Scientist Stories: Stuart Schreiber, Using Chemical Biology to Overcome Cancer Resistance

Thank you very much. And I would like to introduce today's invited speaker, Professor Stuart Schreiber from Harvard University in the United States. And I will make a short introduction of Professor Schreiber. Stuart Schreiber started his PhD studies at Harvard University in 1977. With Robert Burns Woodward, and after Woodward's death, he continued his studies under the supervision of Yoshito Kishi. In 1980, he obtained his PhD, and in the same year, he joined the faculty at Yale University as an assistant professor. In 1988, he moved to Harvard University as a professor. Schreiber started his research work in organic synthesis. He made important accomplishments in total synthesis of a large number of natural products, complex natural products many times, including— I mentioned just a couple of them— epoxidactyamine and the immunopressant FK506. And the latter one, which is also called Fujimycin made, I think, Schreiber very famous because I read about that and many people knew about this 23-membered macrolactone. Now, his work now moved more and more into chemical biology and in 1996 he made an important breakthrough. That year Schreiber and coworkers discovered histone deacetylase by the use of small molecules such as trypoxin. He was the first one actually to isolate a histone deacetylase. Schreiber's discovery that year together with David Allis' discovery the same year of the histone acetylase made the two, Allis and Schreiber, quite famous and it had an enormous impact on the research area after that publication. Actually, thousands of papers have been published in the field and a large number of histone acetylases and histone deacetylases have been identified since then. Now pharmaceuticals that are involved in these processes are among other things used for treatment of cancer. Major areas of research in Freiberg's lab include next-generation synthesis, cancer therapeutics that overcome resistance— I think that will be some, some part of today's lecture— microbial therapeutics with novel mechanisms of action, and finally therapeutics having novel mechanisms of action for the treatment of genetic prion disease and Alzheimer's disease. Schreiber's development of diversity-oriented synthesis has led to the discovery of many promising agents, including novel mechanism of action antimalarial agent being developed in collaboration with the pharmaceutical company AISA. Professor Schreiber has obtained many awards. It was a long list, so I just select a few here. The Paul Carrer Gold Medal in 1994 from the University of Zurich, Tetrahedron Prize for Creativity in Organic Chemistry in 1997, and then recently, 3 really prestigious awards: the Arthur C. Cope Award in 2014, Nagoya Gold Medal in 2015, and then finally the Wolf Prize in 2016. And the Wolf Prize actually awarded him for the discovery he made in 1996 about histone deacetylases. Today he will give a lecture with the title "Chemical Biology Approach to Understanding and Overcoming Resistance of Cancers." to treatments. Please do it.

Jan, thank you very much. Good evening, everyone. It's a great pleasure to be here today, and I'm very honored to be asked to speak before the Academy this evening. So let me tell you what I'd like to share with you today. I'm going to talk to you about cancer, and I'm going to talk to you about something we all know all too well about cancer, that many patients with cancer, even if they respond to treatments, become resistant to those therapeutics. So that's the topic. I want to present my work in 3 parts. I'm going to tell you about the approach that we've undertaken. I'll call it a chemical biology approach. I want to explain to you what I mean by those words. I want to share with you a discovery of a therapy-resistant cell state of cancers, which I think is very common across many different cancers, a pan-cancer effect. But also I'm going to share with you a vulnerability of this resistant state that we hope to exploit therapeutically. And then in the final section, I want to tell you about the mechanism that underlies this cell state change, because I think the one thing we've learned about translating basic discoveries in science to impact on human health is that we need to know the molecular biochemical mechanism that underlies the phenomenon that we study in order to be able to make that translation. So I think everyone here knows that the last several decades have seen amazing advances in the treatment of cancers, from surgery to radiation to chemotherapies and targeted therapies and immunotherapies. But even in the case of cancers that respond to these therapies, we know all too well that they are frequently able to demonstrate their powerful ability, these cancers, to ultimately resist these attacks on their vulnerabilities. And so resistance arises. So there's been a great deal of interest in studying the resistance mechanisms. Studies have primarily viewed resistance through the lens of genetics. Since mutations give rise to cancers, it's not difficult to imagine further mutations being able to block the actions, prevent the actions of the drugs. And in general, this approach has therefore studied resistance on a drug-by-drug basis. But increasingly, resistance is being viewed through the lens of cellular plasticity, this intrinsic ability that cancer cells and many cells have