SLIDES & TRANSCRIPTS
Tuesday, June 18

Breakout Group A Summary and Recomendations with Group Discussion: Pathology, Molecular Profiling, Prognosis

Jonathan A. Fletcher, MD

Just to quickly revisit -- the issue which has plagued us is "exactly how this kit was identified as a central, activated, central oncogene in GIST?"

Fundamentally, this was by a confluence of information from developmental biology and pathology. The problem that I am going to address is that for most of the other types of sarcoma, we had a few fundamental observations that led the way in GIST that we just don't have in most other types of soft tissue sarcomas.

One was the observation that expression of the protein kit is absolutely required for the differentiation and proliferation of interstitial cells of Cajal in the gut or the ICC. Further, that kit is expressed strongly and diffusely in that normal cell type, further the observation that emerged from the surgical pathology literature in the 1990s that GISTs might indeed arise from this Interstitial Cell of Cajal lineage and further, that most GISTs in turn express kit diffusely and strongly, which is a protein that is expressed in rather few cell types.

So, it was that body of rather compelling incremental information that led Hirota et al. to publish their landmark paper showing that kit was mutated in most GISTs, which got the train rolling. The question that we are struggling with here, of course, is what infrastructure do we need to more empirically identify these same sort of clues in sarcomas where we don't have this serendipity of the confluence of developmental biology and pathology?

TOP

We are going to roughly classify the potential targets into three categories, a rather simplistic approach, but we would like an infrastructure set into place such that one could realistically triage or establish a hierarchy of targets that would range from being targets but not particularly compelling, down at the bottom, to those targets that we could really get worked up about.

Expression alone would identify a potential target. It is expressed in the cell type, but we don't know anything about the function or whether it plays an especially meaningful role in that cell type.

If it is expressed and it is activated, if we know that functionally it has turned on in that cell type, we get a little bit more excited and that is a promising target.

Finally, if it is expressed, it is activated, and we have the understanding that there is a specific genomic mechanism which has singled out that target, focused on it, then we really have a sense that it is playing an essential role in that cell type, and then it becomes much more of the model of kit and GIST or other oncoproteins that have been targeted successfully in various human cancers.

TOP

Again, the problem is that in many types of sarcomas we don't have this background of biological information that allows us to identify those targets which are the best for a particular tumor.

So, we need an infrastructure which enables the identification of the essential activated oncoproteins, particularly those which are required for maintenance, ongoing maintenance, of the transformed state of that tumor. And, by definition, lacking the above, we need an approach which is somewhat empiric, which probably involves high throughput data sets as well as an ability to identify the candidates that do the best, and to focus on those, to identify the most promising targets.

TOP

Further, as an assumption, we would posit that in many types of sarcomas -- and as we heard yesterday from a variety of speakers, including the GIST model of the compelling kit target -- that ultimately we need to target therapy to encompass both that primary target as well as ancillary targets that may be downstream of the target or may be parallel targets. Otherwise, in most patients, no matter how successful the monotherapy, we will be dealing with the eventual emergence of resistance.

So, we need an infrastructure to guide the appropriate selection of combination targeted therapies, an infrastructure based on gene expression, protein expression and, importantly, activation -- and ideally mechanisms of activation because, again, very often, the mechanisms of activation will tell you which of the targets are the most compelling.

TOP

We just show an example here of a potential trajectory which one could take a given type of sarcoma, retroperitoneal sarcoma, MPNST, and follow this through to identify what we would regard as very potentially compelling targets. Starting perhaps with high-throughput gene expression analyses, simply to demonstrate which genes are expressed at relatively high levels in a particular type of sarcoma -- that might be cDNA arrays.

Now, the techniques that I have listed here are just examples of some of these cases. Many techniques are now available, but just to make the point. That would be coupled with high-throughput protein expression analyses to confirm and validate that the cDNAs that we think are overexpressed are actually translated to a functional target or a protein target that would be relevant to drug intervention. This might be evaluation of sarcoma tissue microarrays by immunohistochemistry.

Additional analyses to determine whether the candidate protein targets that we have identified in Step 2 are expressed in the activated manner, say, phosphorylation. That gives us a sense that, functionally, they are at least potentially playing a role in that particular sarcoma type. This might be immunohistochemistry or immunoblotting with phospho-specific antibodies.

