 Well, thank you very much for that kind introduction. It's a pleasure to be here and I'll be talking, indeed, about searching new therapeutic targets in melanoma. So I have no financial relationships to the close and these are the goals of this talk. So it's well accepted that cancer is a genetic disease and one of the best examples for this is colorectal cancer since it develops morphologically clear distinct phases starting from a normal epithelial cell onto early adenomas, intermediate adenomas, later adenomas, ultimately leading to metastases. And these have been linked to molecular changes, some of which are seen in this slide. And this has become really a paradigm for genetic development in solid cancers. And we've been investigating this in depth, however, we still believe that the data is the tip of the iceberg and there's more to be found out. And we aim to do this for sporadic metastatic melanoma. So there's over 70,000 new cases of melanoma per year reported for 2011 in the U.S. And over 8,000 people died of this disease in 2011 in the U.S. The alarming part is that the median patient survival is only six months following diagnosis of late stage disease in most cases. So we believe that identifying novel genetic alterations may provide new opportunities for drug development. And so similar to what I showed you for colorectal cancer, this is the histological progression for melanoma development where a melanocytes resides within the epidermis and it develops into a benign nevers, a dysplastic nevers, radial growth phase, then it starts invading the dermis in vertical growth phase and disseminates this metastatic melanoma. Some of the mutations associated with these stages are seen in this slide. The best one, of course, is known BRAF mutations. And although more of these are known when I started my program here at NIGRI, there wasn't a comprehensive analysis of the genetics of melanoma. And so our goals were to perform mutation analysis of the melanoma genome. We then aimed to functionally analyze the most highly and most interesting mutated genes. And then we tried to translate this information back into the clinic. We established a tumor bank together with Dr. Steve Rosenberg at the NCI, as well as Dr. Gershon-Wilford N.B. Anderson and Dr. Robinson at the Colorectal Cancer Center. The majority of the samples, of course, are from here at the NCI. You see the numbers here. So we have metastatic tumor DNAs as well as matched normal DNAs. Most of the time, this is blood. There's OCT blocks, meaning the original tumor that was resected, frozen down. For all of the Rosenberg samples, they're matched low-passage cell lines, which are extremely useful, especially if you want to start evaluating functionally the effects of the mutations. You can make matched RNA from these cell lines as well as protein lysates. And for all of these samples, we have clinical information. This is a somatic mutation analysis where you have the tumor, sometimes the cell line, you extract the DNA, you sequence the genes of interest. You do the same for the normal tissue, as I said, mostly blood, and you compare those sequences. And we only follow up on somatic mutations, so mutations that occur specifically in the tumor. So I'll give you an example of this kind of analysis that we performed for the tyrosine kinase family. So this is the human kinome, and on the upper left hand corner, you see the tyrosine kinases. We decided to sequence these because it's known that tyrosine kinases are highly mutated in cancers, and they're known to be amenable to pharmacological inhibition. Now these have been sequenced in melanoma by the Sanger Center in 2007, but they only sequenced six melanoma samples, and so the mutation rate that they observed was too low to definitely implicate these genes to tumor genesis. So now that we had this unique tumor bank, we decided to revisit this family in that tumor bank. So this is the kinome screen, and every one of our genetic screens is made up of two phases. The discovery phase and the validation phase. And so in collaboration with NISC, here at the sequencing center, we sequenced 29 samples, so a low number of samples, for all the tyrosine kinases. We only sequenced the kinase domain because it's very expensive to do the sequencing. It's a lot of work, and one would expect that if any mutations take place, they would be found in the kinase domain. At this point, we searched for somatic mutations, and any gene that had one somatic mutation or more, and there were 19 such genes, moved on to the validation phase, where we sequenced in my lab 80 samples, and in this case, we sequenced all of the coding exons. And then we analyzed to see which was somatic, and then interesting genes moved on to functional analysis. The one that was most interesting was ERB4, which was mutated in 19 percent of the cases. And you see here the schematic of ERB4 and the various mutations that we identified. So ERB4 is part of the ERB family, which consists of EGFR, which is ERB1, EG2, HER2, which is ERB2, both of which are known good targets for inhibition, ERB3 and ERB4. These either homodimerize or heterodimerize on the membrane, and upon ligand binding, they transfer, correlate each other, forming docking sites for various signaling molecules. So we decided to clone seven of the mutations that we identified, and we chose these seven based on the crystal structure that was made by Dr. Leahy that year at Johns Hopkins University, and also based on conservation through the ERB members. What we did is a biochemical assay. We looked at the kinase activity of