to change their cell state without any alterations in their genome. So I'm going to focus on this latter, this latter ability of cancers and the role in resistance. Now the project started when my lab members noticed that in each of these instances, each of these treatments— the treatments kill cancer cells by inducing a form of death called apoptosis. So we wondered whether it's possible that cancer cells could adopt a cell state, a stable cell state, without changing their genome, that prevented this form of death, apoptosis. If so, you could imagine a common mechanism for resisting therapeutics. So my presentation today is going to mix about half— we'll tell you about a couple of papers that we published last year and about half that have not yet been published. And I'm going to try to make the argument today and make the case for the existence of indeed a common therapy-resistant cell state that underlies the resistance of many cancers, if not all cancers. Now, I want to tell you about a hypothesis that I'm going to end the presentation with. I'm going to begin with the end. I'm going to try to make a case so that you're not confused about the issue of genetics versus cell plasticity to say they're both important, but I think there's a possibility of a rethinking of the role of both, including genetics. So what I'm going to present is evidence that the first thing that happens when cancers are given any of these therapies is they first adopt, without changing their genome, a stable cell state that's this common one that I've described that allows the cells to survive. This non-genetic mechanism allows the cells to survive. Now, the reason that people have focused on the genetic approach is that we find mutations in resistant cancers that prevent, for example, the drug from binding its target. So it seemed pretty clear that it had to be a genetic mechanism. I think these genetic alterations are very real, but they are optimizing the fitness of this resistant cell state. Because one of the things we found about the cell state I'm going to describe is that it's a slow-growing one. So the cancers first adopt— they first solve the resistance problem. Then genetics takes place not to confer resistance. But to allow those tumors now to grow rapidly. So that's, that's how I'm gonna end today. Now, how do we get there? I told you I'm gonna talk about a chemical biology approach. So this word refers to small molecules or chemicals that are used as probes or tool compounds. It's very analogous to dissecting biology with genetic approaches where you would make alterations in the genome, but here we use small molecules or chemicals to perturb proteins and, and make inferences to learn about biology. But we— this doesn't come out of a vacuum. This comes from a century of major advances by many scientists who have advanced this field of small molecule science. I have been showing this slide for two decades now Some of my heroes that showed us the role of small molecules in human physiology, using them as tools, as small molecule probes, and showing us how to take these insights into small molecule medicines. But there's an irony here because the work of three of these gentlemen, three of these scientists, probably well known to many of you in this audience, who in the 1960s and 1970s illustrated the role of polyunsaturated lipids, arachidonic acid and derived lipids, especially the prostaglandins, it turns out is critical to the discovery I'm going to share with you. Completely unbeknownst to us, or a complete surprise, we truly are building on standing on the shoulders of giants in the story I'm going to tell you about today. I'm going to tell you about polyunsaturated lipids and how they surprisingly play, I think, a key role in the resistance of cancers to modern therapies. So a chemical biology approach, for example, would involve measuring sensitivity of cells to compounds and then looking for patterns of sensitivity to uncover biology. So what we did was we started with 500 compounds that I'm gonna call an informer set. I'll start by telling you how we selected these compounds, but briefly, we picked compounds that have highly selective interactions so we could make inferences about their targets, and that overall, in an unbiased way, connected to as many nodes in cell circuitry as possible. We then used hundreds of cancer cell lines that had been thoroughly characterized genomically, and we used data that are all publicly available. The key step here is to measure the sensitivity of each compound to each cell line in dose, quantitatively, 16 concentrations, and measure a dose-response curve of each compound to each cell line. Then we measure the area under the curve as a metric for high sensitivity if it's a strong dose response of this sort, or if there's very little response, then it would be a large area under the curve. But before I go into how we uncover biology from that, Because the informer set is so critical to what I'm going to describe, I want to give you a sense of how the chemists in my lab selected these compounds. A few words about the members of this informer set. So, for example, we want compounds that are highly selective in their interaction so that if we see a pattern, correlate a pattern of sensitivity with a compound and a feature of a cell, we can infer something about its target. So we included this compound, rapamycin. Rapamycin, we now