SLIDES & TRANSCRIPTS
Monday, June 17

Other Potential Targets in Soft Tissue Sarcomas

George D. Demetri, MD

We all talk about translational research. I kept this one in even though it is simplistic so that you will know where I am coming from. The whole idea of to translate is to change something from one form, function or state to another, and I would like this to convert to "translate ideas into reality".
I think as a field, in sarcoma, that is our challenge right now. We need to figure out how to do that while being visionary, but not delusional. I think that is an important thing for us all to keep in mind that has somewhat to do with the technology that we have at our hand, to keep us honest and to keep us functional here. We need to understand the research and figure out how best to apply it, knowing that it is moving forward so quickly, to understand and conquer these many different diseases.

TOP

Now, for the public at large, sarcoma hasn't been on the radar screen. It is not a common cancer. I would actually like, on behalf of everybody in the room, to thank CTEP and the NCI for having us be the first of the non-common cancers to actually have a State of the Science meeting. I think it certainly plays on the excitement around Gleevec and it really helps us, as people interested in mesenchymal cell biology and clinical care, to really focus the NCI in on this very important fact, that we have some lessons we can learn, and that is going to be the theme of my whole talk here.

When we look at these mesenchymal cell issues, we are being given clues to human biology. We think of tumor targets that we should be going after, not just in sarcoma but in all sorts of cancers. Sarcomas just happen to give us more insights, often, than many other diseases. So, the ideal tumor target is often a single, validated element, and that is the key word, "validated". We don't have that for a lot of solid tumors. It is critical to the pathogenesis of the cancer in humans -- that is expressed and active in the tumor -- and, very importantly, is unable to be replaced easily by the tumor using alternative pathways. If it could, it would be trivial, but it has to be a fairly important consequence to the tumors. Ideally, it should be necessary and sufficient to the cancer cell survival. So, these diseases driven by a single genetic mutation are often highly informative, proof-of-concept diseases.
I, like everyone else, like Marc Ladanyi's way of summarizing this -- that we are being given sarcomas, some of which have single genetic mutations, others of which are kind of a mixed bag. In some ways, this is a microcosm of cancer in general.

TOP

We know that we need more help from our diagnosticians. Sarcomas are all over the map in terms of what they look like and even what they behave like.

They are increasingly being well defined by better strategies on behalf of our anatomic pathologists, and better tools on behalf of scientists and molecular pathologists to help us use these as useful model diseases, and not just a mish mosh, which is what Dave Parkinson told us sort of our field used to look like 10 years ago and I think, to a great extent, that drove away industry. We can help bring back resources into our field, the more we whittle this down and the more consistency we get into the field.

TOP

What are the next challenges? I put this up as I was thinking about all the speakers from today, and where the meeting was going. I think we see overall that surgery and radiotherapy, we are seeing outstanding local and regional control rates. We get good functional outcomes overall but patients are still dying of systemic disease. Murray kept coming back to that point, that the ones who had the bad outcomes are really -- these are the new challenges.

That, to me, falls into the realm of finding advances for systemic or metastatic disease, and that is the primary lesson of GIST and imatinib -- that systemic therapy really can be effective against sarcomas if it is properly chosen. There, where you had a disease where everyone agreed nothing worked, and it was pretty much a chip shot, if something was effective, one could show it was effective because nothing else ever worked before.

TOP

So, the lessons have been enumerated already by other speakers. Kit is a critical target in GIST. We know that now from the human experience. Kit is expressed in virtually all the GISTs and it is ubiquitously activated. Brian Ruben and Jonathan Fletcher had already shown that in the laboratory, and the clinical experience has borne this out.

Here is my thought and my proposition. The analysis of resistance to imatinib in GIST will help us to identify new target pathways. That is an imperative right now, that we take advantage of this next-generation challenge.

TOP

In many ways, the simple questions are the most powerful.

TOP

Why are these cancer cells cancer cells? Why are they responding or why are they resistant to imatinib?

TOP

There are some clues here. These rare cancers are giving us those clues that we then can take forward.

TOP

We know this issue. Marc brought this up, that single-gene cancers are the simpler ones and, in some ways, are informative. Jonathan Fletcher had identified this group of leiomyosarcomas that later turned out to be GIST as cytogenetically rather simple diseases.

This one happened to have lost a single copy of chromosome 14

TOP

and contrast that with a leiomyosarcoma. This could equally well be a far-advanced lung cancer, really aneuploid breast cancer, whatever.