these mutants versus wild type ERB4, and what you see is that all the mutants have increased kinase activity compared to the wild type and compared to the kinase dead, despite the fact that they express similarly. So clearly these mutants have an increased basal activation. We then did a classic transformation assay in NIHVT3 cells, and what we saw was that these same exact seven mutants formed an increased number of foci compared to the wild type ERB4, and strikingly similar to the oncogenic rats. So these mutants are transforming. The next question was, are these mutants important for the growth of the cells in which they are harbored? And this is where the power of using these low-passage cell lines comes in. So we could go back to the low-passage cell lines and use SHRNA technology to knock down the endogenous ERB4 within the cells, and we could do this in cells that were either wild type for ERB4 or mutant for ERB4. We could then make stable clones and check various phenotypes. The first we checked, of course, was whether this affects growth, and what you see here is that when you knock down ERB4 in cells that are wild type, you see no effect on growth. Similarly to the control. However, if you do this in cells that have mutant ERB4, it reduces their growth compared to control, and these results were seen in several of these melanoma cell lines. So mutant ERB4 is providing a cell survival signal to these cells, otherwise known as oncogene addiction. So the next step was, is this going to be a good drug target? So we decided to test lapatinib, which is a small molecule inhibitor. It's FDA-approved for breast cancer patients, and when we exposed many melanoma cells that are either wild type for ERB4 or mutant for ERB4, to lapatinib, to various concentrations of lapatinib and plotted an EC50 graph, you can see here that the cells that were mutant for ERB4 were more sensitive to the presence of lapatinib, compared to cells that were wild type to ERB4. So clearly a mutant ERB4 is sensitizing the melanoma cells to lapatinib, and we extended this to additional melanoma cell lines. So we do see a range of sensitivity, which is 10 to 250 in more sense, a fold more sensitive than the wild type, there's a heterogeneity, but it clearly seems to be a good potential target in the clinic. The hypothesis being that about 20% of carmelanoma patients would harbor ERB4 mutations, an exposure of their cells would make them sensitive to lapatinib. So based on this, Dr. Udo Rudloff became the PI and wrote a protocol here at the surgery branch for a clinical trial where with the help of Dr. Meltzer at the NCI, there's clear certified ERB4 sequencing, CTEP is providing lapatinib, and this is a multi-site clinical trial, as I said, involving the NCI as well as Memorial Sloan Kettering, and this particular trial is ongoing and recruiting patients, and at this point we don't have any reports, but hopefully we'll have positive things to say about this. So if these are the mutations that we identified in our first study, these are the mutations that have been identified so far through the clinical trial, and you see that the frequency remains the same or even higher, so we're still finding these ERB4 mutations. These mutations have been identified since our publication through whole exome and genome studies. The ones in purple have been identified by Nick Hayward in Australia, and this orange one has been identified by Levi Garroway. The point here saying is that the frequency is still high, and we're also starting to see some mini hotspots, so recurring mutations in particular locations, which is a characteristic of oncogenes. This is some preliminary data through collaboration with Dr. Elenius, where he looked at ERB4 in the sera of some of these patients that are mutant for ERB4, and he actually found an increase in ERB4 ectodomain in these sera compared to healthy individuals on the T test seem to give a significant p-value. So this is actually very interesting because maybe we could have a biomarker to detect these patients rather than sequencing their ERB4. Okay, so I'll shift gears, and I'll start talking about the melanoma landscape because now we can start using different technologies. We can use whole genome and whole exome sequencing. And so this is in collaboration with the foundation of NIH, Eli Lilly, Elliott Mogollis and Stephen Parker. Our pilot was to first compare low-passage cell line to the original tumor from which it was formed. So melanoma cell lines are relatively easy to form compared to other solid cancers, but the main question is the somatic mutations similar in these low-passage cell lines to this original fresh tumor. And so we did whole genome sequencing, as well as whole genome sequencing of the normal of the blood. The main point to be made is that when you intersect the somatic mutations, 90 percent are concordant when you look at the copina at the CDS, at the coding regions. So this suggests that it is fine to use these low-passage cell line derived genomic DNA in order to look for somatic mutations because these are maintained. When you look at copy number variations, these are less concordant. It's close to about 80 percent, but we're focusing now at somatic mutations. So to continue looking at landscapes, we decided to do whole exomes since it's more practical than whole genomes. It's much less expensive. So we use low-passage DNA to perform whole exome sequencing of 14 tumors on their match normals with Agilenture Select and Illumina Sequencing. And any of the dep validation was done by Sanger Sequencing. We got 12 gigabytes