Thank you. I have one con remark. Everything that Jeff has just told you is wrong.

Usually I am a very optimistic person in nature. So, when Vern called me and said, you have to wear the con man hat, reluctantly I said, I will do it, but it is going to be very difficult for me because I, myself, use these methods and I think this method is going to tremendously contribute to our understanding of the etiology of cancer, and especially in melanoma, provided -- and I emphasize provided -- that we recognize the limitations of these tools we just heard about.

So, let's review together what are the limitations of these methods, and potentially where these methods are going to help us, and where we are going to use them.

TOP

So, let's review some of the limitations that microarrays and CGH have. So, first of all, there is the issue of sensitivity.

The genes expressed at low levels are frequently missed in these methods. So, there are some transcription factors that, even after induction, you get one copy per cell induction, which is going to be very hard for the microarrays to detect them.

We heard about the BRAF that are expressed, the mutated BRAF expressed in very, very low levels. So, it is going to be very hard for us to detect it. So, this is one limitation that we need to realize.

The other limitation is the gene isoform. Genomic complexity is derived from isoform expression. Only a few microarrays currently distinguish between the multi isoform.
Then there is the issue, are we going to detect all the mutations that we know, especially if, for example, one example is BRAF.

The most important issue that I think is the discrepancy between the messenger RNA and the protein. Changes from messenger RNA do not necessarily mimic protein expression changes.

We need to combine these tools together with proteomics, as mentioned here, and even the proteomics profile, it is not going to tell us much. We need to know the function of these proteins, so we need to run protein protein association.
In the case of transcription factors, who are the partners and what are the target genes?

TOP

So, let's review what did we learn from microarrays so far? Well, there was one really positive hit, which was the ROLFC.

We heard about another gene, which was the WNT5, and recently our lab discovered that the thrombin receptor PAR1 is over-expressed in highly metastatic melanoma cells.

The vast majority of genes still lack assigned function. We have so many genes in these arrays, some of them are just the sequence, PCR amplified. We don't know yet the function of genes, which limits the ability to select and declare critical target genes to treat cancer or, in our case, melanoma.

Then there is the issue of purity of samples. You take a tumor. It is contaminated with other tissues. Of course, you can say, now we have a lesser micro dissection. We can really take only the tumor cells.

Then we run into the issue of how much RNA can you take from these cells, which really limits our ability to analyze pure samples.

Then there is the issue of delayed versus early stage. The late stage, by definition, those are highly proliferating cells and there is lots of noise with genes that do not relate to the progression of melanoma per se.

So, how reliable is it to obtain the profile? I would submit that the number of patients can change the clustering, and I am going to show you an example in a second.

Then, one issue that really we need to consider is the lack of standards for presenting and exchanging the data. We are seeing all these beautiful colors between red and green.

So, one lab would produce it, the other lab would not. We need really a consensus of how we are going to present and how we are going to exchange this data.

TOP

Let me show you an example that Vern submitted to me by e-mail. This is the SWOG group analyzing 100 melanoma patients for clustering.

 

TOP

[No text is associated with this slide]

TOP

Just by adding 10 patients here, 79 of the original patients moved in the clustering. So, we need to be very, very careful in our analysis.

TOP

Then, in the future -- and I see it as the very near future -- we are going to be bombarded for information by the guys doing these analyses.I would like to submit to you that, even if we identify targets, such as the WNT5A or the ROLFC or others, first of all, we need to verify and, second of all, we need to know what is the biological function of these genes before we declare them officially as a target.

For instance, the WNT5A was discovered. What do you know about its function? We need to put it back into the cells that don't express it, we need to inactivate it in cells that do express it, put it back in nude mice to see how it is affecting tumor growth and metastasis, and we need to do that for every gene that we will identify by this method.

TOP

Let's go back now to the issue specifically on melanoma. Will it replace or supplement morphology pathology, for instance, for classification?
In my opinion, we are not yet there. For instance, Golab Banai(?) and his group have shown that they can do profiling and they can distinguish precisely between AML and ALL.

We cannot do it yet with melanoma. We cannot give it the samples and we cannot tell, this are metastatic melanoma and this is RGP, and these are upstream melanoma. We don't have this tool yet for melanoma.

Metastatic melanomas are not all the same. There is heterogeneity, which will bring me to the next subject. Are we ready for a melanoma chip, as Jeff alluded to you? I think we are not.

We cannot just take 60,000 genes or 40,000 genes and put it. We need to narrow it down to the genes that are associated with the progression of melanoma.

How many genes is it going to be? We don't know, 50 maybe, but we haven't identified these genes yet, and we are not ready for melanoma.

So, where can we use these tools? For classification, we are not yet there. For prognosis, maybe.

One area that we can really use these tools is for monitoring treatment. Let me give you an example, the Gleevec one. We talked about Gleevec so many times today.

I am not saying that Gleevec is going to work on melanoma, but let's say we want to evaluate what is going to be the effect of Gleevec after treatment.

So, let's discriminate here between two issues. Gleevec is supposed to block the phosphorylation of PDGF alpha and beta receptor, c-kit and BRC Abl.

C-kit is lost within the progression of melanoma. Probably c-kit is not going to be a target. Now, we need to monitor whether PDGF alpha and beta receptor is phosphorylated or not.

With the chip that we have in our hands right now, we will be able to discriminate between activated and non-activated receptor? Maybe in the future it will, but not right now.

If the Gleevec is going to change as a result of the blocking of the receptors, the signalling pathway in the cells, and we will be able to show that there is differential expression of protein, then yes, we might be able to use these tools to monitor treatment of melanoma.

This is where we are.

TOP

I thought that I will end up again with the question that Vern has presented to us. The answer is probably, we need the combination of both.

We need the help of the pathologists like David Elder, and we need the molecular biologists, combined together. In the future, we will have the right answer. Thank you.

TOP