TOP

The point is that these have very unstable genomes. You don't know where to look in this genome. Where is the key element of this one?

TOP

So, we are looking at this as two different types of equations. We have, in the form of GIST had the advantage of simple arithmetic to go after. As we develop, as we saw in the equations on the left, we develop the tools to solve for the equations on the right.

TOP

So, everyone has their own complex wiring diagram of the cell and, rather than bring that in, I wanted to just point this out.

As I think about what the next generation targets are, I think from the outside of the cell toward the inside of the cell, actually there is a nuclear membrane in my version that didn't show up on this slide. In the extracellular domain we have many different targets -- the antigens, NY-ESO-1, heat shock proteins, whatever. Extracellular domains of kinases and other flags sitting on the surface, translocation, chimeric antigens, if they are presented to the outside world. These are all targets of immunotherapy or other things, potentially. Intracellularly, you have got all the complex signaling and kinase activities.

Then, in the nucleus -- the aberrant transcription factors, even some of the normal transcription factors, like the PPAR-gamma pathway. Maybe all of these could be used. I decided not to focus on any one of these right now because I think that is the challenge for the field and that can be what we all discuss this afternoon.

TOP

Certainly, our first foray into this has been pretty darned exciting, the whole idea that just looking at one class of target -- a mutant tyrosine kinase -- has been able to yield something so important as imatinib for the therapy of GIST.

TOP

So, how do we take this morass of data and choose the next generation target? What is the next rabbit to be pulled out of the hat? I think that is an important thing. It can't just be a rabbit. It can't just be serendipity. We have to look at the strengths of the pre-clinical rationale. We had very strong pre-clinical rationale in GIST.
Jonathan Fletcher and many others had shown that GIST was a cytogenetically simple disease. The group in Japan had shown that these GISTs have an activated gain of function mutation, and that was consistent with the biology. So, we sort of had a good rationale there. Jonathan's cell line was pivotal to helping Dave Tuveson and others use the observations of Mike Heinrich; and Brian Druker had shown that imatinib could inhibit kit and then move that forward into the GIST model.

So, we need to be able to identify the target in vivo, and then reliably link the target and the pathophysiology of the sarcoma.

TOP

What that gets us is this side of the equation -- the whole idea of rationally designed, target-based, translational research. You know the target. You really have a good sense or good pre-clinical data that it's relevant to the sarcoma and you move that forward. Sounds great.

TOP

How many times do we really have that? For how many other sarcomas do we have that level of certainty? I would argue not that many. The transcription factors are fascinating; but the transcription factors, the chimeric transcription factors that Marc talked about are challenging targets. They are harder to deal with than this mutant kinase was.

TOP

What we are seeing is a push back on the other side, the idea that empiric exploration, so-called "discovery in the clinic", may be another way of going about this. Maybe we don't know all the cell wiring diagrams and perhaps, using different tools, we can take advantage of that and take advantage of leads that we are hearing about. That mechanistically investigating responses to identify new targets, either in resistant GISTs or in other sarcomas, may be a way to go about this. Certainly, the rationally-based target design ideal has been imatinib and GIST and CML. It has been a very nice story.

TOP

As I think about this, I think about how we should design translational studies. Let me just walk you through this next exercise, because I think it might be useful as we think about how to keep moving our field forward.

What I have got on the left here is a target incidence. Let's take a molecule that's not necessarily 100 percent of tumors, but maybe it is. In the middle, let's say that tumor is absolutely critical to that tumor. So, here we have got target incidence. We can vary that from uncommon at the bottom to very common at the top.

The target activity means the tumor really relies on this 100 percent. If you knock it out, you kill the tumor -- 100 percent target activity. What you will see is, depending on what the population you treat is, you will dramatically change the observed outcomes in a clinical trial. For example, if you had 100 percent target incidence and that target is 100 percent active, the observed outcomes are simple. 100 percent of the patients should respond to your intervention.

What about if you have only got half the cells that have that target -- 50 percent target incidence -- and yet the target is still 100 percent active. Your observed outcome will only be 50 percent. You walk down the sensitivity analysis to an infrequent target, five percent, but a very important target; it is still 100 percent; and yet the observed outcome would only be a five percent incidence rate. One can make whatever analogy you wish. This could be estrogen receptor in breast cancer. This could be kit in GIST.

It really shows that, unless you have a good idea of what you are looking at in these tumors and the relevance of the target activity, you are not going to necessarily be able to interpret the outcomes.

