms and development of ways to get around this problem which in cancer is perhaps the inevitable and fatal step in the disease.

so i was thinking about mauricio on walking over here and you'll permit me for telling another brief story. a long time ago, a famous conductor bruno walter was due to conduct the new york philharmonic and everyone came to hear him but he was sick with

the flu. and a young protege named leonard bernstein took the podium for the very first time. people began to think about walter afterwards but not with the same gusto because leonard bernstein swept away the field and became one of the preeminent

people in modern music and other such things. so i'm not suggesting that bruno is replacing -- ever grateful to mauricio from getting off a plane from a meeting in panama and he's going to discuss some very exciting work which he's been doing.

he graduated from medical school in chile, has been here at nih only for two years now or close to two years. he's a fellow in oncology and has done remarkable work, his bibliography is very impressive since he's been here and we're very grateful to you for both

substituting and on your own. and so people may ask you questions hopefully in the middle of your presentations. we welcome you to do it. so some of the questions that go through my mind, and i put them up only to stimulate those of you, particularly students and

fellows who don't think of this all the time, is some of the old problems is drug resistance intrinsic in cancer cell. is it due to [indiscernible] that survive after therapy, and if so, really, how does that happen fundmental cellular biology.

the the other side of the bridge, how do you determine whether cancer is responding a solid tumor, where it's responding through therapy. how do you do this in vivo. how good of measurements that are made. we're going to hear more about

that. i suppose the big question is why are responses in solid cancers, maybe there are some exceptions. but for the general reason, they're not curative. why is it? if we step back and take a look

at it in the big picture which is what we're going to do this afternoon, hopefully there will be some new thoughts and ideas, and from you as well. this is perhaps maybe the single biggest problem in oncology today. i had one other slide which i'll

just show briefly, and those of you who were here during john coffin's el elegant discussion of hiv a few weeks ago may recognize this slide. this was the point that the hiv virus decays in basically four phases. after starting somebody on

anti-retroviral therapy. you notice way over from my far left phase four, those cells which are as little as 50 or less per milliliter of blood. this is why a person's on multidrug suppressive therapy. those cells persist and there's an infinity sign there because

nobody knows how long they persist. they've been clearly identified. but ... well, i managed to shut it off. oh, there. if you stop therapy, the white bar, it's this clone of long-lived cells that contain an

integrated retrovirus that goes way up when therapy is interrupted. very interesting. you sort of wonder, i'll just throw this out perhaps but some of the discussion of the speakers whether drug resistance clones in the case of cancer,

they require a new drug to try and get around it, i guess. but there were similarity of these with hiv infected cells that maybe there's something intrinsic in the cell. for the hiv, it's the integration site near responsive to a variety of promoters.

it is conceivable that there's something like that in a cancer cell that actually renders it drug resistent along with, at any rate, you didn't come to hear me. so the first speaker is michael. thank you very much. >> thank you.

let me just say by way of clarification that i had been at nih since this coming 1976 but i've been in intramural research since 1993 which is long enough. so for 20 years. >> [indiscernible] >> you may have, yes, but i was not yet.

so i was struck, i'm always struck by the image of this bridge, the brooklyn bridge is being constructed and part of the roadway is not yet. when we began to work on drug resistance in cancer we discovered a gene which is responsible for pumping out

anti-cancer drugs and had a very broad specificity or lack of specificity. we have a major cause of drug resistance in cancer and we thought the roadway was almost complete and clinical studies began and so on. i would say now 25 years later,

we removed some of pieces of the roadway and there are bigger gaps between the science and the translation. and i think both i or mauricio will tell you what the problems are. but of course, a better understanding comes from wisdom

and it certainly come the kind of modesty we reallyshould use in approaching these difficult biological problems. so what i would like to do is sort of give you some background information, give you a sense of where the field is, and then let, we'll answer questions

obviously. and then let mauricio give you some more detail about some real clinical situations and how the clinical information we gathered helps us focus on one or more of the difficult models that have been proposed to explain drug resistance.

so, if you don't, if you fall asleep and it is late in the day, i'll understand if that happens. what i would like to point out is that this slide probably summarizes everything you need to know about drug resistance. first thing is that as you know,

oncologists have been very interested in developing drugs that are targeted to specific oncogenes. we've been sequencing cancer genomes now for many years. and many cases we know what the driving genes are that are responsible for the cancer.

if you target those specific gene products with specific drugs and big pharma is happy to produce these drugs, you should be able to block the cancer promoting pathway. this turns out to actually work reasonably well in chronic my logic leukemia on the surface of

the cell that drives the cancer. there don't seem to be a lot of other ways that that qurns can exist. if you inhibit that you get very long term responses. in some cases resistance develops -- or in pathways that are down stream from the target

that bypass the target. so this is what i would call a fairly straightforward problem in drug resistance and it works in understanding resistance of cnl through the i don't believe but it doesn't work as well for the other targeted drugs. you see the system developed and

it's usually not straightforward. there are other ways of resistance developed multiple drugs which is called multidrug resistance and some of them i would say are in the realm of cellular farm call or organismal pharmacology which gets pumped

out of the cell or metabolized in the cell or the body, all of these are pharmacologic mechanisms that affect pharmacodynamic and pharmacokinetic. when we started our work in the 1980's not very much was known about these mechanisms and we

were able to discover that energy-dependent drug, drug efflux is an important mechanism for handling a lot of the drugs that are anti-cancer drugs. and a variety of other cellular pharmacologic techniques are important as well. this kind of wisdom has come

from observing patients in the clinic who are responding and then failing to respond to chemotherapy. resistance can result from what i would call changes in the life-style of the cell. so a cell that grows [indiscernible] that's blocked

by anti-conference drug and another one chooses another growth pathway. or the changes in pathways that are normally associated with differentiation and one when has been very much studied in this area is called emt or epithelial parenchymal transition where a

cell that looked epithelial is now growing as a sort of parenchymal and they have different patterns of sensitivity and resistance to drugs and that drives their resistance patterns. or in many cases resistance can develop because we're dealing

with a tumor that may be homogeneous to begin with but are breast cancer and there are cell types that are -- and those subtypes each have different patterns of sensitivity or resistance to drugs. or a tumor cell can develop along pathways which are

starting from the same cell but are somewhat different depending on the initiated factor. so it's the biology really that determines in the third case that resistance. it's this kind of problem that has led many people to kind of throw up their hands.

