Chapters Transcript Video Pre-clinical Discovery of Actionable Metabolic and Epigenetic Targets in Kidney Cancer Abhishek Chakraborty, PhD By saying it's really an incredible pleasure to have Doctor Chakrabarti, uh, who's agreed to give this translational talk. Um, I, I know him very well and it's hard to think of anybody for me as a scientist who combines both a rigorous approach and creativity all together at once. And sometimes you find one, sometimes you have to find the other, but having both together is, is incredibly unusual, uh, for someone who's mechanistically rigorous, um. Doctor Chakraborty's background is that he did his PhD work with Bill Tancy at Cold Spring Harbor in uh Mick and transcriptional biology. Uh, and then he moved to Dana-Farber Cancer Institute where he worked with Bill Kalin, uh, who, as you all know, uh, is a Nobel laureate in the area of, um, VHLI biology, that's particularly relevant to kidney cancer. He spent 7 years there, uh, and afterwards we were really fortunate, uh, when I was in Cleveland to convince him to come to Cleveland, and establish his own independent laboratory. And you know, he has incredible energy and um uh over 5 years he's 56 years now actually he's developed an amazing program and it really happened very quickly. Um, and just to touch briefly on the the types of discoveries that he's made over his career, first of all, he's published in all the most visible journals, um, and these are basically mechanistic areas that touch on. Um, what we think of as sort of the traditional signaling components of kidney cancer and hypoxia, but seem to regulate not only um the traditional transcriptional responses, but also epigenetic levels of epigenetic regulation that were not previously anticipated. He's also grown his um uh work to basically um uh identify mechanisms that are relevant to areas of metabolism and transport using in vivo approaches to establish new, entirely new targets in kidney cancer and other types of cancer, and he's, uh, started to take this into discovery of, um, small molecules that I think over the coming years will end up being drugs. That we can utilize the kidney cancer, perhaps even other tumor types. Um, he's funded by a variety of different mechanisms including the DOD, the Kidney Cancer Research Foundation, American Cancer Society, and, and many others, um, and I think I'll probably just stop talking here for the sake of time and turn it over to, uh, Doctor Chakraborty. So Abhishek, thank you so much for, um, joining us to give this talk, and we, we really look forward to the latest on your work. Awesome. Uh thank you so much, Nina. You make me sound smarter than I actually am, but, but I'll try to do justice to that introduction here. Uh, I will try to make this, uh, can you folks see presentation mode now? OK. Perfect, perfect. OK, great. Uh thank you. All right, so as you heard, my, my focus really has been in preclinical target discovery in kidney cancer, uh, but at its core, uh, my scientific philosophy really centers around understanding, uh, this regulated oxygen sensing, and that's what I did in Bill Celen's lab, and I continue to develop those ideas and take them into exploitable, uh, therapeutic, uh, uh, modalities in, in, in a variety of different cancers. Um, so, uh, uh, essentially, uh, this is the template on which I think the textbooks are now operating, right? This is summarizing nearly. Couple of decades of work, I would say of what we understand as oxygen sensing. We think of three major players. We think of proy hydroxylase, uh which belongs to the Egele family. This transfers using molecular oxygen in the environment, uh, hydroxylated residue onto prolelines on its substrates. Uh, the most famous, probably proline hydroxylate substrate is a transcription factor from. called hypoxia inducible factors, the alpha subunit really is the recipient of this molecular oxygen. But then this hydroxylated residue becomes a flag on the protein for, for, uh, it to be recognized by the VHL uh dependent ubiquitin ligase complex, and, and this really uh puts a ubiquitin chain on this transcription factor targeting it for proteolysis, so. Whenever there's oxygen in the environment, HIF is being rapidly proteolyzed by the VHL complex. If there is no oxygen or no VHL, uh, then HIF becomes stable uh and can now mount its oxygen sensing response. Uh, so the reason why VHL, uh, biology or kidney or oxygen sensing becomes linked to kidney cancer is because of the genomics of, uh, clear cell RCC, which is the most common form of kidney cancer, and in this case, if you look at TCJ data, you find that either by promoter hypermethylation or we are somatic loss of function mutations, uh, inactivation of VHL is a virtually a hallmark lesion, uh, in these cancers. And as a result of that, chronic activation of HIF, uh, is also a feature of these cancers. Uh, HIF, uh, the alpha subunit can have then, uh, uh, uh, then binds to its obligate binding partner in the cell called HIP1 beta or RN, and this is what drives the transcriptional response, uh, which is, which is critical for the tumor tumorgenic programs in kidney cancer and, uh, in fact, in some other cancers as well. This also is the rationale behind. And why molecules like Balooan, uh, which have now become FDA approved in both familial as well as sporadic RCC, uh, really were justified to target if in these cancers. Now historically, if you studied, uh, this particular pathway, you would have looked at uh angiogenesis or anthropoiesis, metabolic changes uh in the glycolytic pathway or some cycle changes, uh, as a downstream targets of his. But when I joined Bill Kalin's lab, one area that was just becoming Very uh uh recognized was the relevance of his biology in chromatin function. Uh, and given my background