of sequence per sample. The mean depth was extremely high, 180x or greater. We got over 90 percent coverage. We only had 2.4 percent false negative rates and 81 percent sensitivity. This is just the stages for this discovery screen. I told you we captured 14 melanoma samples. We got over 5,000 somatic mutations. However, once you assemble the data and you filter it through various different filters, you find about 4,000 somatic mutations that are divided into missense, nonsense, blycides, insertions, deletions, and synonymous mutations. So mutations that do not affect the amino acid sequence. And it's important to capture this because then you can look at the ratio of the non-synonymous to synonymous mutations. In this case, it was 2 to 1, which is what you would expect to occur by chance. So if this is what you expect by chance, suggesting most of these mutations are passenger mutations. Meaning they have a neutral effect. They should not affect the tumor genus process. Okay, so the challenge here is to find out which are the passengers and which are the drivers, meaning which are neutral and which actually have a role to play in tumor genesis. And this is an open question. Many people are working on this. There are no clear answers. But what we use are statistics, bioinformatics, and functional studies. So what I mean by statistics is to look for recurrently mutated genes, meaning hotspots, the exact same mutation occurring several times in different samples such as BRAF, highly mutated genes, non-synonymous to synonymous ratios. And the mutations should occur above the background mutation rate. So if you start looking for recurrent hotspot mutations, meaning the same exact mutation would occur in at least two samples or more. So of course, we found BRAF, that is a positive control. It's known to be a hotspot mutation. But to our surprise, we found nine novel genes with reoccurring mutations. So we validated these in a larger number of samples. And we found TRAP to have six mutations exactly in the same position. The likelihood of this happening is extremely low, you see the p-value. This is just that it's being selected for, and it probably has a functional effect. Now this particular mutation was also identified by Ruth Hallerban in Yale, as well as Paul Meltzer here at the NCI. And so this is a nuclear protein. It functions as part of a histone acetyltransferase. Its disruption causes defects in cell cycle progression. And our functional studies in melanoma cells suggest the same. Okay, so now if we look at the other side, searching for highly mutated genes, we found 16 such genes. Taking into account also the size of the gene and the background mutation rate. And so again, we looked at these 16 genes and validated them in a larger number of samples. And these are the 16 in this table. So fortunately, we see BRAF up here on the top, a good positive control. But all other genes were never linked to melanoma before. These are the p-values and these are the percent of the tumors that were affected in this discovery screen. When we validated this in a larger sample set of 52, you see that the percentages are concordant to what we found in our discovery screen, suggesting that our methodology is working. So we decided to focus on Grin2A. The reason being is it's the second most highly mutated gene and it's novel. So we focused on it and scaled it up in even more samples. And this is the schematic of all the mutations that we identified. You see there's a very large number. You see they're scattered throughout the protein. We have five nonsense mutations, meaning these would be truncating the protein. And we also have a few places where you have recurrent mutations. The same exact mutation. And actually, this particular mutation was previously seen in cosmic, which is a well-known reliable cancer database. And these ones in red were recently published by Garouetel in nature last week, again showing recurrent mutations along Grin2A. Now Grin2A is inotropic glutamate receptor subunit. And I will touch upon the glutamate pathway again later in my talk. So if this is a schematic of Grin2A and it's other subunit, Grin1 and the additional two that are not seen in the schematic. This is really a calcium channel. Upon glutamate and glycine binding, the channel opens and calcium enters the cell, activating and inactivating a few pathways, ultimately leading to apoptosis. Now, Todd Prickett in my lab decided to clone a few of the mutations in Grin2A seen here with red arrows. And first check the binding of wild type and mutant Grin2A with wild type Grin1 by overexpression. And what we saw here is that when you have the wild type, it binds Grin1. However, the various mutants have reduced a no binding to Grin1, even though they're expressed in the lysate. So since this is involved in calcium signaling in collaboration with Sylvia Gutkind, we decided to look at calcium. And so again, if you overexpress these in cells and activate with NMDA, this is the profile you get for calcium entry. However, if you overexpress a few of the Grin2A mutants, you see disruption of the calcium entry into the cells. So our model now is that if these different stars are mutations in Grin2A, we now have reduced calcium entry into the cell, leading to survival, which is what you would expect in a cancer scenario. So this is just preliminary data showing that when you knock down Grin2A in cells that are mutant for Grin2A, you do not see an effect on growth. However, if you do this in cells that are wild type for Grin2A, you see an increased growth. And this is actually consistent