mauricio and i were at a symposium at physical science and oncology just a few days ago at nih. we were on a panel in which the number of members of the panel said the problem is insoluble because it's intrinsic to the nature of the biology of tumor

cells. we don't believe that you're not going to hear that negative message today but i think hopefully we'll convey you some of what the complexity is of this problem. so there are certain factors that lead to the development of

and one which i think is only in the last few years been fully appreciated. i mean, you know remember when we learned about cancer, we were taught that there was a single cell that underwent malignant transformation and that's how it was a colonial population to

develop fairly homogeneous population of cancer cells. and if you believe that, then you believe that you should be able, one drug or combination of drugs to wipe out all those cells because they're so similar to each other. the reality is that if you look

at cancer in detail, in molecular detail, including sequencing of the dna of individual cells as a cancer, each cell is quite significantly different. and they are different for a variety of reasons. some of it has to do with the

fact that there are just mutations arising in cancer. fact that the expression of genes in those cancers changes epi genetic cancers. life will not survive if we didn't have really good ways of dealing with toxic products in the environment.

so all of these add to the difficulty when you're starting with a fairly large tumor population and finding a drug or a combination of drawings that will kill all of the cells. so my talk will have four parts and i'll go through this pretty quickly.

the first part has to do with these transporters that we've been working on now for 25 or 30 years. what will they play in drug secondly, a kind of technical note about one of the reasons that we've had these moments epiphany thinking we understood

drug resistance and we don't, that has to do with the model system we're using which is not always relevant system for studying cancer. then i'll talk a little bit about some work we've done sort of preliminary work we've done looking at expression of drug

resistance genes in clinical cancers to show you that even with simplifying assumption that the genes we're interested in are important genes, that the complexity of actual tumors is something quite amazing. and then a little bit, thank you about models introduced in

mauricio's talk that might be important for clinical cancer. so this is just a summary of what i already said to you in terms of cellular mechanisms that could lead to resistance. just to give you some idea of the scope of this problem, we know in the human there are 386

transporters that are called solute carriers. there are [indiscernible] plausibly about 10% of those on the order of 40 are involved in the uptake of drugs into cancer cells and into normal cells as well. on the toxicity cancer drugs

probably relate to the expression of these genes in pick organs or tissues that are sensitive to the drug. and it appears that in order for a cancer cell to be sensitive to a specific drug the drug needs to get into the cancer. we used to think it was simply a

matter of a hydrophobic drug across the membrane but more and more evidence exists that there are specific transporters. those transporters expect that higher levels are easier for drugs to get in. they can determine the sensitivity or resistance of

tumor cells or normal tissue to grow. there are also these exflux pumps which were discovered as a mentioned around 25 years ago. and these include a number of nbc atp binding transporters and i'll talk extensively about that in a moment.

several of these are involved in export of drugs themselves. the same way that uptake transporters can keep drugs out or let them in. the e flux comes also to rules the accumulation of drugs in the cell. then if the drug does get in and

frequently they can because the pharmaceutical companies are looking for drug agents to get into cells. there's a whole variety of changes that go through the more detail involving regulation cell growth and apoptosis. repair damage.

this is particularly important. a lot of anti-cancer drugs damage dna and there's whole bed of important dna damage repair systems that play a role in drug and compartmentalization which sometimes i call sequestration which is a terrifying word. so what i'm going to do now for

a while is focus on these efflux pumps which i'll get back to in a moment when i show that this is not the single most important explanation for drug resistance. so the atp binding family is a big family. there are 48 different varieties in the human.

if you look at all published results, there are about 2000 members that have been identified many of which have unique characteristics. the major one is that they use atp to powell drugs and other compounds out of cells. these include ion sugars glycoin

ns. these are shown here. this is a family of human atp binding proteins and this is a figure made by mike dean who has been studying the -- some of the transporters consist of the six transmembrane domains and atp binding cassettes.

the form of transporter has two of these six trans membrane domains and atp binning cassettes. but you see againations of various types and i'll show you some more examples in a moment. there are seven different varieties, families of these

transporters. and drug resistance transporters exist in several of the different families. so there's nothing unique about one of the families. the families are determined mostly by homology within these transmembrane regions since the

atp binding cassettes are pretty similar throughout all the different members. that's what makes this a family. there are some features that are kind of interesting because there's a fair amount of homology these are difficult to detect without micro ray and

i'll talk about that in a moment. i would like to focus on these transporters and i chose these because are the broadest spectrum transporters. if you select for cancer cell and culture that are resistent to a whole variety of drugs

these are the transporters that tend to come up. abcb1 which was caned by victor ling where p stood for permeability and now p stands for pump because it's not a drug that regulates, it's not a transporter that regulates -- but pumping thing out.

and this has these domains and two htp binding cassettes. the abc family multidrug resistance 13 one is example has a terminal extension of five transmembrane units. you can cut this off and it's still an active transporter. we're not sure exactly what this

is doing. it may have something to do with cell localization. and then there's the abcg2 which looks like half molecule except the atp binding is on the terminal side and this works as a homo dimer to her form two regions a total of transmembrane

domains. we think that basic mechanism is similar for all three of them. let's specificity varies a little bit is shown in the diagram. abc1 recognizes all of these substrates and many more. they recognize these and g2

recognize these and they all recognize these drugs in common. all of which are anti-cancer drugs. the -- those are shared in the human. if you look in other organisms for example the mouse we see some of different specificity.

in fact the mouse has two abcd1 gene so it has four gene products that are involved. all of which is a complexity that makes it difficult for companies that are using engineered mice and so on to be able to study this problem a little bit at a loss.

so if you look at the 1280 amino acids that make up this transporter in a plainer structure it looks something like this. the yellow represents. specific point mutations in p glycoprotein that changes the substrate specificity.

so then though it represents hundreds of drugs you can make a mutation in any of these sites and you get changes in drugs that are recognized and just beginning to try to understand why that is. there's a region called a link region which is phosphorylated

but the phosphorylation doesn't seem to regulate function. you can eliminate the sites or dephosphorylate it and still work with the transporter. the green bars represent affinity labeling studies with end up labeling the transmembrane region which led

us to believe that both regions were involved in drug binding. if you knock out either atp site, you knock out transport function. and so we felt these two parts probably came together and we came up in the 1980's with this model based on the sequence in

some of the biochemistry. the most amazing thing about this model, if you take, and i don't have with me the high resolution em, prior em picture. looks exactly like this c glycoprotein actually looks like this at a resolution of about 10 to 20.

what we think is happening is because of the hydrophobic anyway of the drug, they're actually being recognized within the plasma membrane. atp is hydrolyzed and i'll show you a little more detail in a moment and these drugs are being pumped out.