in transcription biology, I really got interested in that aspect. And for many years now, I have been studying how chromatin, uh, modifications or histones are altered in these uh diseases, uh, especially in the context of oxygen availability and why that matters to cancer cells. So this really became a very simple hypothesis from the context of kidney cancer. We had the VHL null tumors having high levels of HIP, but among the HIF target genes were also very well known histone modifiers, particularly those belonging to the KDM histone D methylase families, and we made a very simple hypothesis. If this is true, then the histone octama around which our DNA is wrapped around in the in the nucleus to form these tight nucleosomal structures. Um, the, the, the modifications on these histone octier would be inherently different in VHL proficient versus VHL deficient tumors. More importantly, we argued that if this was occurring in the VHL deficient tumor, there probably is a selection pressure for some kind of oncogenic activation downstream of these epigenetic changes, and we directly profiled for this in human tumor samples. So in this case, tumors were recovered from Uh, Sabina Signorredi's, uh, team, uh, we took 12 representative tumors that were either clear cell RCC having VH VHL deficient HIPI, or we also took 12 each from, uh, papillary or chromophobe, uh, RCC patients. These are now VHL proficient HIPALO tumors, and we looked for histone modification changes using a mass-based, uh, uproot. And what we found really remarkably was that 11 out of the 12, these black boxes on top marking CCRCCs clustered together in a single clade, and what was driving this cluster really was histone H3K27, lysin 27, hypomethylation, or an increase in histone H3K27 acetylation. And this was really exciting for me as as a postdoc at that time, and the reasons for this were twofold really. One is that these hypermethylation events are really it's a yin-yang sort of a process. They're occurring reciprocally with acetylation events. Uh the hypermethylation and acetylation go go opposite to each other in directionality, uh, but, but really this gave us. A template on which we could start asking what might be the epigenetic vulnerabilities in kidney cancer, and we identified that activation, HIP dependent activation of these histone demethylases is really promoting the hypermethylation in these tumors and perhaps counteracting this is the activity of the the histone methyl transfers of the ECH1. Which make them actually very suscept like these tumors very susceptible to inactivation of ACH1, uh, and we published this, uh, work some time back. But then there was a flip side of the question, right? We were seeing accumulation of H3K27 acetylation. What might this be doing in the cell? Uh, and the reason why this was important is because H3K27accillation is a well recognized gene on, uh, mark in the genome. So this is typically occurring on promoters and enhancers of genes that are Very highly expressed. More importantly, if you look at the identity of this gene, Rick Young and Jay Bradner at that time had started showing that in cancers, but also in several developmental programs, accumulation of H3K27 acetylation was a marker for master regulators of developmental programs or of oncogenic signaling. And we said, OK, well, if K27 acetylation is occurring so highly in the VHL deficient tumors, what are the oncogenic low side that are marked by this uh gene on event and perhaps that's going to tell us something about uh the, the relevance of this epigenetic dysfunction in kidney cancer. So we did this very simple uh analysis. So you can actually grow VHL proficient and deficient cells without any problems uh in cell culture and we could actually use this comparison to profile for differences in K27 asset like the non-tumogenic versus turogenic cells and ask where exactly were changes occurring in H3K27 asset relation. Now, if you have not seen these types of pictures, this is just a simple. map. Every row really is one part of the genome over here. And so essentially what you're looking at is signal. The dark blue is the signal of H3K27 acetylation. So you're actually looking just at the, at this vortex and comparing it between two different conditions. The the places where there are changes occurring tell you actually the functional identity of the genes and, and allow you to generate hypothesis to pursue further. If you look at the VHL deficient versus proficient cells, as you would imagine. If target genes are turned on highly in the VHL deficient cell and in line with that K27 acetylation also occurs very highly in these gray tracts uh in the VHL deficient cells compared to the red tracks which are in the VHL proficient cells. But this was not the surprise. This was what you would expect from this experiment. What was the surprise was there were many if independent or non-canonical genes that we would not have a priority expected to be marked by H3K27 acetylation uh to also show a lot of signal over you. This was both exciting but also a, a cautious sign because anytime you see this in cell culture, you wonder what's the physiological relevance of this finding. And so what we did is we took the template that we got from cell culture and fortunately for uh Patric Times group at that time from Singapore had published a data set of H3K27 acetylation in patients. And what we did is we overlapped our data set with the K27acylation data set from tumors. Or normal adjacent from human samples. And as you can see the dark blue signal which is a K27 acitation at the same low side that we see up regulated in VHL deficient cell lines are also highly up regulated in the uh in the tumors but not as much in the normal adjacent tissue, suggesting that our cell line