[No text is associated with this slide]

TOP

Thanks. I am going to give kind of a global perspective on some of the problems of genomics in cancer, and try to tie that to melanoma. I asked to go first, because I think I will try to set the stage for the speakers that follow.

 

TOP

Of course, everybody is keenly aware of the potential of genomics and the incredible opportunity that we have through the availability now of the human genome sequence.

We have just finished an arduous celebration of the completion of the genome sequence and the 50th anniversary of this paper.

What is little known is what these guys were actually talking about at the time this photograph was taken and, being very visionary scientists, they were aware that we were going to identify tens of thousands of genes with many exons each, and that it would end up being a huge, huge haystack to look for needles within, and we were looking at trouble down the road dealing with all this complexity.

So, that is what I want to talk about a little bit today.

TOP

First, in a very, very broad sense, what are the tasks of cancer genomics. I think this can be put in very simple terms.

First, to define the genes which are targets of mutations in cancer, to understand the mechanisms of genomic instability that lead to these mutations and to the development of greater genomic complexities in cancer, and to understand completely their phenotypic consequences that lead to the evolution of tumors, as we recognize them in the clinic.

Finally, to take this knowledge and translate it in some practical sense into the clinical arena. So, that is what we have got to do and I am going to talk about sort of how do you do this, and how are we doing at it.

TOP

So, this is sort of the classic Vogelgram, where we have a very simple concept of an orderly progression of cancer, with each step involving a distinct gene. So, we go from a normal cell out to finally the invasive cancer, which is the thing that actually kills people and, in each stage, we have a gene. This is an oversimplification and we recognize that, of course.

TOP

First of all, we know that the genome becomes quite chaotic at the end of the process in almost all cancers.

TOP

But we don't know exactly when that occurs during the development of cancer, but it may be relatively early.

Of course, the bigger headache is that it isn't just a simple matter of having a single gene at each stage of cancer progression, but there are probably multiple genes, a menu that the cancer can choose from as it progresses, and that is a big part of the problem.

Of course, this linear model is also incorrect. We really should be looking at a more branching, ramified kind of progression, where heterogeneity develops within the tumor. I haven't depicted that well here, but that is also obviously the case.

So, we really have quite a challenge to identify all of these genes. We also have a feeling that the number of genes involves increases as we go through the progression so that, by the time you get to the most advanced, clinically apparent tumor, there are probably multiple genes that are probably abnormal. So, that adds to the hassle here.

TOP

Let me talk a minute about how well the field has been doing in terms of the sort of pace of discovery and the scope of clinical applications.

So, we can find point mutations reasonably well now, but the pace of this is still sort of low to medium, because the technology and the costs of high throughput sequencing are still high, and very few centers can actually do this sort of total genome mutation screening.

The clinical application of information on point mutations is still really quite limited. The pace of discovery of high penetrance hereditary mutations in cancer is very, very low right now, and perhaps there are a few left to discover, and its clinical applications are also quite limited, although important and familiar to everybody here.

Low penetrance hereditary gene mutation, it is really quite a painfully slow process, and I think anybody who has been involved in trying to find these genes will agree with the pain part, and the clinical applications of this are still very limited.

That is not to say it is not incredibly important for the future, but it is not something that we are really racing along a fast road of discovery.

Numeric changes, we are actually getting really quite good at finding these with improvements in CGH technology, but the clinical applications of these are really very limited. The only two I know of are ERB-B2 in breast cancer, and N-MYC in neuroblastoma. Otherwise, they are very, very limited.

Translocations are still being discovered and even these sort of classic signals of mutation in cancer have a fairly limited clinical application in fairly rare cancers.

Epigenetic changes, we are just really starting to dig into those in a good way and, as far as I know, that has no direct clinical application.

So, I think there is a huge challenge and a huge opportunity to take the burgeoning genomic knowledge that is developing and bring it into the clinic in meaningful ways, both for diagnostics and, of course, for therapeutics.

TOP

On the side of defining phenotypes, we are probably doing much better, in the sense that using high throughput expression profiling technologies, we actually can approach high throughput of the transcriptome in cancer cells.

We are just beginning to now enter the era where this is hitting the clinic and, at the protein level, similar things are just now beginning to happen.

So, this is a little bit more encouraging.
One of the things I want to talk about today is, can we use this more rapidly progressing area of cancer research to leverage our efforts in mutation detection and discovery.

So, can we harness the relatively fast pace of expression analysis to accelerate gene discovery and the translation of genomics to the clinic.

TOP

Here is sort of the Vogelgram for melanoma, and completely devoid of any representations of chains. So, our task, of course, is to figure out which genes are involved at each of these steps. With only 30,000 or 40,000 human genes, it really oughtn't to be that difficult.

TOP

Of course, we know there is complexity. Actually, this is a CGH study from a former student, Rodney Wiltshire, where he micro-dissected cells from an individual patient's samples during the melanoma progression when they were available.

He noted this important fact, that we could see, even in radial growth melanoma, numerous changes in copy number, indicating that complexity really develops early in these tumors, and I think that gives us some sense of the height of the mountain that has to be climbed.

TOP

So, in terms of genes that we already know about, you have heard about many of them today and I won't read the list, but we have several genes that we know are important as targets of mutational change in melanoma, and the question is simply to find the rest.

TOP

One of the things we looked at early, as already summarized this morning, was the question of when BRAF mutation occurs in melanoma progression. This is a study that Pam Pollock, who is sitting in the back of the room, did in a big hurry after the publication of the Davies paper.

We were really shocked when we found that actually the mutations were found when we micro-dissected nevi and analyzed the secrets of the BRAF gene. We were really shocked to find that BRAF mutations were very early in melanoma progression.