TOP

Because for example, let's take more of what might be the case in GIST. Virtually nothing is so important to a tumor than that you have 100 percent activity in every tumor. So, let's take GIST. 100 percent of the GISTs probably have kit and it is activated. Maybe only 75 percent of them are really needing the GIST and really leaning on that pathway.

So, the overall observed outcomes may be something like a 75 percent effect. That is about what we are seeing, around 60 percent, 65 percent partial response, with another 15-20 percent stable disease. You can quibble about it, but that is probably about what we are seeing in GIST. Now, if the target incidence drops to 25 percent, the observed outcomes will only be a 19 percent response rate. So, had we selected the patients poorly, we might have actually missed a lot of this important activity.

This has more relevance to the more subtle things -- contributing factors, targets that are contributing to the cancer phenotype of a sarcoma -- but may not be as important as kit is to GIST.

TOP

So, let's now change it to a 25 percent target activity. Now, in your best possible scenario--100 percent of the cells have it. If it is only a significant contributor, you still are just on the verge of being able to detect anything in a clinical trial, and, if the target incidence is less and you don't know it, you haven't looked for it, you are going to have a negative clinical trial. This gets back to the idea of relying on our diagnostics to really have any hope that in the clinic and in our translational programs we will be able to detect significant activity.

TOP

The conclusions are we need to pick targets which have significant impacts on the tumors, they are very relevant. That is always a little bit of a wishful thinking game because we don't know how many of our models in the laboratory really translate into sporadic human tumors. These contributing targets can be detected only if the tumors are ideally chosen. So, this diagnostic screening is critical.

That makes me worry a lot about how good our target detection technology really is, something we can talk about later in the afternoon, the sensitivity, the specificity, and the reliability across different sites. We worry about that.

What about, do we need frozen tumor? In that case then we are going to be just that much slower along. How much does sampling variability also affect results? Like Chris Fletcher said earlier, if you poke a needle into two different sides of a sarcoma, you might come up with very different results, depending on how much normal stroma is in whatever sample you are, what the differentiation status of the tumor is.

TOP

So, the more empiric way is to sort of give this drug to a lot of people, which is not something I am encouraging us do, but we are seeing being done now. We need to be careful about that, and yet reproducibly try to analyze and make sense of this in a thoughtful way. Possibly, by carefully studying these empiric responders, we might get some more information on these pathways relevant to a subset of cases.

This might be an efficient way of really carefully analyzing responders, not just with this drug, but with other drugs that are in various pipelines, to try to get an idea of which pathways are reliable and which pathways are important to different sarcomas.

TOP

So, what other lessons are there? The most important lesson -- Allan kept stressing and I will stress again -- is that all GISTS are not the same. Here we have a relatively small, presumably homogeneous population of cells like GIST. When Jonathan, Mike Heinrich, myself, Chuck Blanke, Meg VonMehren and Burt Eisenberg were talking about setting up this study a long time ago, we knew that the molecular prognostication could either be really, really interesting or really, really mundane.

We didn't know what to expect. What we found was more interesting than any of us could have predicted.

TOP

Mike Heinrich and Jonathan Fletcher have shown this nicely -- that the axon 11 is the most common site of the mutations to be harbored. Axon 9 is another group, and then there is a smattering of other axons which can harbor mutations.

TOP

This is the more updated version of this. Again, Mike's and Jonathan's data, showing that the partial response rates are dramatically different, depending upon the molecular genotype of these. We do not understand mechanistically 100 percent why this is; but it is certainly more interesting than anyone would have guessed.

TOP

Mike Heinrich presented this at ASCO as well, that it is not just related to response rates. It also has to do with time to progression, or the duration for which this drug is able to keep GIST under control.

TOP

The other nice piece of data that makes this all dovetail together is the fact that this dovetails beautifully with the imaging, using FTG PET scanning as a surrogate marker. That is another point of our research that has to be moved forward, that the surrogate markers give us rapid readouts of what is going on.

This is an SUV analysis of the PET scans of patients at baseline, before they get imatinib. This is from Annick D. Van den Abbeele at our institution. This shows that if you have got the null-mutation wild type -- the dark blue on the left where the purple is axon 9 and the yellow is axon 11 -- at baseline everybody has got about the same uptake. You can't tell the difference based on the molecular genotype.

TOP

After one month on drug, they segregate quite nicely in the same way that they segregate later by partial responses. The wild types don't lose their uptake. They stay a little bit hot on the PET scans, where the axon 11s go really nicely cold within a month, and the axon 9s are in the intermediate group, beautiful dovetailing of imaging as well as the molecular biology.