data really was very nicely recapitulating the effects that you're seeing in human tumors as well. So how do we then now you see there are hundreds and thousands of genes that you're going to identify by these types of analysis. Which ones do you think are functional? And to do this, my lab, we routinely rely on some kind of genetic screening strategy. So you can do two different approaches over here. You can take genes that are highly expressed in the VHL deficient cells compared. proficient ones and you can inactivate them using CRISPR and that is what would a negative selection or a necessity screen. Identify genes that are required for survival. You can do the opposite, which is the VHL proficient cells can be now tested in a positive selection or sufficiency screen. You add bad genes to the VHL. Efficient cells and see which ones can convert erogenic potential and we have done this in both different directions and I'm going to show you a very quick example of both directions, uh, to identify like that has allowed us to identify our targets. So this is brand new unpublished work. Now this is actually for the first time I'm actually presenting this in public, uh, our, uh. Example of a tuber oncogenic driver from a necessity screen over here. So we have taken here the HCK 27 acetylation data, identified genes that are marked by higher K27 acetylation but also are elevated at the transcriptional level and made a CRISPR library targeting these genes. So this is just a way for us to. Genetically shut off these genes and we can now do a very simple in vivo assay. VHL deficient cells form tumors in vivo proficient cells do not form tumors. Can we take the deficient cells, put in this CRISPR library in there, switch off the genes that are highly expressed, and identify which ones of them are necessary to form tumors. So a lot of data on this slide. The, the first part is just a schematic. We start with a population that is completely mixed in terms of our CRISPR library and then over time we select for uh either enrichment or depletion of uh uh of targets. And so essentially the SGRNAs that are depleted are essentially uh targeting potential oncogenes, those that are enriched. Targeting uh potential tumor suppressors. So and and not with the actual, uh, uh, quality control metrics that we did, but the, the assay worked really well for us. Uh, we have tested two different cell lines in vitro and both cell lines in vivo, but unfortunately one cell line does not form tumors in vivo, so our in vivo data became dependent only on one cell line to begin with. But nevertheless, there were 3 genes that stood out in our initial analysis. And there's a gene called NF1A, a gene called RIT1, and a gene called ZC3HC1. Notably, these three genes when we looked at our in vivo data. Now in blue you're looking at genes for who's like genes that are scoring as potential candidate oncogenes. SGRNAs over here are depleted. Every row over here is a different tumor, so 6 biological replicates are being compared. When it's in red, it means that the SGRNA. those genes were elevated. The three genes again common in all arms of our experiment in cell culture as well as in in vivo assays, uh, the three genes that scored as standard onto genes were, uh, NF1881 and GC38C1. I won't show you the validation data, but among them NF18 turned to be the big winner. Uh, I'm showing you the. Your final validation data for NF1A. This gene encodes a transcription factor called nuclear factor 1A. Folks often confuse this with NF1, the neurofibromatosis gene. This is not the same gene, completely different, uh, gene, totally different, uh, cellular function. It's a transcription factor, and that's going to become relevant in just a second. We also noted not only that loss of NF1A was causing these like really small tumors compared to the controls, but we also noted the color of the tumors. I think this is critical probably down the line, uh, because this is sometimes suggestive of differences in angiogenesis, uh, in, in, in the tumor. The vascularization is completely different in these tumors. Uh, but, but remarkably worked in the subcutaneous model over here, but it also worked quite nicely when we, uh, did orthoopic kidney cancer studies. So the control, uh, uh, cell lines continue to grow tumors quite efficiently. You inactivate NF1A and we saw a significant reduction in tumor burdens in these. So all in all, we were, we were convinced that NF1A was indeed a bona fide hit from our screening, uh, and we began our validation studies in, in patients, uh, patient samples. So just using the same data set that we had previously uh used uh for our validations, we also. looked specifically at the NF1A locus and found that in many of the tumors that we got from human patients, uh, the K27 acetylation marking occurred much more highly, uh, in the tumors compared to the normal adjacents at two different regions that were regulatory regions around NF1A. Using uh tumors that we received from across the street, uh, from our urology colleagues at uh Cleveland Clinic, we also noted that we could get by Western blood, uh, the, an increase in NF1A in the tumors compared to the normal adjacent. So this was again showing that there was at least some selection for higher expression. Uh, we also looked at TCGA data. This is open chromatin regions in TCGA, and we looked at the regions that are particularly, uh, specific for the kidney cancers like, uh, clear cell and capillary. There's a very clear signature that you can identify using open chromatin analysis. And if you look at what are the transcription factor motifs that are enriched over here, you find that the NF1 family motifs are one of the most highly enriched motifs in open chromatin and kidney cancer. And this again, our original K27 