 Good morning. Thank you very much for this opportunity to present here. It's very exciting for me. I'm a surgeon working with Dr. Shea, James Shea, Sloan Kettering, and we're part of the KINI TCGA project. We've had a distinct opportunity to work with Chad Creighton and Richard Gibbs on this project, and we're going to present some of our work that we've submitted as a companion paper to Journal of Clinical Oncology related to some of the clinical correlations that we can get from some of the sequencing data. So just for background, RCC is the sixth leading cancer cause of death in the United States. There are several malignant subtypes of which clear cell has been the focus of the initial TCGA project, although they're popular in chromophobar also in the pipeline. Nearly one-third of these tumors present with metastatic disease, and these tumors are characterized by the fact that they're chemo and radiation resistant with the urgent need then to identify novel pathways for treatment. In the past two years, there's been explosion of sequencing data, largely from the Sanger Institute, that has identified several recurrent mutations, some of which we've talked about extensively in this morning. VHL was really the only known gene to be mutating kidney cancer for many, many years due to the pioneering work of Mars and Linehan, et cetera. But in the last two years, several chromatin modifiers have been identified, including PBR-1, CET-D2, and BAP-1. And the fascinating thing about these genes and the reason why we were interested in them initially even before the TCGA project really started to confirm this was that these were all located on 3P21. And, you know, in fact, VHL is on 25, but these three other genes are really reside within the same location, so this is really an unparalleled phenomenon to have potentially four tumor suppressors lying in the same region. And the reason why I say they're tumor suppressors is because one of the hallmarks of kidney cancer, this is from the TCGA project, is the fact that 3P is lost in 90% of the samples. So the fact that you have four genes in the same region that's lost in 90% makes this very intriguing topic for us, and therefore we pursued any clinical ramification of these. As I mentioned, PBR-1 is one of the ones that was reported in nature in 2011, and this is a classic tumor suppressor in that you see mutations that really span the entire length of the coding regions of the gene. And this has been shown in this paper, which is really just defining this as an event, as being associated with tumor proliferation, but really made no assessment of the clinical ramifications of this gene. The other thing that was intriguing to us was that this more and more papers that were sequencing from other institutions were reporting these genes being frequently mutated. This is from the B.A. Genomics Institute, which also highlighted these four genes. And as you've seen from a couple of the presentations this morning, these are very significant regardless of how you analyze the data. The other thing that was interesting is that these are all part of the histone and chromatin modifying family. And one of the interesting things from a therapeutic sample is that some of these genes now, they're designing specific drugs to target mutations in these pathways. So potentially, besides just being a prognostic marker, there could be treatment-related design based on these alterations. In terms of what's been known about kidney cancer and clear cells, specifically, VHL mutation, as I mentioned before, was really the only known frequent event until about two years ago. And in terms of the utility as a prognostic marker, there's really been no evidence to suggest that a VHL mutation alone or the location mutation, the type of mutation, has any bearing on prognosis. And that has to do with the fact that VHL is probably lost upwards of 80, 90 percent. Obviously, it depends on how pure the tumor samples are and how deep you sequence, but you see mutations very frequently, certainly upwards of 60 percent, and the methylation rate is about 7 percent. So a couple of that together, it's probably a fundamental event of carcinogenesis as opposed to a prognostic marker. So what we did initially was we sequenced these four genes in 188 tumors from our own cohort. This was distinct from the TCGA, and we use this as a discovery cohort. We looked at pathologic correlations associated with tumor size, grade, and stage, and we looked at clinical correlations. And then we validated this data with a much larger data set than TCGA, which is four-engine 24 samples with whole exome data available. Now, one thing I want to point out here, which I think is very important for anybody that's going to be looking at any TCGA project in terms of clinical data, is to really understand, and this is where the key is, I think, involving both the computational biology of the clinicians and pathologists, is to understand what is an event. So a lot of the tumors that are sequenced in the TCGA, certainly some of the earlier projects like GBM and high-grade ovarian, most of those patients were dying from their respective cancers. However, the TCGA project for a clear set was a very well-balanced cohort of patients. These were a lot of patients with early-stage disease, intermediate-stage disease, as well as advanced-stage disease. So when you actually look at the number of deaths from overall survival, 144 patients died in the cohort. But when we actually sat down and I sat down with one of our biostatisticians to work on this, about a third of those were non-cancer-related deaths. So it's very critical to understand, in terms of identifying events that are associated with poor survival, is to really understand what is going on in that particular analysis. Anyway, but our two cohorts were very well matched in terms of clinical features. There were more events in the TCGA dataset. Our tumors tended to be have slightly less follow-up and less cancer-specific mortality. This is an overview of the mutation between our two cohorts in terms of the frequency and the patterns of overlap. And you can see, obviously, as I mentioned before, 3P is lost. This is where all the genes lie, and the fact this isn't a race CGH, showing that at least one copy is lost in about 90 percent of the samples. The frequency of overlap was very similar between the two cohorts in terms of VHL being mutated at least 50 to 60 percent, PBM1 in the 30s, and CT2 around 11, and 8 percent respectively. And the types of mutations were also similar in terms of the spectrum of truncating versus mid-sense mutations. This is our cohort of the 188 showing very similar patterns to what's been shown previously, that these, although there is some clustering they tend to be spanning across the genes in these, just, again, the hallmark of a tumor suppressor. So in terms of associations, this is something that we noted right away in terms of our cohorts. This is showing the associations with size, stage, grade, necrosis, lymph node, metastases, and meds. And these are all the classic factors that we use as clinicians in terms of prognosis. These are what are in our multivariate models. And we looked at right away, we saw some striking associations with one particular gene, which is BAP1. We noted that BAP1 was associated with essentially every single poor prognostic factor, larger