TOP

Although this does cause you to stop and think about whether this will really be a good target, since nevi are quiescent for so many years and, indeed, in the vast number of nevi, they are permanently benign as far as the patient is concerned, however, we also have done some RNAi on BRAF recently, and see that it has a major effect on melanoma cells in culture.

TOP

So, I am hopeful that this may still be a viable therapeutic target, although I would hasten to remind everybody of the treatment failures that have occurred in patients that have been treated with Gleevec. That still means we need to do something more sophisticated than treating any one target.

TOP

Our challenge is still to identify the additional hits that occur during melanoma progression, and to understand what they do to the tumor, to bring about progression, and how they tie into the underlying melanocyte biology. That is really the challenge that we have to overcome.

Even the genes that we know about, such as CDKN2A, we really don't know when they might be inactivated during melanoma progression.

TOP

So, there are two potential approaches to high throughput mutation screens. One is sort of the whole genome, as Mike Stratton is carrying out in his huge project, to take a relatively limited number of samples and large, or essentially all genes and try to discover their mutations.

An alternative is to try to prioritize genes and look at a larger number of samples on relatively fewer genes.

Until we can achieve the $1,000 genome and actually sequence 100 melanoma genomes and find all the abnormalities, we are stuck with this sort of a dichotomy, and I want to take a few minutes and talk about that.

TOP

So, this is really the question, as to how well the total genome approach will work, if we can only do a small number of samples.

Are there many target genes we need to find? Are they dispersed widely throughout the genome, or are there only a small number of targets? What is the situation?

If you can only look at a small number of samples, and it is important to find genes that are mutant in only a few percent, then you can't expect to reliably find those, and it would be more important to think of ways of prioritizing genes.

On the other hand, if there are really only a few genes involved, then really, if you sequence one or two melanoma genomes, you will get the answer. Certainly, the BRAF case makes you think that this would be a possibility for critical genes.

TOP

We haven't been able to address this question directly, but one of the things that Pam Pollock has also done is to sequence all of the known melanoma genes, plus some candidate genes that we have identified through prioritization strategies in melanoma cell lines and, as this data isn't completely compiled yet, I am not going to talk about the specific genes.

This is more what her data looks like at this point. So, there actually are a fair number of genes that we can find the mutations in, at a relatively low frequency.

That is a huge challenge for the biologist. If you find a mutation in a gene, even in a conserved domain in, let's say, three percent, four percent, five percent of tumors, how do you follow up on that?

Each of these is an overwhelming task for an individual laboratory, and it is a challenge as to what to do with this information as it comes along.

TOP

So, what are some approaches we can take for prioritizing for mutation screening? One would be to fully define the transcriptome in melanocytes throughout melanoma progression.

TOP

You are going to hear from the next speaker a good deal about this, so I won't talk about it in detail.

From expression profiling with microarrays where we are now accumulating significant amounts of data, you can identify genes and pathways of interest, you can achieve significant sample sizes.

However, the current data is still sub-genomic and there is also the potential to integrate new approaches for mutation detection with microarrays such as nonsense mediated decay-based screening.

This is really a quite powerful approach to define the transcriptome with the limitation of arrays that it is always so far tending to be sub-genomic.

TOP

That is the reason that we, about a year ago, decided to supplement this approach with a SAGE approach. SAGE is the technique illustrated here, developed by Burt Vogelstein, to sequence representative tags from transcripts in the genome.

Ashani Weeraratna in the lab has made three SAGE libraries from melanoma tissue samples, and these will shortly be available to the community on the SAGE data base through NCI.

We did this because of the difficulty in getting exhaustive sequence data from clinical specimens. So, she carried out a micro SAGE procedure very elegantly and obtained libraries that are fairly comparable to the best SAGE libraries that are in the database.

We think this is a step toward a complete catalogue of the genes expressed in melanoma, and that hopefully will be useful in identifying the genes that are there.

Once you have got the genes, are there things that you can do to find genes of interest?

TOP

I will just mention my favorite example, which may not translate to many other cancers, but this is in gastrointestinal stroma tumor, where we were able to compare the gene expression pattern between GISTs, which of course respond to Gleevec because they contain kit mutations, and some non-GIST morphologically similar tumors, and demonstrate that actually the kit gene itself was the strongest discriminator for GIST in our entire study. This is a result that we have even reproduced in an independent sample.

So, I think you can find the gene that the tumors need to express relatively easily that are characteristic of a particular cancer.

It is tricky when you consider that some of these will be shared between different cancers. So, the comparison group has to be considered very carefully, but I think it is possible to prioritize genes on this basis for a mutation analysis.

TOP

Another approach is to map regions of copy number change, and that is also something that you are going to be hearing about in detail. So, I won't talk about that, other than to say that the technologies are rapidly becoming quite marvelous and we can, indeed, identify regions of gain and loss in the genome with a high degree of precision and a relatively high throughput, and that is a great way to prioritize genes that are undergoing selection during tumorigenesis.

TOP

In addition, we need to think about pathways, as you have already heard. So, BRAF is important.

TOP

It may or may not turn out to be drug-able, but tying it into pathways will be an extraordinarily valuable way to find additional genes.

TOP

In addition, we need, of course, to examine familial predisposition, which has been the great way to find genes in cancer genetics, that turn out to define pathways and potential targets for intervention.

TOP

I will conclude with just a comment that the challenge remains to translate the rapidly growing immense amount of genomic knowledge and data from melanoma genetics into clinical applications, and I think that is what we all hope to do. With that, I will thank you.

TOP

[No text is associated with this slide]

TOP