the important thing of recognizing the bilayer that explains the lack of specificity that there are hundreds of thousands of drugs and the central feature has to do with the -- of the drug. some other features as well but there's no clear pharmaco for

it. it's the physical property that determines the substrate of these pumps. these are higher resolution pictures that are based on crystal graphic studies that have been done. the mouse, mbr1a gene which is a

hall log of the human abc1 gene has been crystallized by steve and jeff to relatively low resolution. but they were able to crystallize the protein with a drug bound. this is the bilayer here and you can see the drug is indeed found

in the lipid bilayer. and it's found in a region which is the same region that i defined before as being part of the molecule which is -- so the notion here is the drugs get into a hydrophobic pit, they have some specific interactions with amino acids depending on

the nature of the drugs. when atp is bound, the two halves of the mouse come together. this is called the open confirmation. this is a closed confirmation. and what that does is extrude the drug from the transporter

and then atp is hydrolyzed and resets the pumps so it can pump again. the precise details are being studied a number of labs in our lab specifically by [indiscernible] who could give you more detail about how it actually works.

so when we first discovered that c glycoprotein was a major efflux pump in cancer cells that were drug resistent. we said where does this come from. it's normally spread in tissues. in pack it is. not uniformly in the body but in

tissues that are a barrier so it's present in the small and large intestine pumping into the lumen of those organisms. it's present in the liver. and hepatocytes that are lining the billiary deduct tiles so it pumps into the bile and into the intestines.

in the kidney it's approximal -- tubules and into the urine. in cells, and i have here circulating cd4 cells that express p glycoprotein. and just to get back to what nguyen told you earlier about the existence of the waish to hiv, anti-retroviral theorem

some of the drugs used to treat hiv are really the substrates for p lyco protein. evidence that there's a subset of cells that circulate that express p glyco protein. some of the resistance of the virus may be due to the cellular resistance to these compounds.

and then there are cells that are in the interstitial space, in particular in this case tumor cells that can express pg. what i want to talk about is pgp and thesal multidrug transporters is in the brain, the capillary and endothelial cells out of the brain.

the p glycoprotein and other transporters as well. maybe i have -- no. i don't. basically the pgp normal function is a barrier function. it keeps bad, puts thing out of the body and keeps thing from getting absorbed into the body.

so just a couple words about the blood brain barrier because this turns out to be rather important in treating brain tumors, for example getting drugs into the brain. a lot of people who are studying neurobiology in trying to cure diseases in the central nervous

system are trying to get thing into the brain. a lot of drugs don't get into the brain. and why is that. if you look at this graph which shows the solubility of drugs, that is their -- along this line this get more and more

hydrophobic. this dissolve better in oxynol. in comparison to uptake into the brain, you can see a lot of compounds behave failurey normally. the more hide phobics they are the more they get into the the compounds fall off this

curve and those include glucose which gets into the brain quite readily despite the fact that it's not very hydrophobic. and the reason for that is clearly there's glucose transporters that pumps glucose and then there are all these drugs -- cyclosporin -- they all

occur in anti-cancer drugs and hydrophobic products. the fact they are hydrophobic don't get into the brain and the reason is because of this blood brain barrier which is made up of first of all capillary endothelial cells junctions so things can't leak into the

and then there are these pump systems whenever anything tries to dissolve into the brain gets pumped back. so the same system that protects cancer cells from hydrophobic toxic anti-cancer drugs seem to be involved in this important blood brain barrier and this is

another picture showing the hydrophobic vacuum cleaner that tries to get into the cells and is recognized. so matt in the lab has been particularly interested in understanding how to image the blood brain barrier with the idea of being able to circumvent

its activity, particularly if you want to get drugs into the brain and neurological diseases and also for example cancer. and so we set out with bob inus and victor pike who are colleagues at -- has anyone talked about -- here? anyway, it's a very sensitive

way of detecting molecules in different parts of the body. and so we made a derivative of a compound called low pure made. it's what you buy over the counteras e mode annual -- it's a potent opiate and the reason is it's not intoxicative. you don't get addicted to it

because it can't get into the the reason is because it's a substrate -- you can see here these are cell lines in which pgp is expressed or not expressed. abcg2 or c16789 and the other multidrug transporters are only that presence of abcb1 you

see this big reduction in accumulation of the compounds which are pet labels. if you look at knockout mice that are lacking abcd1 you see a big increase in uptake in the brain of this compound. whereas in the g2 and c1 knockouts you don't see uptake.

so this is another bit of data that suggests in the mouse that uptake is dependent on b1. and so you can use a resin derivative of -- as an imaging agent injected into mice. if you do that, you see that this is the mouse brain and to show you this is a ct scan.

the mouse has a small brain but it has a brain and this is the brain and there's nothing in the brain until you add an inhibitor of p glycoprotein which is in this case is a [indiscernible] and then you get nice uptake of material into the brain. so that shows that the system is

working. you can do the same experiment in a monkey, the dramatic uptake in the brain. in the human you just see an increase in uptake. it's not as dramatic as it is here because the drug we use, we can't give in high enough doses

not because the drug is toxic but the agent used to dissolve it is fairly toxic. and so we're still working on finding a better way to deliver but you can see clearly substantial increase in uptake and this actually points out on you difficult it is in the human

to totally block p glycoprotein expression. and pretty high levels of potent inhibitors which is one of the reasons why some of the clinical trials to reverse drug resistance probably had not been so successful because they hadn't done a good job of

reversing pgp expression. so i just thought i would end this part on p glycoprotein by analyzing the data we currently have that does support a role for p glycoprotein or abcb1 in cancer. the first thing is that it's expressed in human cancer at

levels that are quite sufficient for drug resistance. so at least 50% of human cancer's expression the reasonably high levels of pgp. there's data showing that after you treat certain tumors with drugs that are substrates for p glycoprotein, the recurrent

tumors, the relapse tumors express higher levels of p glycoprotein. this is true for leukemias, myeloma -- cancer. you can for example on -- show increased response to anti-cancer therapy if you inhibit p glycoprotein.

but for many of these tumors that's not the case and even though pgp goes up, inhibiting it doesn't solve the resistance problem. there are a bunch of tumors which express pgp at the time of diagnosis including those derived from cell types that

normally, press pgp remember i mentioned in the small and large bowels pgp is expressioned colon cancer is expressed, kidney cancer is expressed pgp, pancreatic and liver cancer -- frequently at quite high levels and those are intrinsically resistant tumors.

then you can make animal models with human cancers or models which are endogenous to the mouse in which for example you can use p53 mutations in brca1 mutations to get breast cancers in mice. when you do that and select the drug resistance you get pgp

expression and you can inhibit pgp expression and you get response to the tumors and they grow back and they're resistent but not because of pgp. so the conclusion of all of this, and the reason why i said the roadway is getting more and more, the roadway was nice and

clear and so we started doing studies. it's the problem of getting -- probably should avoid it. pg glycoprotein is sufficient. every model suggests that. i show you we have a transgenic mouse we put in pgp. those cells that express it are

drug resistent. it may not be necessary. there are many other causes of resistent. own when it's expressed it may not be the only mechanism. that's the clinical problem. so we went back to the drawing board and we said okay it's not

going to be pgp. since we're interested in studying drug resistance we should probably figure out what it is. we went to the literature and we found about 380 drug resistance genes that people described over the last 20 or 30 years.