TOP

So, the lessons are beyond all GISTs are not the same but are far more interesting than we would have guessed, and the mechanisms may give us tools to really figure out what is going on at the molecular level. The corollary is that sarcomas are not the same, and there are likely to be many more relevant microsegmentations of sarcoma yet to be determined.

TOP

What I would propose is that the resistant GIST analysis will now give us the tools to find other pathways that will have relevance to other sarcomas and possibly also other solid tumors. This is a very feasible goal with the appropriate analysis of the tissues.

TOP

People like Jonathan and Mike are doing this as part of our study right now, and I will just show you one sample patient.

On the left hand side here, there is a cell line that Jonathan developed, these GIST cells that have nicely activated phospho-kit. On this, we have another downstream target that Jonathan has nicely assayed for -- the phospho-MAP kinase. In the middle here, we see an axon 11 GIST mutation patient who doesn't show up all that well from a distance, but whose pathway before imatinib is turned on, the MAP kinase is downstream and turned on as well.

What you see is that, as this patient became resistant with continued dosing of imatinib, suddenly you see the reactivation, the reappearance, to a greater degree, of that phospho-kit and the phospho-MAP kinase with it as well. So, how can that be? The target is still there but the drug is no longer shutting it down. That gives you a few different possibilities of how that can be developed, likely to do with how the drug is fitting into or not fitting into the binding pocket.

TOP

Again, these are reasonable things. They have been nicely worked out in CML to a different degree, and it is something that Jonathan Fletcher and Mike Heinrich are beautifully working out as part of this resistant GIST consortium work that is going on.

TOP

So, leaving all the resistant GIST and back to the main question, what is our next target? What is the next rabbit to come out of the hat? Some of it we can't even think about yet, but Paul Meltzer and others are giving us some of these tools with the discriminator lists.

TOP

As we see some of this work come from the NCI or from Memorial or other places that are doing nice gene array work, we will get some ideas of other targets.

TOP

The caveats we have to keep in mind are that expression alone does not equate with activation or relevance of a putative target.
We know that immunohistochemistry is still somewhat more of an art than an exact science. My analogy here is that if you give the same oils to Matisse and Monet, you get very different paintings at the end of the day. So, even if you validate the tools, the immunohistochemical art may still give you different answers, and that is something that I look forward to a colorful discussion about.
The functional assays which sound so easy are often technically challenging.

TOP

well, if you wanted to get that Jaguar ornament off the hood, if you could effectively target that hood ornament, you would have something that is highly specific, but not have the functional outcome right? You could hit every Jaguar but still not stop them.

TOP

What if you could target the rear view mirrors? You could take those out, but the car would still keep going. On the other hand, you would sort of cross-react with a lot of other things.

TOP

If you really took out whatever tires Jaguar uses -- I have no idea -- you pretty much stop it, at least for a while until the driver could get out, change the tires with something impermeable and keep the car going again. I think that is what we are looking for, something with that sort of a functional outcome.

TOP

So, what we have to do is decide, what are the hood ornaments and what are the wheels in this very complex circuitry that we are defining of these cells?

TOP

To go back to that, how are we going to choose this next target? Certainly, the careful planning and the teamwork is absolutely critical. Having the basic scientists, the molecular pathologists, the anatomic pathologists, medical oncologists, surgeons, everybody work as a team has been one of the great privileges of my life, really, working with this team of people from academia, industry, all over the world, on the GIST story, and in collaborating to ensure that we get the right patients, get the right samples and get the right analyses done, and then mixing this target oriented and post facto analysis of empiric leads might be the only way we can move this field forward.

TOP

I think, with that, we also have to keep in mind that there are a lot of barriers to translation and collaboration. This cross-functional teamwork is not an easy thing to do. It is easier said than done. There are still a lot of barriers that have to do with resource allocation, that have to do with academic credit that we are still working on as a field, and I think still have some work to do. This collaborative teamwork is the only way to do it. We can develop open communications and structures to support this and, again, I think that is part of why we are all here, to help the NCI help us to do just that; and to optimize our use of the Internet to get these patients.

Some of our friends from the patient advocacy group are here today. They have been really critical to helping these patients find this before their doctors knew it was coming and to help us educate not just the patient community, but the physician community out there as to new treatments that are available, so that we can get the right patients on the right studies. With that, I thank you and all my collaborators as well. Thank you.

[Applause.]

TOP