Thank you. I would like to thank Dr. Saxman and the chair for this opportunity. I have been asked to talk about RAS and p16 but, really, I would like to use this opportunity to bring to the table a bigger question, which is how do we translate our targets into discovery, which is the title of this State of the Science meeting.

These questions will be the focus of discussion tomorrow, but I would like to begin by putting on the table and sharing with you some of my thoughts.

I think one of the very important questions that we need to talk about tomorrow is what constitutes a rational target for therapeutic development.

Some of my thoughts are listed here, and we will obviously address this, which is the availability of genetic evidence in human disease, the requirement for this lesion in tumor maintenance, and then tractability, which is probably, most of us in academics are not going to be in a position to address.

What I would like to do today is to use the mouse, talk about how the mouse can be used to address these questions, and I will use the RAS and p16 in a melanoma model as an example to try to address how we can determine the role of these genetic lesions in tumorigenesis, how to correlate the genotype to the phenotype, and how do we use them to discover alternative targets and identify biomarkers.

What I will hope to say is that the points I would like to make about the mouse as a model system is applicable to other model systems.

I think other model systems have their strengths, advantages, and limitations, and I think understanding what those are will allow us to use these model systems.

TOP

To start with, as you just heard from Peggy, the genetics are not entirely understood, but there are some common themes that have been described.

Deletion at the 9p21 locus, which encompasses the CDKN2A or N4A of tumor suppressor gene, that has been found, it depends on what criteria and subset of melanoma, familial melanoma, and it is also found in a smaller percentage of sporadic melanoma.

Activating mutation in RAS has been described, but consistently at a very low rate compared to the role of RAS in other solid tumors, but it has been described consistently in about 10 percent of tumors, and typically is correlated with some exposure, and p53 mutation is infrequent, in contrast to most other solid tumors.

So, to try to develop model systems that would allow us to verify the oncogenic role of these mutations, we developed the mouse model where we express or activate RAS in the melanocyte, and cross that transgenic allele to the mice that are deficient for N4A/ARF or p53. We did the study on both

Just to summarize -- these are published studies -- what we found is that these animals do develop a melanocytic tumor that is confirmed by their immunoreactivity to TRP1.

Histopathologically, they share commonality with nodular melanoma in humans but, most important, we have identified secondary genomic changes in regions gained or lost, that is similar to gain and loss that is described in human melanoma and if I may have time, I will show you one slide on that.

The point here is that, in order to use a mouse model for a particular type of study, we need to know what that mouse model is reflective of.

Traditionally, we have characterized these tumors in the mouse using routine immunohistochemistry characterization. These days, I think we have the tools to go a big step deeper, which is to look at genomic changes and genetic changes.

What I am alluding to here is that the tumor that developed in the mouse developed alteration in the genome, in the chromosome, in a region that is attendant to human melanoma, suggesting that the process in the mouse to get to the point of having an established melanoma recapitulates some of the genetic events in humans. That provides us a system to study what those events are.

As Peggy alluded to, the 9p21 INK4a/ARF locus actually encoded for two different transcripts, p16 upstream of the RB pathway, and p19 upstream of the p53 pathway.

TOP

As an illustrative example of how we can do genetics in the mouse, which we cannot do in humans, we can look at the human data which shows that the involvement of the RB pathway in melanoma is clear, because there have been descriptions of germ line mutations of p16 INK4a in familial melanoma kindred.

There is the Arg24Cys (R24C) mutation in Cdk4 which abolishes interaction with p16. So, this pathway is clearly involved.

Up to very recently, the role of the p19, p53 pathway is not so clear. First of all, germ line mutations affecting ARF were not identified until very recently and, as I mentioned, p53 mutations are very rare.

So, how does one go about addressing the question of whether both products of this particular tumor suppressor contribute to melanoma suppression in vivo.

TOP

This is a question we can address in the mouse model, where we take the melanoma prone transgenic allele, cross it onto mice that are deficient for either the p16 or p19 tumor suppressor.

What we found is that both loss of either p16 or p19 can cooperate with ras, leading to the development of melanoma, indicating both of them function to suppress melanoma development in vivo.

Further, in the tumor that emerged from the p16 deficient background, we can identify mutations in the p19, p53 pathway and vice versa in the tumor from the p19 null background we identified mutation in the RB pathway.

This suggests on the level of tumorigenesis, inactivation of both pathways is required. That allowed us to conclude that inactivation of the pathway is achieved in melanoma that has the CDKN2A or N4A/ARF deletion, by disrupting the two upstream components, p16 and p19ARF.

TOP

Now, that is just an example. So, that study basically allowed us to conclude that the two products are important, and that obviously is in concordance with the human data showing now that germ line mutation that has been described.

Now, what else can we do with such a mouse model? We can look at the interaction of these genetic lesions with environmental risk factors.
For melanoma, UV obviously is an important factor. So, we looked at this using the p19 null model, where we asked, can you expose these newborn mice to UVB, a single dose, and does that affect the latency and penetrance in normal development.

This is the Kaplan-Meier curve showing that, in animals that have been exposed to a single dose of UV at day one of life, there is a significant shortening in latency to melanoma development, in addition to increasing in penetrance.

What is not reflected in this curve is actually the additional increase in multiplicity.

TOP

When we analyze these tumors, we obviously first looked at the RB pathway, because these are tumors from the p19 null model, and we anticipate, and we expect to see RB lesions.