know, binds two proteins simultaneously, an FKBP12 protein and an mTOR protein. Eric Brown, who was a graduate student in my lab who carried out some of these studies, also discovered a mutation in mTOR, what's called a dominant drug-resistant mutation. Where simply changing serine to threonine prevented rapamycin from binding mTOR. And when that allele is expressed in cells, it eliminates all the measurable effects of rapamycin. This is a proof of the specificity of rapamycin for mTOR. Relatedly, we included these two compounds, cyclosporine and, as Professor Bockvoll mentioned, FK506, two compounds that even I hope the non-chemists in the room can see structures are very different from one another. Yet we now know that these compounds interact with proteins in cells like the previous case, bringing multiple proteins together, but in the end their effect is on a protein phosphatase called calcineurin. So if we see a pattern of sensitivity to cyclosporine and to FK506, we're even more confident that this is relating to their targets, calcineurin. This ability of small molecules to bind multiple proteins simultaneously, we, we say to change the interactomes of proteins, was first seen with these three naturally occurring compounds. And I confess, when I saw that, I thought, this is just astonishing what a billion years of natural selection can do to bring together proteins. Chemists will never be able to achieve this. And in fact, I was dead wrong. It turns out this probably is a very common phenomenon. And we now know that really simple synthetic compounds have exactly this effect. So we included thalidomide as an example. My lab had the opportunity to participate in understanding how thalidomide functions. And lo and behold, we now know it binds two proteins simultaneously. Thereby brings a substrate, which is a transcription factor, together with an enzymatic activity that causes the destruction of the transcription factor. So by compounds changing interactomes, there can be highly selective and profound effects on cells. We included a set of compounds that were electrophiles. Electrons like to attract these because they're a little bit controversial in medicine. Some people say they're very good because really effective drugs like ibrutinib, electrophiles, and some people would argue because they're chemically reactive, they may not be selective. It turns out we were lucky in doing this because this compound called RSL-3 is an electrophile, turned out to be key to the insight that I'll share. As Professor Bockvall mentioned, we, um, in— the early days of studying chromatin found that small molecules bind chromatin-targeting small enzymes can change gene expression inside of cells. And today, the chemists and chemical biologists and medicinal chemists have created many highly selective chromatin-targeting compounds that change gene expression in very specific ways. We synthesize these compounds, included them in our informer set. Likewise, today we know that kinases play critical roles and phosphatases and the proteins that bind the phosphorylation, phosphate trans— post-translational modification. So we synthesized these compounds having a range of activities involving protein phosphorylation inside of cells. The last group of compounds also turned out to be critically important. But these are compounds that have novel mechanisms of action that I don't think would have been possible to discover 10 years ago. And the reason is they build upon advances in synthetic organic chemistry. So just like the previous slide, I want to tell you that we stand on the shoulders of giants and we're able today to utilize advances in chemistry, olfamethathesis, asymmetric catalysis, organocatalysis, and as you'll see, especially transition metal catalyzed reactions. To make candidate probes that can do things that traditional probes cannot, we can include in our compounds chemical features, for example, that one might find in naturally occurring substances. Products of natural selection, but do it synthetically in a laboratory. It turns out that the transition metal catalyzed reactions have been especially powerful. And I call out my friend, Professor Bachval, who's been a leader in palladium catalyzed chemistry. And without those methods, many of the compounds I'm going to show you next simply would not be achievable. So maybe the chemists will see right away just the chemical features of these compounds come about by this new chemistry are not like traditional medicines or probes. Each of these compounds was included in the informer set. They all do things that are very different from traditional probes, and I'll say nothing more about them other than that this comes from a review that was just accepted by Nature Reviews in Drug Discovery, which is going to relate these advances in synthetic chemistry to these novel actions. That is the informer set. Well, looking at sensitivity data from the complete matrix of each of these cell lines, each of these compounds, we use computational methods. We correlate a pattern of sensitivity first with binary features, the presence or absence of a mutation. Then we correlated with continuous measurements like low to high levels of a protein, all of the proteins of a transcript of a metabolite, and we look for correlations in either case. You'll see some box and whisker plots, just a few I'll show you, where the dots represent members of the