SLIDES & TRANSCRIPTS
Monday, June 17

Strategies for Therapeutic Targeting

David R Parkinson, MD

I represent a research and development organization that is engaged in creating therapeutics against defined biological targets. It has been very clear that our organization and organizations parallel to ours spend enormous amounts of resources in developing exquisitely profiled agents, and then testing them in completely, often undefined patient populations, which I frankly consider to be a non-test in many situations.

So, what I would like to do is take sarcoma as an example and give you some thoughts about how really organizing our particular development organization, how we are trying to put the tools together to go about this in a more efficient and more productive way.

I would love some feedback and some push back, because this is really for discussion, because a lot of what is going to go on in the panel discussion but, more particularly, as far as I can tell from this afternoon, relates to what this meeting is all about, which is to position the field to be helping develop therapeutics for sarcomas.

So, here is the setting. I don't need to emphasize this too much: a very, very heterogeneous group of tumors. It is part of the reason that they have not been the focus of particular therapeutics development, and frankly it has been difficult to get readouts in clinical trials of new therapeutics when they consist of odd collections of relatively poorly defined patients, for all the reasons we already heard about this morning.

TOP

Now, I think this is the wrong audience to spend too much time on the needs and benefits of diagnostic reductionism.

If the lumpers aren't in full retreat at this point with this set of diseases, then I don't know what it will take, and I think it is like the evolutionary discussion. You just ignore them and move on. Because, it is quite clear that if we are going to make any progress in this disease, we need to be able to profile patients from a biological perspective, considering that we are developing agents that target specific biology.

If we can't bring those two together, then we are not going to move forward, and I will show you an example very briefly of when that is done, things move forward very quickly.

We need, therefore, to -- as we heard so nicely from Marc Ladanyi -- begin to understand the biological homogeneous subsets. The reality is these tumors are relatively rare. Frankly, if the field is going to move forward, it is going to require an even greater degree of cooperation and coordination than has occurred in the past. As the GIST example has shown, where there is a compelling need, the field can rise up and respond to that.

TOP

So, those are challenges, but these are fantastic opportunities also. Because it is quite clear that many of the new targeting agents that are under development for particular biological targets, which have been defined often in the context of other tumors, are nevertheless quite relevant to many of the soft tissue sarcomas.

The GIST example, which I will just spend a couple of very brief slides on -- not to take away from Allan von Oosterom's thunder this afternoon -- is a good example, and I think we need to pay attention to that example because it tells us the direction in which to go to begin to develop effective sarcoma therapeutics.

I would say that at least my vote for one of the outputs of this meeting is a concerted program, whatever it looks like, to begin to define these patients biologically in the context of whatever therapeutics development. Then, take advantage of drugs that are already under development, and we can talk about those in more detail this afternoon. That is not the purpose of my discussion this morning.

TOP

Just some principles from the GIST, not the details. Allan can handle that better than anybody else. First of all, what did we have there? You know, people talk in the sense of chronic myelogenous leukemia. How did you develop a drug so quickly?

The answer is very clear. We started 40 years ago. That is true. 1960, the clear cytogenetic association, 40 years of work by generations of biomedical scientists setting up the biological context.

Then it became a technological problem to develop an agent against that particular enzymatic activity. We did it. By the seventh patient on the phase I trial, we knew we had a drug, and the rest is history.

It turned out as a benefit that that exquisitely targeted drug also happened to have a couple other targets that it hit, and one of those was kit, and it turned out that kit happened to be relevant in this setting, although -- as you have heard from Chris Fletcher earlier -- just because you are immunohistochemically positive for kit doesn't mean that that is a target. More about that in a second.

The reality is, if we can begin to define tumor subsets and understand the biology behind those subsets, then the clinical development becomes quite straightforward.

Availability of targeted agents with confidence concerning relevant dosing schedule. We knew what the IC50s were for GIST for Gleevec against the various targets. We knew, therefore, that the doses and schedules which we developed for CML, which is quite easy, because CML is quite easy: you have a biological readout, you have a clinical readout, you have easily sampled tissue. We knew that the same doses and schedules should, in large part, pertain to inhibiting activated kit. Therefore, it was easy to move into the original patient with a dose and schedule.

Rapid confirmation of target relevance, we knew a lot of that preclinically. So, we had confidence when the patient opportunities presented themselves. The fact of the matter is that it came easy clinically.