we talked to our colleagues and we put together a set of these genes using a system called a low dense array. to measure very accurately and precisely the rna in both tissue culture cells and tumor samples for each of these 380 genes. the genes involve an uptake of

drugs, sequestration of drugs, metabolism efflux of drugs. genes involved in pathway of -- synthesis which seemed to be involved in drug resistance. pathways involved in uptake of drugs into the new close, dna repair. and all of these genes involved

in regulating cell growth and so on. and we have 380 genes and we could easily distinguished them, we could measure them with high specificity over a long linear range and at very low levels of and so the first thing we did we said well let's take a look at

our cell lines we've been using to study drug resistance and see whether those are good models for tumors. so we got tumor samples, we started with ovarian cancer for reasons i'll tell you in a moment because it seemed to be a good model for both intrinsic

and acquired resistance. it's a disease which responds to chemotherapy but all too often the measures get resistent and can die. we got a lot of ovarian cancers and those are shown here. this is p slot. these are the 380 different

genes, the level of expression. red means high level of green means low level. and these are about a hundred different ovarian cancers. and what you look for is the pattern of gene expression to see whether the ovarian cancer actually have pattern of gene

some are different from the cell line we're using as models. and you can see that pretty similar to each other, these are ascites sumers which are slightly different histological kind of tumor. these are the cell lines we've been using, 40 different ovarian

cancer cell lines. their pattern of expression of these 380 gene is totally we've been depending on these to understand drug resistence but they don't seem to be representative of the tumor cells that are coming out of the patient.

so you said is this the kind of general problem or one you need for ovarian cancer. we looked at melanoma. these are the tumor, these are the cell lines. we looked at breast cancer. these are the tumors, these are we looked at -- tumors in the

colon cancer. interestingly different patterns but these are the cell lines, these are the tumors. so essentially every tumor we looked at with maybe the exception of hepatoma. we've had these discussions before.

this is a tumor derived from the liver and i think the liver cell already expresses a lot of these interesting drug resistence genes so when it gets established in cell culture it doesn't change its mode of behavior so continuously express the genes.

whereas these other cell types lead to the back of the tissue culture. so just basically says that our current models are not very good, we need better models. why do we need better models. so when we established cells in tissue culture, i have to say a

lot of the cell line that we use have been in tissue culture almost forever. literally longer than i've been in this field. 30, 40, 50 years. so they established themselves in tissue culture. people think we should human

cells growing for 30 years and culture should be a different species. everything about them is really gene expression is very and they are particularly different with respect to these genes that we're starting that relate to environmental response

to environmental adversity. so we think that what survives is probably either a small subset of the original tumor or more likely something which is mutation and adaptation changed the way of living by being a tissue culture. the other thing about culture,

the way we culture monolayer cultures is the culture condition is very different. we used -- when the oxygen tension in tissues is much lower -- versus 3t cullure. in tumors there's other cell times and there's more and more evidence a lot of the behavior

of tumors relate to other normal cells which are probably not normal that surround the tumor. and of course growth factors. and finally, we're not happy unless we grow cells very quickly in tissue culture. we want them to double every 12 hours, every 16 hours, maybe

every 20 hours. normal tumor don't grow that most of the time they're not growing because if they did, can you imagine how very quickly those tumors would overwhelm us. so we set up conditions that are quite abnormal in terms of forcing themselves to grow and

of course it's not surprising that they express genes that are very differ from those that are expressed in normal ones. these things work against us so we decided we would try to start developing a better system for growing tumors. more like organ culture ex vivo

from the body and try to grow them in conditions that were as close as possible and succeed them under those conditions. i was very fortunate to have a post doc in the lab -- bio engineer and is now in medical school. and we worked with -- and his

colleagues at -- who are bioengineers. the idea was to develop artificial capillaries so that we could create oxygen great -- gradients so you imagine the capillaries running through the capillaries they are just

diffusing away and the further you get the less oxygen there is. the starting concentrations of oxygens are actually quite low. so the idea was to use materials that would allow oxygen to diffuse and we initially came up with a silicon hydro gel.

if you wear contact lenses, contact lenses are made to silicon hydro gel and are easy to mold and they are very permeable to oxygen which is why they are used to make. so we made these things that are capillary size. this is a photo technique that

ashley developed. and the idea is that oxygen in this chamber diffuses up from the bottom. it goes through these artificial capillaries and diffuses away and forms grade -- gradients. if you eliminate the gradient -- little round tumor nodules is a

typical way. the growth is quite different. if you look at an individual tumor growing in capillary and use a compound called -- which is a histochemical stain that detects lower oxygen concentrations. so the staining is for high poxy

tissues. you can see a gradient of oxygen away from the capillaries which mimic the oxygen gradients in these systems and is very close to what you see in actual tissues and artificial capillaries. we think we have a good system

and we're just starting to look at real tumors in this system. and we hope to be able to measure gene expression as a function of distance in the you can see immediately another reason why tumors are hereto genesis. they are different throughout

these tumors. that's one of the drivers probably for heterogeneity. so very quickly we've been looking at three different tumor types representing intrinsic or acquired resistance. so let me just say intrinsic resistance is a tumor that just

doesn't respond to chemotherapy to begin with. acquired resistance is the tumor seems to respond and then relapses with resistance. we never knew and i'm not sure i know now whether these are different mechanisms or two sides of the same coin.

i think mauricio is going to make an argument that acquired resistance is -- acquired resistance resides within the initial tumor. hepatoma responds well to chemotherapy and acute myelogenous leukemia which mostly does respond but in many

patients relapse has become studying this tumor because it's a liquid tumor we could get samples from patients before and after therapy in the same patient which is very difficult for the other two. what we did, we are very fortunate to have an

outstanding -- we accurately measured expression levels so it's 380 genes. we said are there any genes whose expression in ovarian cancer predicts response to therapy. and we knew the response in various tumors.

and came up with 11 genes, all of which were important statistical contributors. together they were the best predictors of whether the tumor will respond or not to therapy. just to remind you, therapy and ovarian cancer. most patients present with stage

three or four disease in the abdomen. the therapy consists usually of a paxal-like compound and a compound which is -- and so on. therapy is pretty much the same around the world. and so each of these genes represents different mechanisms

or resistance -- detoxifying gene. there are apoptosis -- genes involved in metastasis, invasiveness. the patient comes, we can say the likelihood that they respond or not based on the gene expression pattern.