acetylation and, and Western blood just pan out in the TCG MRNA data as well. Uh, if you look across the board, then the kidney cancers happen to be, uh, the, the cancers were most prominently, you see, uh, increase in. NF1A uh in the tumors compared to the. So all in all, there was some functional as well as correlative evidence that NF1A should be a critical player in kidney cancer, and we began testing this pretty aggressively in a whole panel of cell lines that we had in the lab. So here you're looking at crystal violet staining of cell lines where wherever you see this dark blue, dark black signal, uh, that means there are more cells surviving over here compared to the controls, we can see that you inactivate NF1A using two different CRISPR guides and now you see a significant reduction in. Self fitness. So we were able to annotate cell lines based on their dependence on NF1A, and we found that nearly 75 to 80% of the cells that we tested were actually highly sensitive to NF1A activity. inactivate NF1 and these cells died. We also noted some odd cell lines like this SLR 23 over here where NF1A loss was tolerated quite well in these cells. And so we are right now trying to figure out what might be mechanisms by which a cell can overcome NF1A dependent. But along the along these studies we found something really interesting. Uh, so what are the cell lines that the workhorse in the kidney cancer field is a cell line called 7860, and these cell lines, when we inactivated NF1A in these cell lines, they were totally fine, did not care about losing NF1A activity. But Jot massage's group has actually made metastatic. Proficient, uh, versions of these same cell line literally by putting them into mice and then over time selecting out for for populations of cells that actually seed the lung uh and uh you take these metastatic lines, they're genetically virtually identical to the parental line, but I just acquired meta epigenetic changes that make them more metastatic. These cell lines become highly dependent on NF1A, making, making us wonder whether there's some kind of um cell state dependent selection that also occurs over here making NF1 a really potent dependency in metastatic cells. So all of this got us very uh uh keen on, on studying uh what might be the mechanisms by which NF1A is acting in these tumor cells. Remember I told you NF1A nuclear factor 1A encode the transcription factor. So the canonical function would be that perhaps it's regulating a transcriptional program, but I mean just. Because that is the canonical function. Biology always surprises you have to test that idea, uh, formally. So the way to test the idea is in my lab, we do a lot of what's called rescue experiments. We do a back of wild type or mutants of a protein and ask what are the critical domains of the critical residues that are responsible for oncogenic function. The way this works is you inactivate your gene of interest in this case NF1A. Cells are going to die. But now if you introduce a SGRNA or a CRISPR resistant version of that protein. If it is functional, the cells will survive. If it is nonfunctional, the cells will die. So in this case, we put in a functional version of NF1A, the wild type NF1A, the cells can now start to survive in the presence of the CRISPR guide. You put in a transcriptionally dead mutant of NF1A, the cells are not able to survive. Suggesting that the transcription function is indeed critical for its oncogenic role. And, and for good measure, we have done this same experiment in vivo as well. You can make tumors, uh, have or not have NF1A and then add back the wild type of the mutant forms, the wild type form rescues tumorgenesis, whereas the mutant form does not do such a good job. So if it is the transcriptional function, then perhaps we should ask what is the transcriptional program that is regulated by NF1A, uh, and we have done that using a variety of, uh, uh, uh, different, uh, SGRNA that we have in the lab. Uh, the, the transscript signature was remarkable. Uh, we found that there were many, uh, uh, target genes that were in the metabolic processes, particularly fatty acid metabolic. that were disregulated by NF1A loss and actually I did not know this, but I learned this when, when we first did this analysis. The fatty acids are actually precursors to molecules called leucotrienes. These are, uh, chemotactins in the, in the tumor, uh, as well as in just normal physiology, and we found that these leucotrine metabolism processes were also altered in NF1A deficient, uh, cells. So just to summarize, the text on the on the heat map is a little bit difficult to read, but we found metabolic targets. We found cell state regulators, we found also quite interestingly, many of the HIF target genes were down regulated when NF1A is, is, is, uh, switched off in these cells. We can validate some of these, uh, uh, effects. For example, we can actually stain these, so these clear cells, uh, or I mean the kidney cancer folks would know these are literally bags of lipid droplets, right? That's why they're called clear cells because they're like copious amounts of lipid deposits in these cells. We find that you inactivate NF1A and you see a significant reduction in the lipid content of these cells, the neutral lipid composition completely changes. Uh, but more importantly, we also found that uh the amount of oxidative damage that we can score using these lipid peroxides, uh, is much higher in the NF1A deficient cells. So we're seeing both an impact on the neutral lipid uh content, but also perhaps more oxidative stress in these cells as well. We've also done some validation using uh publicly available single cell data. Now, if you do this type of analysis, what you find is that uh these are the tumor cell clusters over here. There's