Imaging, I think this afternoon we will see some images of FDG glucose PET scanning, Allan, with respect to readouts of response. Finally, pathology confirmation.

Because the drug is so efficient, in the context of a homogeneous patient population -- and we saw from Paul Meltzer just how homogeneous biologically that patient population is--it has become relatively straightforward to define the relationship between drug structure, drug pharmacokinetics, drug pharmacodynamics, and the specific molecular mutations and the biological downstream implications of that.

We probably know more about the mechanisms of resistance at this point against Gleevec in both CML and, increasingly now in kit, in extraordinary detail in some of these molecules than I would say in almost any other field of drug resistance in cancer. Because we understand the biology so well.

What does that mean? That has implications for further-generation therapeutics development and for clinical strategies to get around the development of resistance. We have talked about the homogeneity of the biology.

Given that experience -- what we learned from CML -- given we learned from the experience from Gleevec and now a series of other very, very uncommon clinical entities that respond because their biology is affected by the particular drug, we looked at ourselves as a development organization and tried to consider, what was missing?

What was missing was what I discussed earlier, a concerted effort to bring new technologies to profile patients to the kind of exquisite detail that we have used to develop agents. So, in doing a situational analysis, we have basically gone through.

We didn't understand the patients well or the biology of their targets. We have not defined the heterogeneity. We really had impossible problems in establishing dose and schedule in areas where we were developing agents against other targets.

For example, we have an agent that we have struggled with for several years in terms of developing dose and schedule against protein kinase C. These are largely solid tumor patients and, frankly, surrogate markers and biological markers have not been very useful, and at the end of several years, we weren't even sure we had a drug or not.

Suddenly we realized that this agent also affected another target, flt-3 which happens to be in leukemic cells, and now we are on a roll.

Now we have an easy readout. We can look at the mutational and amplification biology behind the target, look at target functional changes with the drug. We can sample using the peripheral blood, and now we can move forward with the drug development. I am suggesting to you that the progress of therapeutics requires this kind of iterative feedback and therefore, we need to get the tools in place to give us that.

If we can't make decisions accurately, we will not move into this area. Development organizations such as ours are very, very businesslike in the sense that you measure what you can get from putting resources into one area against the likelihood of success.

Frankly, if you have a murky area of incompletely defined biopathophysiological entities, where you are not sure where the targets are, you are not going to do trials in that area. So, this problem has to be solved. What I am going to show you in the next few minutes is show you what we are trying to do, and then we will need to work with you in an attempt to do this in the case of soft tissue sarcomas -- to begin to profile patients more accurately, so that we can do accurate therapeutics development.

TOP

What we have done is this. This is the entire continuum of research and development but, just like in the biology of sarcomas, although it may have been a continuum, may have functionally been a continuum, history has managed to create separations. You will have noted this in your own organizations.

So, when we first started out about five years ago, building an R&D organization in cancer research and development, the Venn diagrams actually didn't overlap.

We have spent five years. The Venn diagrams do overlap now, but we have begun to understand that there are quite significant differences between full or registration development and early or iterative or proof-of-concept phase I development.

What we have actually done is create an organization in which the early clinical development is done, but sits in between research and development and incorporates the kinds of tools that I would like to show you.

TOP

So, this is what therapeutics developed looks like. The average cycle is about 12 years in the industry. It is somewhere between $500 and $800 million for a single chemical entity, which is why it is kind of hard to talk about developing specific therapeutics and specific mutations. It is hard to do that in the context of the way a regulated environment forces us to do our business.
R&D discussion, and here are the various steps. Each of these is associated with decision-making. Each of these steps requires more and more resources.

What we have attempted to do is figure out how an integrated, biologically oriented group that sits in between research and development can bring value, help in decision-making, and decrease risk of development.

TOP

One of those is to comprehensively profile tumors at the DNA, RNA and protein level -- more about that in a second. I will talk about how we are going to do that, not in great specifics, but we believe it is necessary, absolutely where we can to the extent that we may treat fewer patients in clinical trials, but examine them with more resources and more detail.

Bringing in blindly, patients who are undefined with some of these agents, is frankly, wasting resources, and in most instances, probably not giving the patient a good chance at success either. So, we are going to do that.