the pack we could find a subset of genes not a single gene suggested that either ovarian cancer is multifactorial. that is all of these resistent genes are contributing. or that all of them simply represent a pattern of resistance but there's one

predominant gene which could be different in different tumors and we would not see that with the statistical approach that we took. a third possibility is that cancer's a different gene expression can arise from different origin.

if you know anything about ovarian cancer you know there's a big argument about the origin of ovarian cancer. it's possible that cancers from one origin are more responsive than cancer from another origin -- or the way it became a so just quickly in hepatoma,

there's a 45 began signature, i showed you an 11 gene signature for ovarian cancer is more hepatomawe independently confirmed this and again there's two different cells of origin or different ways in which hepatoma can develop. aml, we know that p glycoprotein

expression is shown to correlate with poor expawnls but when we look at different samples before and after we see increased expression of pgp but increased expression of g2 or increased expression of c1 or any of the other agency transporters or any of the other drug resistance.

every tumor we look at that's become resistent has a different pattern of gene expression. so we suggest that the basis of resistance may be somewhat different in each cancer. this is the argument for personalized medicine but this is the depressing part of this

study. if every patient has a different reason for resist ex, it's going to be pretty hard up front to figure out what to do about we'll have to wait and see what develops and then we'll have to work with that information. really quickly, a couple models

to account important what we're seeing here. in acquired resistance, what we're saying is we're using two different drugs which are ovarian cancer, blue drug and a green drug. most of the cells seem to be sensitive to the blue drug and

the green drug. but there are some resistent cells in this heterogeneous population to begin with. if you use drug a the cells are blue. mixed in here are multiresist even drugs that are resistent to everything.

same argument with drug green. and if you add them together which is what we do in ovarian cancer, you can kill most of the cells but the ones at that time survive may be the multidrugs. and similarly, in terms of intrinsic resistance, you can say that up front you have a

heterogeneous population of some of the heterogeneous populations are more sensitive cells than other ones. so for example, in this population, there are a lot of resistent cells but in this population there aren't too many resistent ones.

you get a good response but unfortunately in a lot of patients the tumors are back. so just a final thought because i think this may stimulate some discussion if you are interested. so we've been using successfully natural products anti-cancer

drugs to treat cancer. in fact the most successful drugs we have are natural product drugs even though a lot of them are advances that have been made in more targeted so the idea here is that these drugs have evolved and they are like plants and micro organisms.

may evolve over the year to kill this is part of the warfare between plants and animals. so they know how to kill cells. we're not exactly sure how they work. one of the thing we talk about is not so much what the basis of resistance is but the basis of

sensitivity is but they can kill they target multiple pathways that have also evolved over time. to preserve -- extreme environment conditions. we are alive because we have a lot of mechanisms for resisting these toxic things in the

environment. in our cleverness to create targeted drugs, we are targeting single pathways and that these generally don't work and take care of cancer because there are these compensatory mechanisms that allow some subset of cells even in the face of this kind of

onslaught. so that's just a final thought. maybe mauricio wants to comment on that as well. i have a lot of people to think and i'll just say the clinical work was done with the help of -- we have a lot of clinical collaborators -- was responsible

for the biostatistics. i've had a good fortune for many years to work with, the crystallographer and other people in the clab that contributed in many ways that count. i'm happy to end there and we can either take questions now or

wait until after. >> you mention the mdr and pointed out other mechanisms but mdr is the focus of a lot of our attention and you mentioned that there's clearly a session to generate a resistent. what do we know about regulation or expression of mdr and how can

we address that issue? >> right. so we know quite a law. quite a lot. the main way it has influenced anti-cancer therapy is because pharmaceutical companies screen all their new compounds against mdr cell lines.

and they always hope to find compounds that will get into cells that won't be able to be pound out by mdr. so a lot of drugs entering the market are alrea screened to halt be mdr substrates. so that's the big contribution. so that eliminates p

glycoprotein as an important mechanism. what we know about regulation, there are two kinds of regulation. the regulation that's developmental. so these tissues that i mentioned, the epithelial cells

of the gi tract, the hepatocytes and the epithelial cells and the placenta and other places. all are developmentally regulating expression of pgp and there's -- that are essential for regulation under those conditions. in addition, there's pretty good

evidence both certainly in tissue culture and also in actual tumors that pgp is acutely induced by toxic compounds. not only substrate but compound that damage dna that results in pgp expression. they seem to be under control of

these spectors that are involved in turning on gene that protect against -- and again a little bit is known about the binding sites and the promoter region and so on. it's kind of a very interesting area because if it turn out that resistence is due to

simultaneous expression many different resistence mechanisms and those are relatively universal control. there's a single transcription factor or there's a couple factors. those then become really interesting targets for

regulating response. and not enough is known to know whether that tru or whether the regulation is going to be as complicated as the genes that promotel resistance. good question. >> -- good hydrophobic substrates not interfere with

what you're trying to accomplish in terms of [indiscernible]. >> okay. so, it's a really interesting question. so the discussion is broad but obviously there aren't too many, i all get asked, you know, it can't be that the natural

substrates are all endogenous -- compounds that are substrates. there's a whole field that believes -- was an important component of the structure. we don't see it in the crystal structure but the resolution isn't quite high enough to see it actually, that pgp is

recognizing and binding cholesterol is part of it because it does recognize steroids and pump steroids frequent. so statens are wonderful substrates of p glycoprotein. so the answer is, probably in evolution, it was important that

it not pump normal counsel that are present in the lipid bilayer but recognize things that are abnormal. i think it's the kind of surveillance system for making sure that bad stuff doesn't get into the membrane. once we settled on having

membrane, we wanted to be sure that we could clean those. that's why i call it a hydrophobic vacuum cleaner, a little bit controversial. but the other point is that is every one of the compounds that binds to this binds to or is transported by pgp is the

potential inhibitor or module later. so there are millions of different inhibitors. there's no lack of inhibitors. anything which is non-toxic becomes non-toxic. so that's why we know that just inhibiting it isn't sufficient

probably to reverse drug >> what's interesting is that nine of the ten top candidates are causing acute liver injury in man are antibiotic derivatives of micro organisms. which don't affect the organism. >> some of them are -- some of them are pgp substrates, some of

them are both. actually, site control p450 which is metabolizing, the -- derivative has a drug specificity which is almost identical to pgp. they overlap is enormous. so here's a transporter and a metabolizing enzyme but tre's

no sequence -- and probably not much -- regnizing these compounds at least two different ways. save your questions, we'll have more time at the end and mauricio. thank you very much, mike. [applause]