a cycling tumor cell cluster as well as the two different tumor programs, uh, that have been shown. This is this is kidney cancer, advanced kidney cancer, uh, this data set, and you find that NF1A virtually aligns with the expression patterns of, uh, um, HIPA. HI target genes in many ways limited to expression and cycling tumors as well as in the tumor activated tumor mogenic programs. And if we do a correlation, which we did with our friends at Memorial Sloan Kettering at Resnick and Ari Hakimi's teams, we looked at what are the the clinical correlates of high NF1A expression in kidney cancers. We find that Higher levels correlate very highly with uh angiogenic scores, but also have a negative correlation. They're essentially acting almost as immunosuppressive, uh, for, for many of the immune associated mark. Remember I had shown you the tumor samples themselves, how the colors look different, and we think that what we are seeing in patients also is aligning with this difference in uh angiogenesis and vascularization uh in these tumors. Targeting a transcription factor is really, really challenging, and we are trying to do our best to figure out what NF1A binds to in the cell and what is the transcriptional complex looking like to identify drugable vulnerabilities. But in the meantime, while we generate that data, uh, we also started doing some simpler chemo sensitivity or, or, um, uh, uh, synergistic sort of pharmacological assets. year we took either wild type cells or cells that were inactivated for NF1A and against a library of about 5000 FDA annotated pharmacological inhibitors asked which of these would, would preferentially kill NF1A deficient cells. And in this list we were able to identify several kinase inhibitors like the BTK inhibitor. In getting Nema has been interested in also for some time, uh, but also some molecules that are regulating bromodomain proteins as well as, uh, the 11 ACH inhibitor. So, uh, we're seeing this, this, this, these repeated signs that the transcriptional hallmark or the epigenetic hallmarks associated with NF1A are, are really critical in terms of their, uh, oncogenic function perhaps. So based on all of this, this is where our working model really is. Uh, we find that uh VHL deficient cells, uh, uh, we are getting the activation of HIP, uh, and that really has become now the prime target for, for molecules like the ludophan. I did not show you today, but, but the activation of NF1A in the VHL deficient cells is not be a HI. It's via some other transcription factor, uh, which we have not identified as yet. We're looking at it, but effectively we find that NF1A is a completely HIF independent um by which VHL can operate a different oncogenic program. But if anything, We find that by our RNAC data indicates that the loss of NF1A in fact can impact HIF dependent transcriptional programs and drive uh kidney cancer as well. So, uh, all in all, we think that there's a metabolic function over here. There's an immune regulatory function over here, uh, and, uh, both cell intrinsic and cell extrinsic pathways are together driving NF1A oncogenic program, um. So what I'm going to do perhaps is in the interest, given that we have to stop a little bit early, this is a good break point. If folks have any questions about this section, maybe I can stop for a couple of minutes and take any questions for this part because we're going to switch gears entirely for part two. Maybe just one quick question just in terms of um this uh other NF1A and uh it's established role in other other cancer types. Is there anything there? So, so there are only two other lineages in which a cancer cancer role has, so NF1 has been, if you, I mean, I don't know, I've not done this recently when when we first found it on PubMed in cancer, there were 30 papers, like 30, 30. And sort of virtually unstudied, right? And the only other cancers in which there was a role that was described, there was one paper in ovarian cancer and, and a couple of papers in GBM. And in both instances there was this association with stemness and metastatic output that had been previously reported. But no, none of the other cancers. The big link has been with lipid metabolism and glucose, insulin signaling. OK. I'm not hearing anything else. I'm going to continue to the next part. All right. So in this part now, we're going to do the opposite. Remember, we, I said there are two ways by which we can screen this. We can look for the necessity screens and identify hits that, that we get from this approach, but we can do the opposite, the positive selections, uh, uh, sufficiency screen. Uh, and the template over here is exactly the opposite. We're now looking at what genes when added back to VHL proficient cells, uh, can form tumors, and the way we track this is we have a, uh, 24 nucleide barcode at at the three prime end of all of these genes that can allow us to sequence in pretty high depth and, and look for, um, the, the enrichment of potential on candidate onto genes, right? So these are. The actual data looks like, so VHL deficient cells, these, these, these, these cells form tumors at 100% penetrate, 8 out of 8 injections over here. VHL proficient cells don't form tumors. If you put in a library of potential candidate oncogenes, we can start getting some tumors occurring not at 100% penetrance, but nevertheless, there are some tumors that are scoring as sufficient. So when we sequenced these tumors, we were able to find that the negative controls, which you would not expect to score as drivers in this are all in blue, which means they are depleted when the cells are injected in mice to form tumors. On the other hand, there are potential candidate oncogenes that are being scored over here which are in red, enriched in tumors and therefore potentially