informer set that are highly correlated or highly anti-correlated. This has turned out to be a powerful approach to uncover biology used now by many labs And the reason that many labs can do this has to do with the work of your colleague, who I'm happy to say is here, Brenton Seashore-Ludlow. She's the first author on a paper that released the Cancer Therapeutics Response Portal. All of the sensitivity data, all of the analytic methods are freely available to anyone in the world. And that's a big shout out to Brenton. For making that possible. So you can now go— anyone in the world can go and make hypotheses about compounds with specific mechanism of action, features of cells that seem to respond to them. But I started by saying we're going to look at cell states. And cell states are not something easily measured by a protein or a metabolite. So how do we do that? What I'm going to tell you about takes this Cancer Therapeutics Response Portal to the next stage. We extended the approach to cell states, and this is how we did this. We did this specifically in the case of resistance with this hypothesis that there may be a resistant cell state. So we went to the literature, wanted to be unbiased, and we found 3 reports of 3 different patients that had been responding to therapy, completely different kinds of therapy, and then became resistant. And gene expression studies were performed with these patients to find the changes in the transcriptome. We simply looked at those data and looked at the most up- and down-regulated transcripts and considered that a signature, an unbiased signature for therapy resistance. So then we took our hundreds of cell lines and we rank ordered them from low therapy resistance to high therapy resistance. Now, as those words imply, it's not surprising we found lots of compounds that were effective in the low resistance state and not effective in the high resistance state, but we're actually looking for compounds that do the opposite, that are a lot of cell killing in the otherwise highly resistant state but not so much in the low resistance state. It was not at all clear we would find any such compounds. Now, we did find compounds that correlate such that they're less effective in the therapy-resistant state, and it was interesting to us— these are famous compounds, like basically always hitting mitogenic signaling pathways, EGFR, MEK. These compounds are known to induce apoptotic death. Exactly as I suggested before. Well, it wasn't clear that we would find compounds that are more effective in therapy-resistant cell state, but actually we found about 10 of them. I'll start with the first 3. The first 3 are these compounds. This is that electrophilic compound that I mentioned, RSL-3. But there are 2 other compounds we had included at the beginning without knowing the mechanism of action of these compounds, but we had discovered them in my lab, the 2 ML compounds. And we knew they had very strange, very unusual effects, so we thought it would be useful to include them. And we've now determined the mechanism of action of all three compounds, and we find that they are inhibitors of a lipid hydroperoxidase, an enzyme that converts a hydroperoxide from a lipid into the alcohol. It's called GPX4. But most importantly, and what really opened our eyes was that each of these compounds induced cell death, but not by apoptosis. They induced a different kind of cell death called ferroptosis. This is an iron-dependent form of cell death that results when cells build up high levels of lipid hydroperoxides. We could correlate the pattern of sensitivity of these compounds to patterns of sensitivity to knocking down genes. The best correlate was GPX4 again, which made us even more confident that that's the relevant protein. These were cell lines. We started looking at other tissue types and gaining more evidence. And so we worked with Yu Chen at Memorial Sloan-Kettering, whose lab had been studying benign prostate and prostate cancers derived from 4 different patients by converting these into organoids, prostate organoids. These first 3 patients have what are called androgen-dependent prostate cancer. That's the normal form of prostate cancer, but it's known that prostate cancer can become resistant to androgen antagonists, no longer even express the androgen receptor, and become highly resistant and highly metastatic. And this is the form of prostate cancer that most often kills patients. It's called the neuroendocrine form of prostate cancer. And we had one patient that had, in fact, a patient-derived treatment-induced neuroendocrine prostate cancer. And when we looked at the dose response, analogous to what I told you before, only the otherwise lethal and drug-resistant neuroendocrine prostate cancer responded to the pharoptosis-inducing agent. Only the neuroendocrine one lacks expression of the androgen receptor, but we noticed had upregulated this transcription factor ZEB1. More about that later, but it got our attention because ZEB1 is a master regulator of lipid biochemistry. ZEB1 also upregulates what's called a mesenchymal cell state, but I hesitate to call what I'm describing epithelial to mesenchymal transition because we see it in cells that are not epithelial and therefore can't undergo this transition, but they undergo a related kind of transition. So these are melanoma cells. We used 49 different cell lines of melanomas. And recently it's known