 I'll be talking about the emerging targets in renal cell carcinoma, specifically set D2. I have no financial conflicts to disclose. So over the past couple of days, you guys have probably seen some of the emerging targets and some of these mutations in histone-modifying enzymes, one of them being set D2. It's a member of the set domain family of methyl transferases, and it contains a set domain which is responsible for the methyl transferase activity. Similar to BAP1, VHL, PolybromO1, it's located on chromosome 3P, which is an area associated with loss of heterozygosity and greater than 80% of renal cell carcinoma tumors. Sort of in fellowship, instead of picking the first or second most frequent mutation, I thought I could kind of slide under the radar and pick the third most mutated enzyme in renal cell carcinoma, and at the time mutations were reported single digits. But with recent data, it's more double digits. So let's cancel that. So my hypothesis for this project is that the loss of set D2 histone methyl transferase activity alters the cancer phenotype. You have VHL inactivation, 3P loss of heterozygosity is one of the earlier events, followed by mutations in these histone-modifying enzymes. And I was particularly interested in the subset of patients with set D2 dysregulation, which would be evidenced by decreased histone-3 lysine-36 trimethylation. The prediction is that if you have a loss of function of an enzyme function, it will decrease histone modifications as a result. So one of the things I wanted to first do is actually look at patients with VHL mutations, polybromomutations, and stratify the patients by set D2 mutations versus set D2 wild type. I optimize the IHC staining protocol for histone-3 lysine-36 trimethylation. And here when you compare the nuclear stain for the uninvolved kidney, compare that to the set D2 mutant tumor, you see a decrease in histone modification as you would predict from the model. The other thing to note is that you actually see as an in-slide control is that there is still staining of the lymphocytic infiltrate as well as the stroma. If you compare this to the uninvolved kidney to set D2 wild type tumor, so these patients have VHL mutations, polybromomutations, but wild type set D2, you don't necessarily see a decrease in histone-3 lysine-36 trimethylation. So just using a simple IHC screen, you could actually potentially identify patients for further sequencing, and you could exclude patients, potentially, who don't have a decrease in histone-3 lysine-36 trimethylation when you're trying to screen for set D2 loss of function. If you use a computer to actually analyze the data for you where it actually quantitates histone-3 lysine-36 trimethylation, you could look at the percent positive nuclei. So essentially it looks at the ratio of brown nuclei, which is DAB staining, to the purple nuclei, which are the negative nuclei or the counter stain. You get percent positive nuclei. And for these patients, we had the uninvolved kidney, the primary, and matched metastases. What's interesting to notice with these box and whisker plots is you can see that there's a general decrease in histone-3 lysine-36 trimethylation when you go from matched uninvolved to primary tumors, and then with metastases. So this suggests that there's a clonal evolution or clonal population in metastases that will have this unique phenotype of decreased histone-3 lysine-36 trimethylation. So in sort of building to the model, you have VHL inactivation and 3P loss of heterozygosia as an early event, followed by mutations in histone-monifying enzymes. With the tumors that are with set D2 dysregulation, they have decreased histone-3 lysine-36 trimethylation, and this population seems to be enriched in metastases suggesting this is a phenotype that could be potentially targeted. So one of the things that I first wanted to look at was alternative splicing. So just a reminder, alternative splicing generates multiple isoforms from the same gene. So it's thought that 90% of our genome undergoes alternative splicing, 60% of which actually produces different protein isoforms. And this is a way of generating genetic diversity using the same genes. So you can have cassette exons or constitutive exons, depending on the alternative splicing, you can generate multiple protein isoforms. At least from prior functional splicing screens, it's thought that 10% of these alternative splicing events are required for proliferation of at least breast cancer and ovarian cancer cell lines. So why look at alternative splicing in histone modifications? So Tom Mastelli in 2010 published a science manuscript or paper where he found that there's a link between these epigenetic readers called alternative splicing. So during transcription you have RNA polymerase 2 as it opens up these different genes for RNA transcription, the histone 3 lysine 36 trimethylation is actually recognized by a Crobenton adapter protein called MRG-15 and it recruits alternative splicing factors such as polyperimidine tract binding protein. So this suggests that it might be a potential link between alterations of these histone modifications and alternative splicing. So to further test this idea, I actually wanted to go and look at patient tissue. So at Mayo Clinic, we're actively genotyping our patients as they go through the clinic. We identify the patients with a set D2 mutation. As an additional control, I look at the whole tissue and I actually stand by histone 3 lysine 36 trimethylation to confirm there's a decrease in this histone modification. So we have both the genotype as well as the epigenic phenotype. So the first thing I wanted to do is actually look at the genome-wide alterations of histone 3 lysine 36 trimethylation between a paired uninvolved kidney and a set D2 mutant renal cell carcinoma tumor. This allows you to actually compare the differences in histone modifications between genetic loci between the tumor and the uninvolved kidney. At the same time, using RNA sequencing, I can actually identify aberrant alternative splicing events. So this is somewhat alluded to by Ian Davis, but we're looking mainly at mature transcripts and we use RNA sequencing in a little bit higher depth reading in terms of 150 million reads. This allows us to actually identify low-abundance isoforms as well as complex splicing isoforms. So just an overview of chromatin immunoprecipitation sequencing, it takes advantage of these DNA protein interactions. So you can cross-link this with something like formaldehyde and actually binds the histone protein with the chromatin and using an antibiotic specific to the histone modification, you can actually immunoprecipitate the protein DNA complexes. You can reverse the cross-links and then you get these DNA fragments and if you sequence these DNA fragments, you can then align