uh driving an oncogenic. Extremely oncogenic program. The one candidate among these that we really started focusing on was a protein called SLC1A1 or solute carrier 1A1 is an aspartate glutamate transporter. This was the only one that we found scored with 100% penetrance and in multiple cell lines and therefore became our primary target of interest. This confuses many folks, uh, uh, especially in on the clinical side, because the textbooks tell you that aspirate and glutamate are nonessential amino acids. Why the hell should a transporter that picks up a nonential amino acid become an oncogene? And really that's because of the magic of tumorogenesis. These, these cells get transformed and they, their, their appetite for even these quote unquote non-essential metabolites increases manyfold and this is a very common feature in many cancers that transporters of even these non-essential amino acids become very critical for survival. And SLC 1A1 we think is one of these types of uh transporters, uh, and therefore we have been looking at the, the mechanisms by which SLC1A1 drives tumorgenesis in kidney cancer. So I'll skip this part very similar to what we did for NF1A. Uh, we can find that in human cancers also, uh, SLC1A1 expression is marked by higher HTK27 acetylation. Uh, it is VHL dependent and again exactly like NF1A, we found that this is not occurring via his but some other transcription factor. So, but why do we think SLC 1A1 is also important is because of these types of functional studies. Here we can inactivate SLC 1A1 again using multiple different CRISPR guides. Every time that we are able to successfully eliminate this protein from the cell, like three different guides over here, that's when we start seeing fitness defects in the cells as well. So SLC 181 is critical for the survival of the cells. SLC 1A1 takes up aspartate and glutamate, so the most common expectation there would be that SLC 1A1 loss would drive complete metabolic rewiring downstream of aspirate and glutamate. We have tested this using metabolomics analysis. Um, this is a busy slide where I'm going to summarize it with these graphs at the bottom. The key perturbations that we find in cells lacking SLC 1A1. are a significant depletion of aspirate and glutamate derivatives. TCA intermediate. So if you remember your biochemistry techbooks aspirate and glutamate feed into the TCA cycle, and we find that there is a significant reduction of those. Nucleotide intermediates are also decreased in these cells, and why that happens it will become clear in just a second, uh, but all in all, we were able to score a very clear metabolic reprogramming event. So this is why we think that nucleotide perturbation also makes sense. cytozolic aspartate and especially once it goes into cells, um, uh, it, it gets not only it goes into the TCA cycle, but really the 4 carbons of aspartate are the backbone, uh, carbon skeleton on which nucleotides are synthesized. So the fact that when you inactivate a self1A1, you deplete aspartate from the cell, you're seeing a reduction in the nucleotide intermediates also makes sense from this result, this perspective. So all the data that I showed you so far, it's all cell line-based data except for the validation of K27acci is in human tumors. So we went back and again used RENE's, uh, metabolomic data sets this time and asked whether the SLC1A1 expression would also show uh uh uh association with metabolic reprogramming human tumors and this was quite remarkable. For me, on the same pathways that we are finding in cell lines as altered by SLC 1A1 activity, if you stratify tumors that SLC1A1 high versus low, we find that the SLC1A1 high tumors also have uh elevated levels of many of these same metabolic pathways, aspirlutamate pathway, pyramidine, uh, nucleotide bi synthesis, etc. So how do we know that aspirate and glutamate are the reason why these cells are dying? So this is what the, what, what the picture looks like. Are you inactivate cells you want to want, you can see the cells do not like to like that too much. They don't grow. But all you need to do is provide cell permeable versions of aspirate and glutamate in the media, and now the cells are OK. They will now survive in the absence of SLC 1A1. Uh, and this does not happen with other amino acids. I'm just showing you alanine as a control and suggesting that the canonical function of SLC1A1 is really what the critical oncogenic function. And this becomes relevant in the context of other therapeutic modalities, right? Because, uh, there, there was at this time, this, this, this unfortunately has failed entirely now and, and has become, uh, irrelevant, but at least from a metabolic standpoint, the kidney cancers are highly dependent on this enzyme called glutaminase. Uh, you convert the glutamine to glutamate and then you This reductive carboxylation process to make lipids in these tumors, and there was a significant clinical trial by Calithera to to target glutaminase as therapeutic in kidney cancer. But based on our study, we predicted that SLC 1A1 expression and function would impact the response to CBH 39. Higher levels of SLC 1A1 would make the cells resistant because they don't need glutaminase anymore. Uh, down regulation of SLC 1A1 would increase dependence on glutaminase, and we have been able to show this in both directions. You can add back SLC1A1, while type SLC1A1 make the cells more resistant to glutaminase, uh, mediated cell death. Conversely, inactivation of SLC 1A1 makes them more sensitive to, uh, to sell them. And so based on this, we have then asked from, from a physiological standpoint, uh, can SLC1A1 targeting be even relevant? Where exactly is this protein being expressed? What context should we be thinking about its function? And so with Sabina