that many melanomas respond to targeted therapies like RAF inhibitors. That induce apoptotic death, debulking the tumors, but then become resistant. And biomarkers of these two states have been identified. The normal state of the tumor has a high expression of a biomarker called MidF. The resistant one, a high expression of a biomarker called Axl. And so when we use the gene expression signature involving Axl and MidF, we again found these two classes of compounds. And the ones killing the otherwise resistant cell states are the same familiar compounds. And just to give you a sense of the dramatic way in which sensitivity changes, this is the 49 cells we honed in, follow-up studies on a subset of these that are in blue, high BITF and sensitive to targeted therapy. In red, high-axle and resistant to targeted therapy. And you can see the ones that respond to targeted therapy really don't care about GPX4. This is very high concentration of a GPX4 inhibitor. But when they become drug resistant, they become exceedingly sensitive to the ferroptosis-inducing agents. Each of these are examples of basal states of cancers.. But what we really care about is what happens in a patient where the tumor starts out responding but then over time becomes resistant. These are— the field calls these kinds of cells induced to become resistant, they call them persister cells. And I'll show you how we see the same phenomenon in vitro and in vivo. So these are different types of cancers. Here's an ovarian cancer that's responding to Chemotherapy in blue, that's the parental cells. They don't care very much about GPX4 inhibition. But when they become resistant to chemotherapy, they become sensitive to the GPX4 inhibitors. We see that with melanoma, with lung cancer, and multiple types. Well, these are in vitro. What happens in an animal model? Well, if you start with a cancer cell that's already in this drug-resistant state, the high-AXL targeted therapy-resistant melanomas, you can engraft in a mouse, you can make a tumor, and if you remove the GPX4 from the mouse, the tumors go away. Well, what's really of interest, however, is what happens in most patients, in human patients. So here what we did is we engrafted the targeted therapy-sensitive tumors These often have mutations in BRAF. And we treated them with the frontline therapy for such melanoma cancers, a targeted therapeutic called vemurafenib. These cells don't respond to GPX4 inhibition. But if you put them in a mouse, grow up melanomas, and then treat them as a patient would receive with vemurafenib, the BRAF targeting agent, the tumors debulk. They go away, just as in humans. However, in humans, about 8 to 12 months later, invariably, the tumor comes back. It is drug-resistant, resistant to the targeted therapy, and highly lethal, highly metastatic. Now, we induce this state with vemurafenib, and if we don't remove GPX4, the tumors come back drug-resistant, but if we remove GPX4, we have a cure of melanoma in mice by using targeted therapy to induce the resistant state and then tackling resistant state. So what I've said thus far is there's an enzyme, GPX4. It converts hydroperoxides into alcohols. I didn't tell you it's a selenoprotein. I said that when we treat with these compounds, we induce ferroptotic death. So what other compounds are in this category? Well, the next three, statins. Statins, lipid-lowering drugs, they target HMG-CoA reductase. This was of great interest to us because we knew that in order to introduce selenium into a protein, you need a very unusual isopentenylation of a transfer RNA and that that requires HMG-CoA reductase. And indeed, we were able to show that statins are inhibiting the biosynthesis of GPX4. We found other compounds, Arastin and BSO, that inhibit the biosynthesis of the cofactor glutathione that GPX4 is dependent on. We also found chemicals that when we added them protect the cells from the spherotoxic death. For example, vitamin E, which is known to quench the otherwise lethal oxyradicals. Or a drug called xylutin, which inhibits the lipoxygenase enzymes that convert polyunsaturated lipids into the lipid hydroperoxides. Now one last compound very briefly, ML239. We had included it because My lab had uncovered this compound. Once again, it seemed to do very odd things. We did not understand it. But luckily, Matthew Reese teamed up with Brenton again, who I've mentioned, Brenton Seashore Ludlow, and illuminated the mechanism of action of ML239 with a new technique described in this paper. And lo and behold, amazingly, this compound truly is a novel mechanism of action agent. It activates, activates the enzymatic activity of an enzyme that is a fatty acid desaturase, meaning it's an enzyme that introduces unsaturation to create polyunsaturated lipids. So you need those in order to fit into this pathway. So overall, it turned out every single compound yielded a coherent biology around lipid hydroperoxides. But that takes me to the final section of this presentation. So what is going on here? How does this, how does this come about? What is unique about this pathway in the therapy-resistant cells? And how does this hyden, the zinkamyl, hyaxyl, these biomarkers for some resistance state, how does it connect? So to answer these questions, we again use a systematic and unbiased approach like compound sensitivity But now we used other kinds of measurements. We measured metabolites, transcripts, and