Thank you very much for letting me be part of this.

You see that I slightly changed my title from what you had in the program, because I think I wanted to emphasize that all the data I am going to show you is from primary melanocytic tumors, benign and malignant ones.

Obviously, this has advantages, because one is very close to the reality of what is happening in the patients, but also you see the limitations in terms of validation, because this is all fixed material. So, some of this will be descriptive.

TOP

The first slide I want to show you is data with chromosome base comparative genomic hybridization, about 200 melanocytic tumors.
So, here we are using chromosomes as a substrate.

The resolution is rather low, about the band of chromosome that we can see.

What you can see here are DNA copy number changes across the genome. This is a frequency plot of 133 unambiguous melanomas. You see chromosome one to the left, chromosome 22 to the right.

Gains are up in red here, and losses are down in green. You can see about 60 percent of these lesions have lost 9p, including the N4A p16 p19 locus here. About 50 percent have lost chromosome 10, including the p10 tumor suppressor.

You can see it is a non-random pattern here. Certain chromosomes are always lost and never gained and vice versa. Chromosome 7 is very frequently gained in melanoma.

So, there appears to be a preferred genomic deformation in melanomas. Obviously, it is interesting to find out what are the selective forces that lead to this constellation.

I mentioned some of the players BRAF obviously maps right here, and I am going to briefly mention later on that the melanomas that have BRAF mutations actually selectively increase the copy number of the mutated allele, indicating that the mutation precedes this event that we see here.

Obviously, this resolution is actually too low to actually identify novel players here, but already, at that low resolution, it can be informative if it is compared to what one sees in benign nevi.

So, this is a collection of 54 nevi, specifically congenital nevi, blue nevi and Spitz nevi. You can see the profile looks completely different.

There are almost no aberrations present in the benign lesions. There is this one peak here that is a recurrent gain on chromosome 11p, and that happens to appear in Spitz nevi.

You probably all know that Spitz nevi are notorious melanocytic proliferations, frequently occurring in children, that can be very difficult to differentiate from melanoma.

You can see that their genomic profile is actually very different.

TOP

We showed that this chromosome 11p change in Spitz nevi is due to a change in an isochromosome, 11p.

So, if one looks at the normal cell with two FISH probes flanking the centromere, one sees these fusion signals in these Spitz nevi that have this aberration. We see, actually, multiple copies of isochromosomes, 11p indicated by these doublets here.

We were interested, what is the gene that is driving the accumulation of these isochromosomes in Spitz nevi, and Hras happened to map on the p arm here and, when we sequenced that, we were surprised -- this was three or four years ago -- that 80 percent of the Spitz nevi with that aberration had mutations in Hras.

TOP

So, obviously, as you all know, ras maps immediately above BRAF and we were surprised to find an oncogenic mutation in this pathway in these benign lesions. As we know now, by these frequent mutations in BRAF and other nevi which, by the way, the Spitz nevi don't seem to have BRAF mutations. We are just looking at other raf genes in these.

TOP

So, when we think about nevi and melanoma, obviously there is a gray zone here where they are very difficult to distinguish histopathologically. We think that these marked differences genomically that I showed you might be of diagnostic help in the future.

One can think of a limited proliferation of melanocytes and an unlimited proliferation.

Obviously, we would be interested in the starting events.

Obviously, we now have strong evidence that it is actually activation of the MAP kinase pathway that makes melanocytes grow.

Still, we need to find out actually what causes these mutations. Is it UV? Are some patients more susceptible to this proliferative event, or does every melanocyte that requires a pathway starts to grow, or does that need to happen in a specific genomic context which predisposes the patient to melanoma?

We also would like to know what is the event that actually inhibits these melanocytes in nevi to stop proliferating and, as Paul Meltzer pointed out, the role of p16 in this arrest needs to be clarified.

We know there is some relationship with nevi size and age. Nevi can become giant if you acquire the nevi in utero.

It can never become giant if they are acquired later on, suggesting that maybe telomere length may have to do with this arrest here, but these are hypotheses that need to be tested.

TOP

When we look at these Spitz nevi with the Hras mutation, we find that they have high expression of the components downstream in the map kinase pathway.

They highly express the phosphorylated form or ERK. They also express cyclin D1.

They also express very high levels of p16, and that is different from melanomas in the vertical growth phase and, therefore, maybe, their proliferation rate is near to zero.

TOP

So, let me switch gears now and get to the genetic difference between melanoma subtypes.

TOP

You know, there are different clinical presentations of melanoma. There is one that predominantly affects the trunk of younger individuals, mostly presents as superficial spreading or modular melanoma, and that is associated with an acute intermittent sun exposure pattern, and a more pagitoid histological growth type.

There is one that affects older individuals, that is associated with chronic sun exposure, and is characterized by a lentiginous growth pattern.

Then there is one that has nothing to do with UV, probably because of its anatomically protected location. That also has a lentiginous pattern. Then there is one that affects mucosal sites and obviously has nothing to do with UV at all.

We were interested in the distribution of BRAF mutations in these melanomas here, and we found that there are striking differences.

TOP

We sequenced 115 primary melanomas, and you can see the red bars here indicate the mutation frequencies of BRAF, that over 50 percent of melanomas on the trunk of the type that we actually assess by the absence of solar elastosis as a proxy for chronic sun exposure, had BRAF mutations.

Actually, on the face, where we see severe elastosis, massive solar damage, there are almost no mutations.

The mutation frequency is also significantly lower in these sites, and in the mucosal membrane. So, there is a significant heterogeneity of BRAF mutations in different melanoma types.

Obviously, in terms of finding out what is causing BRAF mutations, clearly this indicates that the acute intermittent sun exposure pattern is somehow related, but sun obviously had a much greater effect in this body site. So, the relationship to UV seems to be complicated.