Ideally, genomic studies, which might be sequencing or molecular cytogenetics -- FISH for example -- to demonstrate what are those proteins that seem to be highly and perhaps aberrantly are, again, being singled out by a genomic mechanism -- amplification, rearrangement, point mutation -- that would give them a special significance. Very importantly, in vitro pre-clinical studies of target inhibitors -- going after those proteins that we feel to be compelling.

In addition to the simple, one-drug approach, trying Gleevec in some tumor where we think that PDGF receptor-beta for example is strongly activated, these would become ultimately fairly complex studies. Tying in with other aspects of developmental therapeutics where one would put a given inhibitor together with other inhibitors to determine these very important questions as to which pathways we need to interact with synergistically, or inhibit synergistically, to really accomplish the final endpoint of knocking out proliferation -- inducing apoptosis and so forth.

Several in the group mentioned the exciting developments with small inhibitors of RNA, which are, in many cases, much more selective in inhibiting particular targets than the small molecule drugs that we have been discussing at the meeting, and would play an important adjunct role in telling you that what you think you are inhibiting -- which pathway you think you are inhibiting -- is indeed the correct one in the particular cell type.

TOP

What do we need to do this? I have described a process that everybody in the room is excited about and many of the folks are involved in, and there is a real passion for. There are certain resources which are needed urgently to make this sort of process available to the general sarcoma community.

Coming back to frozen tumors -- if one wanted to develop these whole stories from start to finish, one would need about half a gram of snap frozen tumor -- 500 milligrams of tumor -- to do both RNA, protein analysis, genomic studies, and so forth.

Of course, in a given scenario, not all aspects of this pathway need to be done by any one lab, and this will come back to the theme of data sharing and the idea that we need a centralized portal to exchange this information more efficiently.

Generation of additional cDNA microarray datasets -- there have been terrifically exciting publications in the past year that we hope are very much just the start of the process. The urgent word needs to get out there, that we need a substantially increased amount of support to generate this sort of data so that we simply know which genes are selectively expressed in different sarcoma types.

Serial analysis of gene expression - SAGE -- is a complementary method of screening the transcripts, or identifying the transcriptosome for different sarcomas.

The advantage of this approach is not only in validating or confirming some of the sense from c-DNA arrays, but often identifying genes that we might not even know about yet, and might, in fact, be expressed only in sarcoma types and not perhaps in normal cells.

Generation of sarcoma microarray immunohistochemistry datasets -- this is an area which is just in its infancy, but we see a great need equal to the cDNA microarrays to have data sharing and to encourage development of this data, which is available in the first place.

As part of this -- an important part -- having a reference resource to provide sarcoma mesenchymal tissue microarrays that are highly validated in terms of being representative of their given cell types and optimal tissue specimens.

We need to generate datasets for phosphorylation or activation profiles of the cell signaling cascades in common sarcoma subtypes -- again, getting back to a need to identify the pathways that are at the fore at the moment, with the excitement over small molecule inhibitors.

Also, something that one can integrate with the cDNA array expression profiles, if one knows that phospho-akt is through the roof, for instance, in a particular tumor type. It might cause you to focus on certain genes that are related to that pathway in the cDNA arrays, to see if you develop a greater sense of enthusiasm for that particular target.

There is an urgent need for bioinformatics and data-sharing resources, and some of the group described the ongoing initiatives for this at the NIH. Clearly, that needs to be tied into this effort, so that there is the appropriate power in terms of the sorts of tumors that are incorporated, perhaps, in these tissue microarrays, and that there is the necessary support to enable the sharing of this data with the sarcoma community generally.

Finally -- very important -- coming back on the previous slide to the in vitro studies, we need a sarcoma cell line repository or at least an identified, highly characterized group of sarcomas, perhaps at the American Type Culture Collection, which we all believe in.

As we know, there are leiomyosarcoma cell lines out there now, and we are not sure how representative they are of the tumors they are supposedly derived of. For many types of sarcomas that we all like to work on, there are really no well-characterized examples that are readily available from the conventional distributors.

TOP

So, the last slide is really the meat of the recommendations -- these specific resources which we feel, if they were generally available, would go a large way toward enabling the development of this infrastructure for target discovery and validation in soft tissue sarcomas.

Of course, as part of this effort, there would be a direct tie-in with what I expect we will hear from the second breakout session; but it also comes back full circle to Sharon's presentation, where part of the issue in identifying how pathology specimens should be processed and how we should be gearing ourselves, presupposes that we need an infrastructure in place that will more reliably, reproducibly, handle these important targets that we all know are coming over the next five or 10 years and that will in many cases, be instrumental in therapeutic decisions.

TOP