the activity of genes in an unbiased way and asked which ones are either correlating or essential for this phenomenon. So most illuminating was the metabolite analysis. We measured 400 metabolites in all of our cells and looked at the GPX4 non-dependent ones versus the dependent ones. And it just jumps right out at you. If you notice, a lot of TAG means triacylglycerols. The little number after the colon means the degree of unsaturation, and you'll see a bunch of 1s and 0s and 2s. Whereas in the GPX4-dependent drug-resistant state, you see P, which stands for phosphatidylcholine ethanolamine lipids, and large degree of unsaturation, 4, 5, and 6 degrees of unsaturation, including the famous arachidonic acids. Now, the arachidonic acid cascade, as illuminated by Bergstrom, Samuelson, and Vane, bifurcates to make either triacylglycerols that are tied up in what an organelle called lipid droplets, or the phospholipids that go to the cell membranes. We noticed that our resistance— resist— GPX4-independent cells were filled with lipid droplets, but our GPX4-dependent ones were not. So we said, let's put lipid droplets on this pathway. We also knew how a cell eliminates lipid droplets. This goes back to that compound rapamycin. One of the earliest findings about rapamycin was that it activates a process called autophagy. And autophagy is the mechanism that cells use to hydrolyze lipid droplets to allow the synthesis of phospholipids. So we ask, what if we take a GPX4 non-dependent cell without having seen chemotherapy or targeted therapy and simply treat it with rapamycin, turn on autophagy? Turned out that is enough to create this GPX4-dependent state. Turning on autophagy to eliminate the lipid droplets takes non-responsive to highly sensitive to pheroptosis-inducing cells. So we put autophagy, and a selective autophagy called lipophagy, meaning lipid droplets, onto our pathway. Next, we look at the genes necessary for this kind of pheroptotic death. And lo and behold, in a genome-wide loss-of-function screen, we found two enzymes critically important that take the arachidonic acid into the phospholipid state. So these enzymes go back on our pathway. So we're beginning to see the makings of this cell state. Yet we still don't know what's upstream here. So remember I told you about ZEB1? I mentioned it briefly. It turns out when we systematically knocked out all of the genes using a CRISPR genome-wide screen, not only did we find those biosynthetic enzymes, we found ZEB1. Now, ZEB1 is fascinating. Oh, I'm sorry. Even before that, we were clued into ZEB1. 'Cause we did a different kind of analysis and asked if we rank order cells by transcript levels, what is the pattern of expression of all of the transcripts, each of the individual transcripts that best correlates with the pattern of sensitivity to our ferroptosis-inducing compounds? Number 1 on the list was ZEB1. The knockout experiment says, the first one says more ZEB1, more sensitive. This experiment shows no ZEB1, not sensitive. ZEB1 is essential. What does it do? Well, it drives EMT, epithelial to mesenchymal transition, but so do half a dozen other master transcription factors and they are not on this list. What is unique about ZEB1 is that it is a master regulator of a lipid program that we believe underlies this GPX4 dependency and therapy resistance. So that puts ZEB1 on the list. Now next, we were clued into this one because ZEB1 is upregulated by an extracellular protein factor called TGF-beta. And so when we turned to colleagues at University of Zurich led by Mitch Levesque, who had a number of patient-derived melanoma cell lines and simply treated a large number of cells in bulk with TGF-beta with no cell death, therefore no selection. Over 10 days, we found TGF-beta induces resistance to targeted therapy. And TGF-beta induces sensitivity to the GPX4 inhibitor therapy. And again, huge shifts in dose response. So now we put TGF-beta up there. I won't take you through the rest of the analysis, but we found 3 other transcription factors that work with ZEB1.. And amazingly, together they were known initially in heart biology by converting fibroblasts in the heart to what are called myofibroblasts. And indeed, these factors work together in cancer resistance to turn on a muscle-like state of cancer cells, having many of the hallmarks of muscles, including a voracious appetite for unsaturated lipids. So altogether, we think this is what is going on. To summarize, I said there's the naive state— cancer cell state— undergoes a therapy resistance transition with loss of apoptotic potential. But the good news is these otherwise resistant cells are teetering on the edge of existence, but you've got to nudge them to the therapeutic death mechanism. I told you this was a pan-cancer effect. So most cancers today are very specialized. You look at liver cancer, you look at skin cancer. So what I've told you about, we've now uncovered in high-axil melanoma with mutant BRAF, in EGFR mutant lung cancer that's become independent of EGFR. In KRAS mutant pancreatic cancer that's become independent of KRAS. And then finally, in androgen-independent prostate cancer. So we all have learned that cancers are caused by oncogenes, and targeted therapy rests with the fact that they're dependent on these oncogenes. But as they evolve in the presence of chemotherapy or targeted therapy or immunotherapy, They become resistant, but