>> thank you for the -- in this afternoon. i will talk about a number of models for cancer that might be better treatments now in the clinical setting in treating patients. you can remember in the 60's and 70's was the exciting times for

oncology and chemotherapy for the -- because tried a combination -- in many patients. in other places, other institutions it starts to be wondering what would be the best way to reach to the best combination or the best drug and really starts to be -- therapy.

that's america's important role or objective between institutions. and i want to tell this story that -- told me. all of the -- who criteria for clinical trials are coming from this doctor, dr. moertel in 1976.

and he did an experiment -- with 16 oncologists and 12 spheres and 1920 measurements. what he did is there were 12 spheres from 1.8 to 14.5 centimeters in diameter. and they ask for each oncologies in that meeting to measure these spheres.

the thing is that two pairs were equal in size. the number five and six and seven and eight. by the up -- oncologists --what happened 25% of the time when they have to realize that the difference was more than 25%, they did a mistake.

and two different oncologists measured this and it has to be more than 50%. this 6.5. because the difference between two spheres 25% and 60% is larger. and then the same oncologies to measure and did the mistake for

this -- only 7% of the cases. and in 25% and 19% of the cases. that is really the base of the data uho criteria for -- response for solid tumors nowadays with some difference that we'll talk about. and i say from the beginnings -- looking for really practical

reasons but not really for -- assessment of efficacy of drugs in oncology. and the w.h.o. then was involved we use now in clinical trial here and in many places with some subtle difference in how we measure tumors and not only dimensional measurements of the

tumor. some difference in -- that are based in progressing of his tumors, 50% versus 30%. but really are based on the same experience. here you can see the subtle difference between the final volume when us -- has

progressed. you have to be 10% increase or 40% increase in the volume of the sphere or in the tumor to say that a patient has to rest. and now, with the new -- it's 173% more, a more relax way to define progression. before, do assigned progression

before than now but really is almost the same. it's just some evolution -- it has been 37 years since moertel study and should we be thinking of different ways of assessing efficacy is the main thing. i will start with kidney cancer which is one of the models we

use to try to understand a better way to assess efficacy of this is a simple diagram to explain how the new drugs for cancer in the last ten years has been in addition according to all the notations -- somatic -- that cause the accumulation of these transcription factor in

the cytoplasm that is increasing the transcription of -- many genes related to proliferation and angiogenesis. one of the drugs that was first used for like a targeted agent was -- that is used and approved by the fda for kidney cancer. and it's a drug that can cause

chronic disease, disease stabilizing agent. but really you see -- each of these line is one patient and each line is the best response of each patient. and you can see really 74% of the patient has some decrease in the size of this tumor.

really decrease the size of the tumor, the thing is the threshold for the partial response was not reached. for that reason, it's called stabilizing agent -- but really all you see is the -- of the cells in this drug. for that reason, many years

ago -- how would be a better way to assess these difference between drugs targeted agents, called targeted agents and traditional agents. this is the equation that dr. -- is physics and -- that works with a the doctor for many try to model the -- and

regression of tumors. and here you can see this is the y axis is the tumor measurements that you can see in a daily basis [indiscernible] and this is -- for example if you have a patient that responds to therapy you have regression of the tumor and the size and measurement.

and then resistance appears and you have a growth of the tumor [indiscernible] and the patient is in a clinical trial is put off the clinical trial. but this is in the way that this curve is composed for two component. a component that is the tumor

that is sensitive to the drug that is under treatment and a fraction of the tumor that is -- from the beginning. this is the concept that -- and in these two, we can construct this curve and this curve fits this equation, this mathematical equation that really is

sustained the site of the tumor at any point in time if you following the patient under treatment -- potential exponential of the decay how the tumor, the sensitive tumor decrease with treatment. and how the tumor is growing and these two are constant and the

valueables are really the time and their treatment. the mean es one is the value for the start the equation for the number one that is the start of the treatment measurement. and this is the first analysis that they did i think like seven or eight years ago.

this was phase two trial at nci and drf -- against placebo in the form of a study renal cancer just for you that you don't know -- that targets a gross factor in the blood that targets the receptor of the cell membrane and really open deed that this growth factor

activates this receptors that are involved in angiogenesis -- but also target the same receptor. and this is the -- curve of that paper of dr. -- that shows that the progression free survival was longer in the measures -- you can see here is the patients

that progressed and begin to have any patients progress 100% and over time you see the patients start to progress at the end you can do a calculation and you can say that it's significant in this case that the value was significant for saying it was better with a high

dose -- and then the question was what happened with the data you really can see the equation that i already show to you. and these are four examples. this is a patient with the type of tumor that only progress with the treatment. this is a patient that has an a

response and the tumor starts to grow and this is a patient that more follow up the tumor starts to regress and start to grow again with you more slowly each month. this is a facia that has data that don't fit this equation. this happens more or less -- the

decay part of the equation and you can have and you can have with the growth part of the occasional. and then they are comparing the growth rate constant between the two arm. you can see here the high dose the system of arm has more lower

growth rate constant. this is lower means slower. and compared to the placebo is a statistical dinner between the growth rate constant but not really to match between the decay how the tumors regress and continues to indicate a placebo or the correlation between the

growth rate and the placebo and high dose -- it was really a good -- really significant .6 for both. this is the first experience using this equation that shows really the growth rate constant is an excellent yow by that time we know it's very common end

pointed and can help others to discern effective therapy. what it was that the advantage of doing this and other traditional end points by this time was that you can assess the growth rate constant before -- under treatment. and then i can show you another

example in this case, growth cancer data this was a clinical trial -- cancer patients using different combination of drugs. and again, in this case, instead of tumor measurements were esa values in the blood which is a surrogate for tumor burden in the case of prostate cancer.

and you can see here the same, patients at the beginning of the treatment -- a patient with no really response and this psa and also patients but you cannot see their measurements in the equation. it's again in this five trials during 15 years was i think 9%

of the patients that cannot be seen in the equation. this is the trial between the growth rate and the survivor. in this case it's .72 significance. the regression rate again the e of the part of the equation was not related with overdose

survival. also there's another the initial psa correlates with the other survival -- seeing a patient with a huge tumor burden is less than a patient with a small amount of tumor also correlated with the other survive and the time to get this response.