What we found is the emergence of a new genetic event which is only seen in UV-induced tumors which are amplifications of the CDK6 locus.

Moreover, the tumor then harbors CDK6 amplification exclusive of the tumor that had loss of p16N4A.

Given the relationship between p16 and CDK6 in this particular pathway, that particular genetic profile would suggest that the RB pathway is being targeted by UV.

That is confirmed when we do a study on p16 deficient background. In other words, if the UVB action is to target and inactivate the RB pathway, and we expose p16 deficient animals to UV, there should be no effect, since the pathway is already inactivated.

This is precisely what we saw in the study. These are p16 deficient animals. You can see that, in contrast to the Kaplan-Meier curve I showed you for the p19 null animal, there is no effect with UV.

So, by exposing a p16 deficient animal to UVB, there is no cooperation with RAS or p16 deficiency by the UV.

Now, this is actually -- obviously, this is a study done in a mouse. The obvious question one would ask is, how does this relate to human disease.
That is always going to be a question we have to ask with any question done in any model system. For example, it had been described that CDK6 is over-expressed in human tumor.

The other interesting point is that I want to remind you what Peggy just mentioned to you. The penetrance of CDK into a germ line carrier varies with geographic area.

The cumulative penetrance I believe is highest in Australia and lowest in Europe. That penetrance paralleled the baseline population penetrance which is believed to reflect, to a degree, the UV index of the region.

So, if you look at the molecular genetics in this mouse model, you can speculate that, perhaps, the reason that the penetrance of CDKN2A germ line carriers is higher in Australia is because a higher UV index in that region leads to more frequent and more likely inactivation of the second allele in the carrier.

That would explain, at least in part, the variation in the germ line penetrance of the germ line carrier.

TOP

Now, I said I will try to show you an example how this type of study in a mouse can relate to human disease.

This is actually a slide that I stole from Boris, Bastian, who is going to talk later about where he had compiled alterations in human melanoma with respect to gain and loss by CGH conventional and array CGH.

What we did, we looked at the UV and non-UV tumor from the model I just told you about, and we profiled them also for amplification and deletion, using array CGH.

We found that, among the many changes we identified there are 20 top discriminators that are powerful in separating these two groups.

If we map where these 20 changes are that we identify in the mouse, you can see that many of them overlap with what has been described in humans.

In fact, one of these regions is a proximal chromosome 5 gain which, by analysis of multiple tumors, identified a minimal region which harbors three genes, one of which is CDK6.

So, we have, in addition to the candidate gene approach, where we analyze the genetics of a particular tumor based on our understanding of the pathway, we have also taken a sort of unbiased global approach, and came to the same conclusion, that a gain in a CDK6 locus is a key discriminator that separates a UV versus non-UV induced tumor.

So, these are the types of studies that we can do in model systems that, for obvious reasons, one cannot do with humans.

TOP

Now, I will end, in the last two minutes, with a little bit on the RAS, which is, if we identify a genetic lesion and show that it is involved in tumorigenesis, does that automatically make it a valid target for therapeutic development?

I will argue no. As I told you, in this model system, we show you that we engineered a couple of initiating mutations that have been described in humans.

The melanocytic cell acquired an additional mutation that we can document and ultimately we have the development of a tumor.

At this point, there are many other mutations in addition to the initiating mutation, such as with RAS. So, the question is, is RAS still required to maintain this established tumor?

That is highly relevant if you are talking about treating a patient with tumor as opposed to preventing development of melanoma. If it is involving maintenance, what is the biological role of that gene?

So, the way we are doing this is, we can engineer the model system in such a way that we can genetically inactivate those lesions and then ask, does it cause tumor regression if we inactivate that mutation and, if it does, what are the mediators leading to regression?

In other words, if we shut off RAS and the tumor regresses, what is responding to RAS down regulation that leads to tumor regression?

By identifying what is the phenotypic and genetic correlate of RAS inactivation, we identify potential alternative targets that could lead to the same end result.
Sometimes the particular target you are looking at may not be a very drug-able target. In fact, I think RAS is a good example of that, that we have not been able to develop a good drug against RAS. This may be one approach to identify alternative targets.

TOP

This is to show you that, indeed, this is what happens. If you inactivate RAS in the established tumor, the tumor regressed and, moreover, what you see here shows apoptosis in the endothelial cell.

So, as part of the phenotype during tumor regression, tumor angiogenesis collapses, so clearly indicating that RAS is involved in tumor maintenance by maintaining tumor angiogenesis.

That would be one particular biological function of RAS that you would not be able to predict based on a cell cultured based study.

TOP

Given that RAS is involved as a tumor maintenance target and, therefore, it makes sense to develop therapeutic drug against RAS, RAF is positioned immediately downstream of RAS, and it would be logical -- and I think most people would anticipate, that RAF would play a maintenance role in melanoma.

TOP

However, I would like to remind you that there are many effectors downstream of RAS, and I listed some, and RAF is just one of them.

I think we do have genetic evidence that RAF mutation is involved in melanocytic neoplasia, but we don't know that RAF -- I think we don't have evidence supporting that RAF -- actually plays a maintenance role and we don't know, if it does play a maintenance role, what its biological function is and, therefore, what are the potential biomarkers we can use for its efficacy and toxicity.

With that, I actually would leave this up for our next speaker, who is going to talk about BRAF. I would like people to think about this as we talk in our discussion tomorrow on additional genetic targets, or maybe even BRAF as a target for therapeutic development.

TOP