they also become independent of the oncogene. So we think this is an oncogene-independent state that's contributing to therapy resistance. And it's pan-cancer across many different kinds. These persisters, and we even— although I didn't show it— found it in certain sarcomas that start out life in this cell state. To summarize, cancers are apoptosis competent. All therapies known today exploit this fact. But what was not known, and what I've tried to argue for today, is that they are capable of undergoing a drug-induced myofibroblastic-like cell state switch without any mutations. That's a non-mutational cell plasticity. They become apoptosis resistant. Arjun Raj's lab last year discovered, reported on discovery of a funny feature of melanoma cells. He noticed that there was a kind of blinking phenomenon. He reported that when melanoma cells are growing, 1 in about 1,000 undergo a very strange, emerge the daughter cells in a strange state. He called them jackpot cells. We got together with Arjun. It turns out his jackpot cells, they're our myofibroblastic cell state. But what's really interesting is he didn't have any chemotherapy or targeted therapy in this case. And what he showed was that when this cell state emerges, the next cell division, that both daughters go back to the parental state. So I think it's some stochastic process just sitting there waiting for apoptotic stress induced by cancer therapies. Now these cells selectively survive, and I believe these are the apoptosis-resistant persisters that I've been describing that acquire this GPX4 dependency. Where are we going? Well, I started with a hypothesis. I'll finish, come back to that hypothesis. We know mutations arise in drug-resistant states. They've always been thought to be the basis of resistance, and I'm arguing they're not. That the survival of cancer cells is the cell plasticity phenomenon. But in every instance when we found one of these persister cells, they grow very slowly. So we think mutations are coming along to optimize the fitness of those cells so that they now proliferate. So for example, if you're not dependent on EGFR anymore but the patient is still receiving an EGFR inhibitor therapy, You don't need a mutation to survive, but you need to remove the irritant of the drug if you want to divide rapidly and get signaling through the EGFR. What we don't yet know is whether this final state where the two have joined together remain in this GPX4-dependent state. We have some hints in one case with lung adenocarcinoma. In collaboration with Jeff Engelman, but a lot of work to be done for chemists and chemical biologists and medicinal chemists to make the novel compounds to explore the therapeutic implications of these findings further. So let me finish there with some thanks. I've tried to acknowledge my collaborators as I've gone along. Another big shout out for Brenton, who you've heard, but also in the audience I'm delighted to see is Oscar Verho. Oscar trained with Jan Bachval, came into my lab afterwards and brought some of that Bachval magic of palladium-catalyzed transformations that made these novel chemical agents that I described feasible. I'm very grateful to Oscar, and most importantly, I'm grateful to the— it's okay. Yes. Thank you. I'm very grateful to the Academy for inviting me, for Peter and for Jan, and most importantly, for all of you for your patience in listening to my story today.

Thank you. Okay, thank you, Stuart, for this Very nice lecture, and I mean, very impressive studies that you have made. And it's now open for discussion. So, yeah.

Thank you.

Would you like to comment on the similarities with persistence and in bacterial work and the resistance and the striking similarities in terms of if you could kill the persisters or if the mutation— I sure would.

—resistance develops in the persistent population. I love this question. It's— I'll tell you a little backstory. The answer is yes, and I think you've hit it spot on, that we need to learn from the infectious disease, the world of TB, for example. You know, I come from the Broad Institute. It was a genomics institute. I can tell you there's been a lot of stress and disagreement about what I've shared with you today. I tease my colleagues at the Broad and tell them, of course you don't like this because it's illegal, because it doesn't involve the mutations. But of course it does eventually. And what I've said, because I often hear about cancer and resistance, I've read very prominent articles about learning from HIV and that we need combination therapy like HIV. I disagree with that and I agree with the point that you're making. I think we should learn from pathogens like Mycobacterium tuberculosis, which respond to therapy, then become dormant, but they're still present in a hunker-down state that's slow dividing, which is called the persister state.. And then later they come back with a vengeance. So I believe that cancer is more like TB than HIV. I think that looking for combinations is not necessarily the right way because actually I didn't have time to explain, but this phenomenon, we actually have to burn in this state with the targeted therapy or chemotherapy, then later come in with the GPX4 therapy. If you simply mix the two together from the beginning, you don't see what I've described. That's very reminiscent of TB. So I think the persistent state in pathogens is highly relevant to cancer. Okay.