TOP

I showed you that the mutation frequency varies significantly with the anatomic site and sun exposure patterns.

We had follow up of 90 of these patients over at least three years, and there was absolutely no association with outcome in this group of melanomas.

We also looked for an associated melanocytic precursor, where there was a higher frequency of BRAF mutations in those melanomas that had an associated nevus. That was actually not the case.

I showed you that, when we looked at the copy number of the BRAF region by FISH and array CGH, we found that these tumors that had mutation and a copy number increase, selectively increased the mutative allele.

TOP

That indicates that BRAF is at least one of the selective forces that drives this frequent copy number change of chromosome 7.

Now, I am going to the recent development of CGH, where we use microarrays developed in Dan Pinkel's and Donna Albertson's lab at UCSF, and we have a much higher resolution, in fact, about 1.4 megabases over the entire genome.

TOP

I just want to show you, when you compare these melanoma types -- here I have just three groups -- when you compare acral melanomas, mucosal melanomas and skin.

The cases go this way here. So, this green line here would be one case. This is chromosome 11, the centromere, the p arm -- some of this didn't come out here -- and this is the q arm, and the gains are in green and the losses are in red.

You can see that there are very frequent gains or amplification, in fact, very focused high copy number changes in acral melanoma that affect this region. They are infrequent in the other melanoma types.

In it you have a few homozygous deletions in mucosal melanomas in the distal p arm.

TOP

Mucosal melanomas also have much more frequent deletions on 4, which are infrequent in the others,

TOP

and also more frequent in eight.

 

TOP

This is an example of one of those acral melanomas. This is chromosome 11. This is the copy number along the axis. This is the location of the centromere.

You can see that there are multiple peaks, very high copy number. So, this would be about eight-fold here.

This one here, the most frequently amplified region in these melanomas is the cyclin D1 locus. We could show that this is actually amplified in 50 percent of acral melanomas.

When it is amplified, it is always over-expressed by immunohistochemistry, and together with Meenhard Herlyn lab, we did some studies where we tried to knock down the expression in cell lines that amplified or over-expressed it. That led to massive apoptosis of these cells, indicating that these are a player here.

TOP

When we summarized that, we have one melanoma type here that is associated with this chronic sun exposure, lentigo maligna, kind of prototype.

We don't find BRAF mutations, so we have to find out where the players are. We are currently looking at the other genes in the pathway.

Chromosome 17 is frequently lost. p53 may play a role here, and we also see frequent deletions of chromosome 13 in this type of melanoma.

In the most common type of melanoma -- that is, the one that is associated with BRAF mutations -- these are the typical aberrations that we see in those. Actually, the ones that I show for the others are the ones that differ from this most common type here, and that is associated with histology.

TOP

I just want to mention, when we compared the other types, we did multiple stratification.

The histology, the pattern, lentiginous versus pagetoid, doesn't play a major role in separating those.

When we look at the acral melanomas, all of them actually do have these multiple amplifications scattered over the genome, whether they are SSMs by histology or nodular.

As long as they are non-hair bearing acral skin, they have this phenotype. We call this amplifier genotype. They have gene amplifications, as I showed you, all of them, without exception, whereas the others have about 10, 15 percent frequency of gene amplifications.

These amplifications frequently target 11q13. Cyclin D1 is a candidate here, 5p15, 4q11, and then the mucosal melanomas are completely different in terms of the distribution of the amplifications.

One of the candidates may be CDK4.

TOP

I am just showing here the overview over the entire genome, frequency of aberrations for these three melanoma types, four genomic loci.

You see 11q here amplified in 50 percent of acral melanomas. These mucosal melanomas also have frequent amplifications, but they occur at very different sites.

Here, 12q is one of them and maybe CDK4 is a target here, but this differs, when we use these statistical comparisons to classify these lesions just based on the copy number, we get 88 percent accuracy in separating them, just based on their genotype.

So, what I am trying to say is, there clearly is a genetic heterogeneity here among these melanoma types. When we want to look for targets, we have to take that into account.

TOP

One interesting thing I want to finish with is these amplifications in acral melanoma were found to occur very early, in fact, at the in situ stage.

We used fluorescence in situ hybridization for the amplified loci, and mapped the copy number over the tumor. At the invasive front here you see multiple red signals here.

The copy number in these in situ portion here was similar, indicating that these amplifications were already present at the in situ phase.

What was interesting about this study is, by chance we identified that the normal appearing epidermis adjacent to the in situ portion here had single basal melanocytes that were distributed as normal, looked histologically normal, and also had these amplifications.

TOP

We called these cells field cells. This is a high power of this. You see the dermal/epidermal junction here. Here, the amplification is targeted with a green probe.

You see these single basal melanocytes that have very high copy number, in this case of a mutated Hras here, that are in histologically normal appearing skin.

Five millimeters away from the histologically detectable melanoma, we have some indication that these cells can lead to local recurrences if not excised with a safety margin.

So, we think these are one form of residual melanoma that we try to remove with our safety margins which, as you know, are entirely empirically based.

TOP

So, what I conclude from the field cell observation is that, in addition to the vertical growth and in situ growth phase of melanoma that we all know, there is actually a phase of melanoma progression that precedes in situ growth where the number of melanocytes is still intact, but they already carry genetic abnormalities.

TOP

That would be a field effect, and I am very interested in what the relation of these freckles that all Caucasians carry on their sun exposed skin in much higher numbers than can be appreciated by the naked eye, if you use the right imaging technology, is to these genetically altered fields, and that will be an interesting field of research. I am going to stop here, and thank you for your attention.

TOP