We absolutely believe now it is necessary to link diagnostics development with therapeutics development. GIST was a good example of that. We can discuss that this afternoon.

Implementation of a bioinformatics infrastructure -- you have gotten some insight from Paul Meltzer about the complexity of this. He was only talking about RNA arrays. Imagine when you start to look at the DNA information -- he did mention that -- and even more so now, the new kinds of information that we believe proteomics is going to give to us, which is a huge new piece to this puzzle.

To do this, we don't have all the resources in hand. Although we have enormous resources within the company, we either don't believe we should develop all those resources or, frankly, the technologies in these areas are emerging so rapidly that we don't want to be locked into particular investments and particular kinds of resources.

For example, we have all sorts of SNIP analysis. We have all sorts of RNA microarrays in the far-flung empires of genomics and genetics within Novartis. However, when it comes to actual execution of clinical trials, we need to have vendors in place or, when I say vendors, we need to have agreements with partners. They may be biotech companies. They may be academic institutions, to accurately help us profile these patients.

I will tell you, though, we are going to spend a lot of time on validation of process. One thing we learned, in trying to apply the same kit paradigm to small cell lung cancer, is that we don't believe the literature for most of the histochemistry -- no disrespect to pathologists in the audience because I am sure their papers are absolutely accurate. It is quite clear that the literature has not paid attention to quality control in reagent generation or to processes or to tissue preparation and, frankly, I don't know how you can actually be in this field because it is extremely confusing.

We have just gone out and we are redoing a lot of the work looking at particular targets as expressed immunohistochemically across lots of tumors, because we don't believe the literature. It is just shocking, to tell you the truth.

TOP

The other thing we realized -- and there hasn't been discussion of this very much yet today -- but if we are going to do analyses of either DNA or RNA, we need to separate tumor from stroma. Both are legitimate targets for therapeutics development but, to understand the underlying biology, we need to separate the two out, and that is just a slide of laser capture microdissection. We actually have contracts in place with companies that will work with us and our newly hired pathologists. We now have pathologists within the company within clinical development, to help us make sure that we are actually studying the kinds of patients we want to study.

TOP

Tissue microarray: this could represent an entire clinical trial on a particular slide, and I won't go into the details. I only mention it because it wasn't shown earlier; but this allows us, with a single staining, to look at a particular marker when we have an antibody to it. It could be FISH. It could be all sorts of other different assays. The reality is, you can put either 200 or several hundred now, tissue pieces from different patients in the same clinical trial, or with the same kind of tumor on a slide.

The other thing is the relevance of reagents. It has become very clear to us that being kit positive does not mean responding to Gleevec, as you heard from Chris Fletcher. On the other hand, with many targets, we want to know if we are using the immunohistochemical reagents to look at the resting state of a particular target, or the activated state.

TOP

So, this is just an example to show that it is quite possible now, fairly routinely, to generate antibodies against activated versus resting states.
We heard how bad the reagents have been for platelet-derived growth factor receptor.

We agree with that. We think that situation is changing. Frankly, the development of these reagents has to be a big priority for us and other development organizations like us. Otherwise, we don't know what we are doing with these reagents.

TOP

So, what are we trying to do with biomarkers? Well, we are trying to answer some of the kinds of questions I just talked about. To do that, we need legitimate biomarkers. They may be preexisting. They may be necessary to be developed.

We are actually beginning to do that in partnership with other companies who do this better and more routinely than we do, for all these reasons.

TOP

Why do we need diagnostics and biomarkers? There are many, many, many steps of what we need to do.

But again, it all comes down to, when we understand what we are doing we actually can do a much better job at it.

TOP

So, as we begin to reduce this kind of profile -- I would just say, we heard earlier about the problems with financing and translational medicine, and I agree with all of those. I have struggled with it for years, either as an investigator, or in the NCI, or now as a sponsoring organization for clinical development.

I think what we should try to do is separate the kinds of information that we need to get to accurately apply therapeutics from the challenges in accruing that information or paying for that information.

We are amazed at the rapid development of technology. Just in the last two weeks I have seen a presentation where it is now possible, with a real machine -- a prototype but a real machine -- to resolve proteins and then to actually sequence those proteins at the zeptomole level. For those of you who don't remember, because you probably don't move in those circles, zeptomoles are 10-21. It is just amazing. I just saw on Friday a demonstration from a company which hasn't gone public and doesn't even talk about what they are doing.