but if you see the correlation much higher correlation is .72 if you compare with the other surrogates. this is the final analysis of all these trials over time at nih using first for prostate cancer -- combination of chemotherapy -- hormonal

treatment and this is more aggressive treatment but it combines -- prednisone and then -- vaccine. you see this very interesting -- the growth rate constant was more or higher and more slow growth where you kind more but the last one -- was not

really didn't follow this pattern and it's the best of these treatments -- this is a hypothesis theory about this because this is -- not really agent or cytotoxic agent. this is the way in how vaccines work is to figure an immune response and may be delayed --

treating the tumor inside and for that reason you have maybe a lower -- in this case. continuing with kidney cancer just to see how you want to use -- there's still 14, 15,000 dying from cancer every year. if you compare this number to the year 2002 it's almost

similar. we have seven or eight more drugs approved by the fda to treat cancer. really the number of patients are dying is the same. this is again where the drugs we use for copy cancer -- receptor one and two in the tumor cells

and endothelial cells. but mainly all the drugs target the same pathway. then we start to analyze. this was the registration -- this was the hormonal therapy this was approved for chemo -- this is a big trial with 700 this is again to he show that

this is more molecule -- atp for phosphorylation of atp domain in the tpi receptor and the way how it works. the difference to the -- is that really is the therapist realistic diagram because we know that blocks more than the receptor -- and there's some

subtle difference in this p -- update of this trial. this is the important numbers to remember for the next slide. this is the response rate for patients 47% compared to 12% -- again five months -- and here you can see a couple, the difference between -- almost six

months very similar to different to the other survival. it was clearly a better drug. and then we, at that time for gs load all the data from this trial, all the sheets from each patients measurements of the tumors. and they calculated growth rate

constant and they -- what's court lated with the overall here again in larger scale you can see more slow the tumor, more the patients sleep. and here they are the same but not so strong. the correlation here -- we started to see a couple years

ago when i arrived what happened with the second lie in kidney cancer how we can really know if we are using drugs that similar the same pathway we're really including the treatment of these patients. this is a patient with prostate they start treatment with a psa

of 10. then the treatment works. you have that point of view the psa drugs to two. the patient was keep on treatment. the drugs to point to that is almost -- the lowest value. then it starts to grow again.

everybody was happy with chemotherapy, he started with 10. have now .4. then .8. this is very low compared with initial psa. 3.2, 6.4 -- and this time the patient was stopped chemotherapy

was doing another treatment. this is an example for what? this is the plus of the -- and you can see again the plus it showed this initial decrease in the psa values and increasing the psa values over time. you can see this straight line is the difference really between

this point and this point related to how the tumor grows is seeing that. double in this case the -- slightly the same is double the size. there's no assimilation in the process. the thing is that you have more

tumor, more mass but there's really no acceleration in any point. just to keep in mine this concept we tried to look at what happened -- for a clinical point of view. this is the -- guidelines that we use mainly in the private

practice or in the -- outside the setting for treatment of kidney cancer. suppose you see here first line and second line and you have like seven and like 12 or 14 options. because we don't know really what is best for our patients.

and the option you have a clinical trial. we're talking about ten years of targeted agents and we don't have still one that is really clearer than the other. and this is the trial, the axis trial that was to really see what happened with -- just to

show you the patient that -- the difference between these two agents was this 1.-- progression-free survival. definitely not a huge difference. this will be maybe a good one for approval but not so much in this case, in this analysis.

using this rationale we're thinking what will happen if the patient theoretically are -- meaning first line after progression. how we will compare with this 4.8 map. we analyze the data again, the data from the trial i showed

this is an example of 24 patients randomly choose from these 300 patients. i want to show you two thing. here you have many different patients with decrease in tumor size, partial responses, patients with no partial responses just growth of the

patients like with the and below in blue dots the growth rate concept. you can see that in each case more consistent, more clear growth rate constant, really constant over time. there's really no [indiscernible] you see from

here almost from the -- you can see the growth rate calculated is similar. it's not, doesn't change. the growth of the tumor is being stable over time. and this is another way to look at this graph. this is all the measures.

and we plot the g against the and just seven or eight patients we can see in the gene, in the from 300. this is like .7% of the patient has some different pa --behavior in the tumor. but really in the majority of cases keeps constant.

and really we want to look what will happen to the patient until the second progression designed by clinicians. for doing that we use the simple calculation -- because it's the 20% increase for assist and we divide this by -- of the population considering all the

patients including the patients that have acceleration of the this was 222 days, 7.3 miles. these mathematical calculation is not a trial but it is generated. you compare the -- after submitting in other trials, these trials prospective trials.

in theory, this compares very good in theory. for that reason we want to push what will happen ever but we do a trial comparing these studies -- versus the other drug. and the other thing that i mentioned before and that growth

rate constant in this tumor and this data set is said to us that -- because it's from the beginning at the same rate. and again for you remember the curve of the decay traction, the sensitive tumor, the resistent tumor and we see in the cleaning of the measurements.

it's interesting you can calculate that the growth rate without additional -- that means if you really think the concept for the growth, you can have a better feeling or -- because if you double the greet rate between the blue again and black lines you can see here for the

response is different -- not because you are killing more sensitive cells it's because the population of the resistent cells is smaller or larger. for some reason the -- cell in the tumor that you feel is small. that is not because because

really you have the velocity of the resistent population is growing. we collect data and many times -- a little bit that we have to look deeper in the data to get more information. we see the tumor as a dynamic and -- of a drug sensitive

portion that's digressing and a drug resistent fraction that is going you can measure the rate of growth of the resistents fraction and this correlates with other survival in at least two or three i present to you. i give patient the rate of growth of the tumor is not more

important than the absolute quantity of tumor. i put this in parenthesis. if you have a patient with a huge amount of tumor in the body independent of the how the tumor is growing, the patient may be will die because of the amount of the tumor in his -- but in

general this is very important constunted and maybe very important how the tumor is going. the rate of tumor growth is the constant. it appears to be constant from the beginning, we don't know -- in the core of the tumor.

sometimes the best thing that we can do for our patients is keeping the treatment. that's a strong statement. at least we have to assess or -- trying to assess [indiscernible] keep patients if the patient is well enough and it's clinically normal because the second option

we know and keep patients for example with ppis -- for more than the progressions is something that is done that is not really pa sellsed in clinical trials we don't have a good data for explain that. a response rate that does not mean more tumor -- showing the

last place but -- tumor has no time to maximally regress and more evidence in the, in our evaluation. deep therapies meaning that the earlier rate of tumor rose and -- and as a come clusion, you seen data collected part of the clinical trial and --

constant is very important now we have to do it. we just need the sheets of the patient the rate of growth correlates well with the rate of survival estimating these rates can give us ways to assess tumor roles and a better understanding of hour therapies.

and we can analyze a huge amount of data. it's like to have an experiment of 300 or 700 mice. and you can measure and get a lot of information and i think just the -- fcrpr but it's the is really a part of the data you can get from all of this big

trials. estimating these can give us insight how we might test therapies by continuing in this case than for long -- benefit. this is really word of dr. fojo, i'm just replacing him and just correlating some of his work -- and works very close with

dr. fojo. julia is the statistician that they do all the plots and the calculations and all the clinical themes and all the lab raters of urology and medical branch forral the data i just showed to you. thank you.