with Grin2A being a tumor suppressor. So this is very preliminary, but certainly an interest of our lab to follow up on. So I'd like to remind you that in addition to looking at specific genes, we actually looked at most of the human genes, most of the exons, so we can start looking at pathways. And the pathway that seemed to be most significant was the glutamate signaling pathway. So you see the p-value right here. And Grin2A, which I told you about highly mutated as well as some of its subunits, HerB4, which I told you about being mutated, and there has been a Nature Medicine paper showing that Grin2A and HerB4 interact, and that HerB4 phosphorylates Grin2A affecting its activity. Now, we've also seen Grim3 to be mutated, and this is another glutamate receptor. And preliminary data suggests that it also binds HerB4. So we're now focusing further in this pathway, since we really think it's relevant for melanoma. So in parallel to our exome screens, we have been looking at the G-protein coupled receptors. This is the largest human gene family, and so we had to use capture methods and second-generation sequencing again to look at these and again in two phases. And the gene that seemed to be most interesting, and I already mentioned it, it's Grim3, seemed to be mutated in 16% and 15% of cases in two different cohorts. So again, we have a glutamate receptor here that's highly mutated and has this particular mini-hot spot. So you see this exact same mutation that occurs seen here in the two different cohorts, and again, the p-value of the likelihood of this happening is low, so it probably has a functional role. And just to show you a quick functional analysis, you see here that when these are checked in melanoma cells and activated by an agonist, you see that MEK phosphorylation is increased, especially by this Grim3 hotspot right here. You see a 10-fold increase in phosphorylation in phosphomek. So you do have a functional effect. Now we know that the MEK pathway has effects on migration and invasion, and so one study we did was looking at pulmonary macromatastasis formation, and you see that this is increased after tail vein injection of cells that have a Grim3 mutations. So if this is the paper on Grim2A, you see another paper talking about the glutamate pathway, specifically in this case, Grim1 showing that it induces a melanoma in mice. This is another paper showing as well by people at the NIH showing that Grim5, another glutamate receptor, induces melanoma in transgenic mice, and this is our Grim3 paper, and the common denominator being, again, the glutamate pathway, emphasizing its importance. So what are we planning to do? So we're planning to further delve into the melanoma genome, mainly because if you look at the mutation frequency of melanoma versus other solid cancers, you see that it's extremely high. It's similar to lung cancer, probably due to the UV etiology that melanoma has. And so we really need more samples to be sequenced in order to find out which are the passengers and which are the drivers. And so that's exactly what we're doing. We have, at this point, 18 whole exomes. We have 10 whole genomes, and we can merge the data. And when we do that, we can search for recurrent mutations, and we're finding even more, in this case, that 38. And again, search for highly mutated genes, and we found 15. The good news is that seven of these 15 have previously been linked to melanoma, and these are the seven, BRAF, again, on the top. So this suggests that our methodology is working. It also suggests that we are reaching a plateau, at least with this particular melanoma cohort. So we will further analyze these genomes. The emphasis will be on tumors that are wild type for BRAF, since BRAF has recently had an FDA-approved drug, and we're looking for targetable drivers in tumors that do not have the BRAF mutation. Since we're doing whole genomes, we can look at structural changes, such as amplifications, deletions, rearrangements. We can look at non-genic targets, such as the regulatory regions, the introns. The gold standard, we believe, is doing functional analysis. Since we're doing this in low throughput at this point, we are focusing on Grin2A, ERB4, and the glutamate pathway. And we believe in genetic integration across melanoma data sets. There are many groups doing this at this point, so we are integrating data. TCGA has taken up melanoma as well, and so they will be releasing their data. And so we really hope that integration, such as the one seen here, where we're already integrated with Nick Hayward, and you can see that from the over 1,000 mutated genes that they found, 446 are found in our sample set. We hope this will help us look for common denominators and therefore drivers. So if now the disease is being categorized by clinical and pathological aspects, we hope that in the future, somatic mutation signatures will be added to this particular categorization in order to help to develop more targeted therapies. At this point, I'd like to acknowledge many people, especially in my lab. Todd Prickett, who did the ERB4 in the Grim3 story, Shomu Wei, she's the first author on the Exxon paper. Jared Gardner is doing all this integration on whole genomes, also done by Stephen Parker. Of course, a very, very big thank you to Dr. Rosenberg and his group. Without these samples, we wouldn't be able to do all this work, as well as other sample providers and many other people at the NIH, Elliot Margolis and Stephen Parker, who are helping us with the whole genomes. Thank you very much for listening and I'll be happy to take questions.