Signorretti's help, we had analyzed a large uh TMA panel of kidney cancers, various stages, various metastatic associations, uh, and we found that higher levels were associated with advanced stages. I'm just showing you one small, uh, part of that entire data set, uh. Compared to the M0, the M1 tumors had a much higher level of SLC1M1 expression, suggesting that there might be some functional selection in metastatic cells. Now this is purely correlative, does not show functionality. So we did use the metastas metastasis proficient kidney cancer line then we. Inactivated SLC 1A1 in those cell lines and we actually did the functional assay in mice and we can, we were able to see that compared to the control lines which were able to metastasize to the lung quite efficiently, you inactivate SLC 1A1 and the metastatic seeding is, uh, is decreased quite significantly. While we were doing all of these studies, there were these two papers that caught our attention, and this was really interesting for us. One came from uh Joshua Rabinowitz's lab, the other from Ralph Berard Dennis's group. Both of these papers, they essentially said that our current understanding is that in tumors they undergo what's called this quote unquote Warburg effect. They up regulate glycolytic cycling and they down regulate oxidative phosphoration or PCS cycling. And that is true. As a feature of localized tumors, but really when these tumors start to metastasize, they transition out of this local niche, uh, that's when they actually upregulate the TCA cycle and mitochondrial metabols. And the reason why I'm bringing this up is because if you remember, SLC 101 loss is depleting TCS cycle intermediates, impacting mitochondrial function in this way. And we think that perhaps the fact that SLC 101 loss regulates TCA cycle and it's also impacting metastasis are functionally linked to each other based on all of these studies that have started to just very recently come out. And that is a, that is an area of, of future interest for us. So in the last few minutes, I'm gonna show you what we have been doing to try to target SLC 1A1 directly. Uh, so the first question we asked was whether this was going to even be feasible. So if you look at normal tissue, you find that the most prominent expression of SLC1N1 occurs in the kidney and in the central nervous system. So the neuroscientists have been studying SLC 1A1. They call, they, they know it as E3. And they have been studying this protein for a long time because SLC 1A1 uptakes glutamate from post-synaptic space. And if you remember your neurobiology textbook lutamate is a very potent neurotransmitter and so this is a means of neuro regulating neurotransmission and the neuroscientists have actually identified some tool compounds that they, they, they claimed that were able to target SLC 1A1 and could be utilized in these neuropathologies. I am by nature a very skeptical scientist and so when provided with random chemical structures, I almost always question whether they are true hits or whether this is just off target. And for this particular molecule, it failed what what I would call the sniff test of the drug test for me because this molecule does not look anything like. Aspirate of glutamate, the canonical substrates of uh SLC 1A1. There is no reason why this molecule should also specifically target SLC1A1 and not all the other proteins in this family, as it was originally claimed in this paper. And so, we, we proceeded with characterizing this molecule, but with a lot of caution. So Pune Kuaki, postdoctoral fellow in my lab, she led a lot of the characterization efforts and basically Pune was able to validate that. Indeed, the, the molecule binds the SLC1A1. She did what we call these thermal stability assays showed that the thermal stability was increased. She then did some cytotoxicity assays. We knew which cells were or were not dependent on SLC 1A1, and we found that this molecule compound 3E also preferentially killed the SLC-1A1 sensitive cell lines, but not the insensitive ones. When we did metabolomic studies, we found that compound 3 actually blocks as. glutamate metabolism, which was also good to see. And then finally, the same experiment that I showed you with the genetic experiment, uh, you add back self-permeable aspirate or glutamate and now you can overcome the the cytotoxicity associated with compound 3. So all in all, these validation studies just allowed us to conclude that uh compound 3E might have some role to play with SLC 101 uh and, and perhaps it was not just uh uh off target uh cytotoxic entity. So to start testing what exactly SLC1A1 and compound 3 are uh are doing and, and what their function, the, the, the mechanism of interaction is, we collaborated with Olga Bootker's group at Cornell and we solved the cryo EM structure of SLC 1A1 either by itself or in complex with compound 3E. And what we found was quite interesting. What we found is that this protein can exist in the sodium bound state, so it requires, requires this ion for, for transport. These different colored groups are really the transporter exists as a primer. So these are literally like 3 shoots of an elevator. They're all going up and down, uh, and using each other to kind of leverage movement across the membrane. And what was really remarkable for us to see was that compound 3E almost always was seen bound to the Apo form. This form does not have sodium, does not have the substrate. It was never found in association with sodium, suggesting that the binding with sodium and the compound are mutually exclusive events. The compound is probably kicking off sodium from this transporter, and that's probably one way by which it is blocking the transport process. The other thing that the molecule was doing was if you look at this