tumor size, higher tumor stage, higher nuclear grade, and metastases at presentation. Interestingly, in our PBM1 cohort, we saw, sorry, in our MSKCC cohort, independent cohort, we saw an association with advanced tumor grade. However, what was interesting about that was that that was associated primarily with small tumor size. So in small tumors, in clear cell, tumors that are less than four centimeters, about a third of them will eventually become invasive, even at the small level. And what we found in our cohort was that the PBM1 mutations were occurring in those patients with the invasive tumors on a small level. And the actual frequency of a tumor invasion in the TCGA cohort was much lower. And we think that has to do with the meticulous dissection that our pathologists do at our institution in terms of upstaging patients in that smaller cohort. But we also noted, though, that BAP1 was associated, as I mentioned, with all these adverse features. And this was reassuring because a recent report in Nature Genetics found that by James Berguelis's group from UT Southwestern found that BAP1 was associated with high tumor grade and actually further validated the fact that this is a very important mutation in terms of prognosis. Now, many of you may know that BAP1 is a critical player now in other cancer types, most profiling in uveal melanoma, which really defines the poor prognostic group of tumors. It's also been shown by Mark Ladani and others in mesothelioma as a critical player. So this was exciting to us to see that we were able to recapitulate our findings. As I mentioned before, the associations with tumor invasion in small tumors, we found that not only PBM1 in our cohort, but also any mutation in any of these genes, BAP1s and D2 were associated with tumor invasion. Now, in terms of survival, this is our MSK cohort. We saw associations with poor survival for BAP1, a very strong association even though we had relatively few events. We then looked at that cohort in the set, we then looked at the TCGA and we found very similar associations not only with BAP1, but also set D2. And again, if you would look at overall survival alone, there is not a strong association with set D2, but when you pare down and you actually look at cancer specific survival, and again, this is something I really want to emphasize, you definitely see a strong correlation now with set D2 mutations in poor survival. And this is really the first time in kidney cancer that we're able to report a gene being associated, a mutation being associated with poor survival. We also found, we also were able to, with the help of Kim Rathmel and our bio statistician, Irene Ostrov-Nyaya, we were able to look at time to recurrence. These are patients that have no evidence of disease after surgery, but then develop new disease. We found a very striking phenomenon where these patients were much more likely to have, these tumors were much more likely to have a set D2 mutation. So we now think that tumors with BAP1 or set D2 mutations are critical for tumor progression, and they're also associated with poor survival. Finally, with the help of Boris Riva from our Computational Biology Department, we did a pan TCGA analysis to understand what are the additional events that tumors acquire as they progress. And I look, and I show you this box plot here showing that as tumors progress, and these are the staging system that we use by tumor size, they acquire the number of parentheses here as a median number of tumor suppressors. And we define those as genes that are truncated, mutations that are truncating mutations in the setting of a copy number loss. So high functional mutations. And Boris actually has a poster on that in which I encourage everyone to see where he kind of talks about this concept. But we looked at the additional acquisition that tumors have as they progress. And what you can see is that there's just an additional one to two to possibly three mutations that tumors acquire from a truncating tumor suppressor standpoint as they progress, as they increase in size, as they increase in grade, as they increase in stage. And what this shows here is that the number of times those mutations are populated by BAP1 or set D2 goes up. So what I would suggest is that this is the full change, and this is BAP1 here, and this is increasing tumor grade and increasing tumor stage. What I would suggest here is that this analysis shows that the additional events are relatively few, and they're probably populated in a large part by these genes on the 3P. And these are probably critical drivers for tumor progression and ultimately leading to cancer-specific mortality. And obviously these genes have just really recently been reported, the function of them, other than the fact that they're involved in chromatin modification and remodeling is relatively unknown and we're actively working on that, on those components in our lab. So basically we've shown, we confirm the frequency of these non-mutations, CCRCC. We show that they're associated with adverse tumor features, pathologically, and survival. Further studies with longer follow-up will be necessary to assess the clinical impact of these additional mutations. One thing I'd finally add is that in terms of adding to a prognostic model, which is one of the goals of a biomarker study, because BAP1 and CT2 were so linked to adverse tumor features that are already in our models, they're unable to add to a specific model. We are currently now, and we have some exciting data which we're working through right now in terms of mutations. Copy number to look at additional factors we can do controlling for the current prognostic factors to add to the models, to theoretically add a new mutation, a new gene, a new loss or gain into the current models. Thank you very much. I would just like to close by thanking this really multidisciplinary group that worked together on this project in terms of the surgeons, computational biologists, medical oncologists, biostatisticians, and all the members of the TCGA. Thank you very much. Questions? Nice talk, thank you. One question about your tumor association table, where you listed the association of the mutations in different genes with pathological features. And first in the T stage, you said higher stage associated with like three or four genes. But for the T, I mean size larger than five centimeters, only the last one was associated. I'm confused by this result because if higher stage is associated with multiple genes, then when the size is larger than five centimeters, it's certainly also higher stage. So that phenomena we saw was really only in the small tumors. So we saw that tumors under four centimeters were acquiring invasive characteristics in our MSKCC cohort. And that was really looking specifically at tumors under four centimeters. So that's the distinction that I made. I think now I understand it. Thank you. Do you have any information on treatment and do you know how they impact the survival data based on the mutations? That's an excellent question. As my PI James mentioned a little earlier, James Shea, we are currently working with the TCGA group, the clinical TCGA group, which we founded, to capture, we have about 100 patients that have gone on for target therapy and that's one of the first things we're going to look at. Obviously, that would be a huge factor. Thank you.