[applause] >> so this is an profession that you are using and you would recommend be used. in practice. >> yes. >> no, not yet. it's just mathematical trying to understand from a clinical data

with a clinical data how to assess eventually drug efficacy. but our goal, my goal was to try to push for a trial not big trial but small trial to effectively -- this hypotheses and really see what happened when you compare -- maybe it's not so good when we do a trial

but one thing that will be interesting to see that but at least -- like my country and south america, you don't have the option for second line or third line. and maybe the best way to go for your patient is try to keep the drug working and the patient

would qualify for many -- keeping the drug really slow in the tumor growth. i will not recommend for doing >> you would think this approach is applicable to any solid tumor theoretically. >> theoretically dr. fojo -- i think we have to see the data.

we have data for breast cancer -- multiple myeloma, a plasma cell dysplasia. now i will analyze data in lung i very interested in lung cancer because that's a good model for -- a lot of data in many i don't know if it will be the

same -- as i mentioned before. i think oh is a different animal related to treatment. i don't see that you can really calculate the role and it's a different derivation over -- is different [indiscernible] >> when i see your data i think about maybe -- about diseases

like melanoma for example where a binary melanoma is removed and they then go in five years exsix ars later and look around for remnants other than you don't see any and then four or five years after that we have a brain metastasis. that kind of, that's something

different from latency of tumor growth alone. >> we had to look through development of other mathematical more complex in the equation for -- therapies. how it doesn't work very well. >> so just a comment in terms of mathematical modeling.

i have a mathematician in my lab as well who is very impressed by these kinds of studies. and points out that there's a second level of analysis. for one thing in the heterogeneous tumor you can't assume every cell has the same level of resistance.

if you assume level of resistance you get a slightly different model. the other thing is this exponential growth which is interesting but can't be always the case that the tumors are growing exponentially. they slow down after a while so

you can introduce tumors can grow exponentially for a while and as they get bigger they stop it may be what you're seeing is exponential growth is not the exponential growth of a single ma tusions but the exponential growth of the total tumor which is slightly different situation

in terms of how would you treat multiple metastasis. there is this wonderful weigh which is another tear keen kinase growth promoting pathways. there are a number of different inhibitors for b ras and there are example of patients covered

with melanomas, they get b ras inhibitors and they totally going into remission, the tumor goes away. and then three months later they come back and it's -- consistent with what you're saying. there are very few other mechanisms that explain that.

>> i agree with that. my point of view to the model, i think you can do another more accurate if you put like the great rating of persistence and other things. the only problem i see with that is that the complexity will be left -- protein factors.

because this formula is really is like a physics from other types of put a very simple way. maybe it's not so simple like this but it's easy to do because just measurements. and you take all the thing like -- greater persistence we're still far away from that

level of measurements. >> in some sense i think this question is for you. you started off your talk saying that some people believe that there's intrinsic resistent and you don't take that point of view. this last talk, just your last

comment of the melanomas talk interest the intrinsic resistance so i'm wondering why you've come to the conclusion that it isn't intrinsic. >> i'm not sure i said atment i think i said that some people believe that all resistance is intrinsic.

mauricio and tito i think, i don't want to put words into their mouth. i'm sure you don't think all -- but a lot of the clinical behavior of tumors is as if the resistent is, all resist's we see is due to the heterogeneity of the tumor.

part of that is we're treating tumors that are far advanced. there are examples in bacterial systems. we're always interested in what's happening in amount body resistance in bacteria was things are becoming resistent. if you look for example at

resistance to third generation cephalosporin these treat antibiotic resistent criteria. there's an interesting phenomena which is very early on that you see, induction of sort of general mechanisms of resist else, pump systems and metabolizing systems and so on

that are not very specific for the drugs but allow some cells to survive long nerve so you tart to see fixed mutations that confers resistance. it could be that that's happening tumors as well that there's an initial phase of survival there are systems more

general like transporters and so on followed by a more substantial resistance mechanism that actually allows the cells to grow or not. if that's the case and nobody has actually done a very careful evolutionary study in molecular terms of what's going on in the

we can presumably do this in the -- because you can get samples quite frequently. one of the thing we're trying to do is actually to look using the tools that we have. that would change your awe approach because up front what you want to do is inhibit all

the sort of general mechanisms like the transporters and reduce the resistent population a little further. i think it's going to be different for each tumor and it may be from patient to patient, different mechanisms would be brought to bear.

the other thing about the approach that mauricio is describing it's a very practical approach given the set of drugs we are have now. it doesn't necessarily tell us how to develop drugs that are more effective or combination of drugs so that even those

resistent cells can be treated. >> neither of you mentioned except you did in passing, michael, animal model trables planted tumors as a source for experimental study both of mechanism and of drug responsiveness even the heterogeneity of the

people, you might respond to >> so the animal model which is analysis called the tumor-derived xenograph -- growing ces in tissue culture worrying about the adaptation process that i described. and a lot of pharmaceutical companies are investing large

amounts of money to use those kinds of models. there are tumors growing in immunodeficient mice. what i did mention in order to make real advances we need an organ culture system ex vivo that we can use to study the growth of tumor outside of the

animal. the down side is that ma hip lated to do imaging and molecular studies. real down side is their components missing, the immune system is missing, nutrient flow is missing or a variety of things, we're not sure how

important those are, but there's no perfect system at the moment but we need better systems to study drug resistance because to get new drugs, we need to be able to manipulate the tumors in a way that we can't do in a >> what do you think of the use of the energy the primary

patient derived xenograph. >> the question is about the totally immune deficient mice that allow the growth of most human tumors. by the way, we also didn't say anything about cancer stem cells which for the last few years have been a very popular

explanation for the development of drug resistance. and there are a bunch of reasons and tito i think agrees that it's hard to explain the behavior of tumors based on stem if there's a stem cell when did differentiate in all the different energies and types of

cells you see in a tumor and you kill all the cells that aren't stem cells. when that stem cell grows back in the tumor, other shouldn't be he resistent, it should be -- in fact the tumor that grow back are sometimes resistent. so i think that's an issue.

i think, i think there's a lot to be said particularly for the orthotopic tranplants where you take a tumor, let's say a kidney tumor and transplant it into the kidney capsule. because that allows for the local environmental factors, the cell types and the tropic

factors have an effect on the tumors which we think are whenever you do transplants it's not going to be exactly the same as the tumor growing in its original environment. so it's not perfect but maybe it's better. that remains to be proven by the

way, it hasn't been demonstrated. >> thank you very very much.