pocket, this molecule actually wedges itself between two domains and essentially it sticks the elevator movement and does not let this to move in a frictionless manner. And that's what's also impacting the transport mechanism of this, this transporter. All in all, we were, we were quite excited to see that this molecule was binding SLC1A1 and working with a completely novel mechanism of action. And so we looked at what exact residues in SLC 181 were engaging with this compound, and we identified the key residues that were making direct contacts and based on a lot of functional studies, I think I'm running out of time here, uh, to, to talk you through all of this, but we have been able to identify the key residues in SLC-101 that are making these contacts and that are also driving the selectivity of this molecule for SLC 11. So, uh, quick story, punchline is that we have taken the original molecule, done, done a lot of medicinal chemistry and the original molecule, and now come out with two new molecules. We have called them now PBJ 192 in honor of the three folks, Pune, Biao, and Jessie, who actually drove this work. And fortunately for us, PBJ. 1 and 2 are not only more potent than the original parent molecule compound 3E in their cytotoxic function. Uh, they're impacting the expected metabolic processes, but they're also doing that without compromising on the selectivity for a cell. So, uh, all of this is is going in the right direction. So I'll just leave my, my summary slide and the acknowledgements, uh, uh, while everybody starts to leave the room, I guess. Uh, but essentially, uh, in, in the second part of the talk, what we've been able to see is that epigenetic dysfunction, uh, is driving SLC 1A1 expression and, and downstream function, uh, in the kidney cancers, and this is giving us several, uh, nodes, uh, of the therapeutic vulnerability, uh, in these tumors. So I stopped there. A lot of the work on SLC 181 was driven by Treg Grab and Puneujai postdocs in my lab, and then former members in the lab Carly, Jason and Shannon drove the NF1A project. Uh, this work would not have been possible without help from a lot of folks, uh, both at Cleveland Clinic, but also at various other. Institutes I'm not gonna read them all, uh, in the interest of time, but, but, uh, but really, uh, this would not have been possible without, without this collaborative effort with the whole team. And then finally, of course, uh, uh, the funding that we, that allows us to do all of this exciting work, uh, and, uh, I, I'll stop there and and take any other questions that folks might have. Sorry, I had to rush to the 2nd second part a little bit, but I just want to make sure folks have. No, thank you, Abhishek as usual, your talks leave us both breathless and hypoxic. Um, really just incredible signs, um. I, maybe I'll give a chance for anybody else who wants to ask the question because I, I have several, but I know our time is limited. I, I, I don't have anything lined up, so I'm happy to stay and take questions. Abhi Sheikh, my name is Sana. I had a quick question and it may be very naive just because I'm not a scientist, I'm more of a clinical person, but you talked about, you know, how you did a forward and a backward kind of approach and through both of those approaches you kind of ended up finding. Two different targets, and I guess like when you say you kind of give you have an option to go forward in a backward approach, I guess, you know, what do you think about the fact that that didn't come showing the same target? Does that does that limit any particular targets, you know, suitability? So from a genetic standpoint, the sufficiency screen is always a much higher bar to reach than the necessity screen. It's much easier to, so the way we try to explain this is, you know, take a standard TV and start to take parts of the, or, or throw your shoe at the TV. There are more ways that you can damage your TV than to turn it into an ultra high definition unit, right? So, so it's literally the same. You do a complex system, many ways to kill the cell, not that many ways to make the cell grow better and form tumors. So if you look at the necessity data, if you actually look down the list, you will actually find SLC 1A1 also in that list, in that data set. So it's there, it was just not one of the most potent hits in that screen. Um, uh, the, uh, but we also use a cell line that was relatively insensitive for SLC1A1 and UMRC2 cells. Uh, these are, those are the more insensitive cell lines. So that's also another factor that played over there. Uh, but, uh, the, the opposite for NF1A not showing up in the sufficiency screen, um, again, In my mind, I think probably the odds of it getting there might have been more challenging. In terms of cleanliness, the sufficiency screens tend to score fewer false positives than the necessity screens do. So the CRISPR off is almost always requires a lot more validation, years of work compared to the on assays. But yeah, I, you're, you're right. In, in principle, you should be seeing an overlap in both directions and we, we do see that, uh, to a great extent, but the sufficiency screens, I will tell you, are, are much higher standards to reach from a genetic stand. Thank you. Seems like from a, from a time standpoint, I think it sounds like we're probably gonna have to stop, stop here and turn it over to Laura and the AI, but Abhishek, I want to thank you again for, for sharing your very exciting work with us and it's um this is just, you know, incredible development. So such a pleasure to listen to your work as always. I learned something new or many new things every time I listen to your talk. So thank you again for joining us. Thank you so much. Thank you, everybody. Thank you. Bye. Published August 14, 2025 Created by