You know, in the genome sequencing, the machines that you were shown by Paul Meltzer, they sequence around 30 to 40 nucleotides a second. This company already can sequence 26,000 nucleotides per second, by the end of this year, 260,000 nucleotides per second, within a year or two, a million nucleotides per second per machine. They are talking about sequencing, ultimately, whole genomes from individuals as a commercial product. That is a ways away. However, what is not a ways away is the possibility of sequencing around any one of several hundred thousand SNIPS that you could choose.

So, this is a whole-genome polymorphism analysis, which you could translate also to looking for relevant mutations at any site you think is relevant in the whole genome.

Again, this is what Paul was talking about earlier -- moving from hypothesis testing, which was a legitimate way to attempt to do correlations, as you showed Murray, when all you had were reagents to p53 or to Rb. Now we understand that almost all sarcoma cells have p53 abnormalities or Rb. It is just that there are different parts of the pathways.

So, we are unlikely to find correlations there, as we heard about, in some classes of tumors; but if we can begin to look at every single step of the pathway and correlate it maybe with RNA over-expression as we heard about, then we begin to do pathway analysis as we heard about from Paul Meltzer -- physiological analysis. That is what we care about. Our drugs perturb pathology. They don't perturb anatomy, at least initially.

So, all of these kinds of approaches have relevance to target identification and credentialing, as we heard about earlier; but again, we are talking about -- our ultimate goal is to disturb function rather than to identify structural abnormalities. We are probably going against pathways, rather than individual targets. We are going against physiological rather than anatomic approaches. What does this mean? It means that it is highly unlikely, in these complex tumors -- now GIST is a wonderful gift; we can learn from it; we can benefit patients -- but I don't think there are a lot of those silver bullets out there. There are more of them with different targets but it is not going to help us defeat the complex karyotypic kinds of environments that we see in solid tumors, and it turns out in a lot of the common soft tissue sarcomas.

So, we are really talking about combination therapy to deal with pathways, pathways that are often redundant or parallel or have feedback loops.
If we are actually going to successfully harbor our resources and use them in a constructive way and minimize the time that is necessary to get from what we put into it to what we can get out of it in terms of patient benefit, we need to think that way and begin to develop our organizations and get together the resources to act that way, if we are going to be successful.

TOP

So, just a comment. This was a Petricoin paper in Lancet. There is a lot of controversy in the proteomics field as to whether these results are actually correct or not, but no one will disagree that the process is valid. The point I make, I will summarize the whole thing by saying that it was possible on 20 microliters of unfractionated serum, to examine using an approach called mass spectrophotometry on a particular adhesive-type surface slides, to begin to profile proteins in such a way that you could look at differences in profiles between women who had ovarian cancer and women who did not have ovarian cancer, select five proteins that were in the one and not the other, and begin to develop diagnostics and insights into relevant targets.

Regardless of the correctness of this paper, the approach is very valid. This is actually first-generation technology. We have met with a company recently that can change the adhesive surface up to 10,000 times, to begin to give 104 additional levels of complexity to exactly the same approach. So, pay attention to proteomics.

Although DNA is a wonderful thing -- it came from our mother and a father, so it can't be that bad -- and RNA is a wonderful intermediate what we really develop drugs against are proteins, and this is a field that has finally started to move very, very quickly.

TOP

I don't need to tell people here that -- while these wonderful terms that I struggled with as a medical student to learn about the soft tissue sarcomas, are wonderful -- we need to begin to profile tumors biologically, physiologically because frankly, that is the way we are going to be treating them. We need to definitively characterize patients and establish therapeutic relevance.

Then frankly, we and others have agents against many, many, many of the kinases that already exist.

The issue is which one of those to take forward into development or, frankly, which molecules which have profiles of combinations of different kinase activity to take forward.

TOP

We need your help in defining how to do that and which ones to do it with. So, I speak as representing the entire therapeutics development industry in the sense that we have incredible technologies which allow us to develop molecules against targets, but we need your help in the context of soft tissue sarcomas to help us understand which targets are relevant and in which settings, to make clinical development more efficient. If people smell that that situation exists, then significant resources and these molecules will move into that setting. Thanks very much.

TOP