Some more questions?

So the general strategy or policy The general strategy that is often applied in cancer therapy of having cocktails, having a mixture, is the wrong strategy? That's a strong statement. I'm not prepared to go that far. But I think— I hope these insights cause us to think about the way in which we administer drugs. I'll share another little sidebar story. I got some good news recently. I had good news and bad news. I wrote a grant to renew this work. It got funded. That was the good news. The problem is I said, aim 3, the most important aim, is to understand the pharmacology of drug treatment. How much time do we need a targeted therapy or chemotherapy versus just cocktails of drugs? The reviewers came back and said, well, this is a meritorious study, but aim 3 is not very original or exciting or innovative. And of course, to me, it's right at the heart of knowing how to translate this to human health. So I do think that it's— I think we should at least think about alternatives to simple cocktails of— also, of course, patients tend not to tolerate cocktails as well as In some respects, the idea that the first therapy burns in this resistant state to make it permanent is a good development because patients can go on one therapy, maybe wait some period of time. Sometimes we— in medicine we call that a drug holiday, but the drug holiday was there to mitigate the suffering of the patient. The drug holiday paradoxically may have been critical for allowing the tumor to reach a state where we have a completely different way of killing it.

So having a strain of cancer cells, would it be possible in vitro to design the bullets A, B, and C that should come conceptually?

And A, making this first plateau, and then B, and then C, killing it. So you have outlined exactly the approach that our aim 3 that was viewed as not innovative But we're going to do it anyway. It's exactly the approach we will take now, knowing perfectly well that there are limitations in the ability of cancer cells in vitro to truly mimic the, the patient state, which is why we've tended to move next into the organoid system. At least it's patient-derived, and then patient-derived xenografts in mice. But ultimately, we need to do this in humans. In order to fully explore this. So that's our dream. That's our goal. Thank you. Yes, please. I mean, as you've shown, state 1, removing the first kind and then the second kind— don't remember the names. With the research that you've done, have you with that removed all the cancer cells, or have there been any traces of mutations or other— So once again, a very insightful question. We're heartened by the mouse study, but the mouse study is a xenograft mouse study with all of its imperfections. But as far as we know, those mice, as far as they've been taken out thus far, no melanoma has returned. I think it would be very unlikely that such a simple state will be seen in patients. And so we are anticipating that cancers are very smart, like the pathogens that we're modeling after them. And my guess is there will be some escape mechanism that— and we're going to have to figure out what that is. I'm hopeful that this— the fact that this cell state occurs, you might say epigenetically, or you might say that it's because of a new gene expression program, we have evidence that it takes a long period of time to burn that in in a more DNA methylation permanent chromatin way that we— it's at least different from sort of traditional therapies, and that maybe it'll be more difficult to escape that. But I, I suspect there will be escape mechanisms, and that's just another step in the process of figuring out how ultimately to cure, cure patients of their cancers. We make no, you know, false illusion that stepping into this problem that anything will ever be easy. The cancer will always find some devious way, but at least hopefully this is a step in that progression.

Some more questions? Well, I was— when I was sitting there wondering, this selenoprotein GPX4 apparently plays an important role.

Is the structure of that known? The structure of GPX4 in the cysteine version is known. There's lots of progress that's being made with the selenoprotein version that I'm aware of that's not been published, but it is a very challenging protein, the selenoprotein cysteine version. And I think this is part of what it's doing. I suspect that it's functioning in the cell at the plasma membrane in a multiprotein complex. And when we find the critical proteins needed for this GPX4-mediated effect, we do see other proteins that may be candidates for interacting and forming. So I think structurally it may end up being challenging to really get at like the multimeric protein complex. But the cysteine version structure is known. It's not a very appealing binding site for chemists. The enzymatic activity is a very flattened region. It's not a traditional pocket. And as— but as I say, I think there's a lot of progress coming along in looking at the selenoprotein version.

Thank you. So I don't see any more waving hands. So if there are no more questions, please join me in thanking Professor Schreiber again for a very stimulating lecture.