them to the human genome and identify where in the genome these histone modifications are laid down. So traditionally, histone 3 lysine-36 commethylation is associated with open chromatin, so normally our chromatin is compacted down and in the presence of gene expression, it has to be some opening or sliding of the nucleosome units so that the transcriptional machinery can access the gene promoters to promote gene expression. So this is chip PCR from patient tumors. On the y-axis is the percent input. On the x-axis I have various genetic loci that I decided to look at. As a negative control, I use IgG, which is essentially a non-specific control. And then to look at this particular histone modification, I found an antibiotic-specific for histone 3 lysine-36 trimethylation. If you look at actin, which is a gene that you would predict to be expressed in tumors as well as uninvolved kidney, you see an enrichment of histone 3 lysine-36 signal compared to the IgG control. As a negative control, I looked at an intergenic region on chromosome 12. This is an area that has no known genes, and it's an area that's thought to be chromatin closed. So the prediction is that you would not see any signal from either IgG control or histone 3 lysine-36 trimethylation. For the match uninvolved kidney, you see the same relationship where you have enrichment of histone 3 lysine-36 signal at the actin loci compared to the IgG negative control. And this is across three different tumors that we've looked at so far in-house. So in trying to identify where these histone modifications are laid down in terms of the gene features, we essentially looked at where in the genome, such as CPG islands, introns, exons, untranslated regions, downstream and upstream fragments of the gene. And if you look at everything in orange, you can see that the majority of the exons are enriched in histone 3 lysine-36 peaks. So this suggests that these histone modifications are enriched at exons. And as a control, we also looked at 7-8-6-0 and as well as CACI-1 cell lines as well. So to identify apparent alternative splicing between paired uninvolved kidney and renal cell carcinoma, I worked with Mia Champion, who designed a cufflinks algorithm to actually identify alternatively spliced events. So this is just the chromosome. This is the reference sequence. This is on the antisense strand. So the gene actually goes this way on this diagram. So at least in uninvolved kidney, it expressed the reference sequence. But what's interesting for the renal cell carcinoma, we see these additional three different isoforms. And this is a gene called absentin melanoma. So luckily it's not absent in kidney. So at least that kind of makes sense. So the next thing we want to do is actually look at the cicer algorithms. So what this algorithm does is it actually looks at chip sequencing data. And what we're trying to do is overlay the chip sequencing data with the RNA sequencing data to see if there's a correlation between where you detect differences in histone 3 lysine-36 modifications with our RNA sequencing. So again, you have the chromosome. You have the reference sequence. And the cicer peak algorithm detected a difference in the histone 3 lysine-36 peak islands in this particular region of the genome compared to the uninvolved kidney. And I just wanted to show you what our chip sequence, when you normalize the coverage, what it looks like. And that's the reason why you have these gaps. In the absence of any normalization, you would actually see some signal here. So when you overlay the two, you can have the RNA sequencing where you detected a difference in transcripts in terms of these three additional isoforms. And this overlays with some of the cicer peak algorithm that we were detecting. So this allows us to actually identify which alternative splicing events are associated with alterations in histone 3 lysine-36 trimethylation. So in some of the future experiments that we're doing is we're actually comparing alternative splicing changes between isogenic set D2 cell lines. This is important because we can actually see if this is context-dependent or not. So we've generated some set D2 knockout mouse embryonic fibroblasts. These are VHL wild type. And we have some set D2 zinc finger isogenic cell lines. And these are in the setting of VHL deficiency. So we can compare the alternative splicing changes. The hardest part of this project, and the most challenging part, is going to try to identify functional consequences of these aberrant alternative splicing events in isogenic cell lines. At least for a functional splicing screen in breast and ovarian cancer cell lines, they found that about 10% of the different isoforms or the alternative splicing events promote proliferation. So we would probably use the same strategy. The other thing is we're looking in part of the precision medicine clinic at Mayo Clinic where patients come in for whole exome sequencing, RNA sequencing, CGH. We are now optimizing chip sequencing to start on our first patient in one month and looking at chip sequencing in our frozen tissue as well in parallel. This would identify alternative splicing events in other set D2 dysregulated malignancies. So so far we've detected set D2 mutations in glioblastoma as well as a patient with breast cancer. So this allows us to take a look at other set D2 dysregulated events in a different tissue. In thinking about potential targets in set D2 dysregulated malignancies, as Jim alluded to before, it's very tough to target tumor suppressor. It's tough to target a loss of function mutation. So in thinking about some of the targets, some of the targets may include some of the readers of these particular histone modifications. So far, the three readers that have been characterized that read this histone 3-lycine-36-time methylation contain in this PWWP domain, two of which are alternative splicing factors. One is a DNA methyl transverse. The other opportunity is looking at these alternative splice isoforms that are unique to set D2 dysregulated tumors. We've identified, through alternative splicing screen, we've identified extracellular surface markers as well as tyrosine kinases. And the idea is that potentially you could target a domain or region or isoform that's specific to set D2 dysregulated tumors. And the last is alluded to to some of the presenters before me is using novel pathways identifying synthetic lethal screens of these isogenic cell lines. So the goal is to look for pathways or targets that selected the kill as set D2 deficient cell line, but not the parental cell line. So I want to thank the Kidney Cancer Association for funding this research through ASCO Young Investigator Award. My current collaborators at Mayo Clinic, my former mentors at MD Anderson, as well as my external collaborators. Thank you very much.