Founder Stories: Jay Bradner, Chemical Biology to Power New Medicines

Roderick, thank you so much. I've really enjoyed being a viewer to this lecture series, and I'm so pleased to be invited to have a chance to share some reflections on the space of targeted protein degradation, and also some new data. I tried to bring some new findings from our research at Novartis that I hope you find interesting, and I welcome any questions at the end or thereafter. Especially having trained at the Farber and having had 8 wonderful years on faculty at the Farber. It's a special pleasure to visit with you. I just wish it was in person. As already shared, I have these disclosures to make as this is an academic setting. I'm a shareholder, also serve on the executive committee of Novartis, but this will be a thoroughly scientific presentation of a research nature. And I tried for balance to include a lot of insights that derive from research beyond my laboratory at the Dana-Farber or Novartis itself. But I am sharing this one bias that I am all Farber all the time, Pan Mass Rider, and still carry the torch proudly. This space of molecular glues is getting really hot and exciting, but it's a little bit what's old is new again.— for so many, Novartis included, that has had a long legacy in the study and in the clinical development of medicines that function as small molecule adapter proteins between two large biomolecules. But I did want to share some insights from that long legacy of study, and also the new strategies that we're taking to really understand this new class of agents. So maybe a definition is required. I don't intend that this be a definition for the whole field. This is just the way we think about it at NIBR. Glues come in at least two flavors, and I'll touch on both today. Intermolecular glues—these are molecules, typically small molecules, that stabilize proximity of two or even more biomolecules. Great examples of this from nature would include cyclosporine A and FK506, as shown here. You can see in the crystal structure of cyclosporine A on the left, the way that the immunophilins are arranged around this core macrocyclic natural product. Present there is cyclophilin, as well as calcineurin A and B. And this adapter molecule, cyclosporine, has, of course, become for many years now, an important medicine in transplantation. I myself am a transplant doctor and prescribed it many times. Even more have I prescribed FK506, which is perhaps a better tolerated but equally potent immunosuppressant that bridges calcineurin A and B to FKBP12 and other FK-binding proteins. These intermolecular glues in this instance abrogate inflammatory signaling pathways in lymphocytes and other tissues, But this concept now has been engineered into many different networks to short-circuit pathobiological protein networks, typically with these drug-like small molecules. There's another flavor of molecular glues that I think are really important, maybe underconsidered, and these are intramolecular glues. These are molecules that conformationally restrict multiple domains of a single biomolecule through coordinated bridging interactions. Could be lobes of a protein, it could be subtle features, or— and I'll share an example of that if we get to it in this lecture. So how do you know you're working with a glue versus, say, a heterobifunctional protein degrader or PROTAC? One framework that we commonly use in the parlance of NIBR is that a glue does not exhibit saturable kinetics, meaning there should be no competition for active sites by a second small molecule. This differentiates it from bifunctionals that exhibit this hook effect, as we and others have characterized. I think that a brilliant historical consideration of the space of molecular glues was just recently contributed by my mentor Stuart Schreiber, whose lab I trained in as a postdoc, reflecting on the early days of these immunosuppressants at Sandoz, now Novartis, and so many other elite scientific institutions, to the present consideration of molecular glues in the context of degradation. And this is a— as this is a degradation meeting, maybe the bulk of this will be reflections on where we are right now in targeted protein degradation and where we might go with intermolecular glues. So I hope I get through each of these 3 examples, but if the conversation directs towards one, that's okay too.. So my own personal curiosity with molecular glues really began as a myeloma doctor at the Dana-Farber prescribing this remarkable new class of medicines, the thalamides. And where so little had helped myeloma patients since the mid-'50s with melphalan and prednisone, we had these new weapons, thalidomide itself, lenalidomide, and ultimately pomalidomide. And these were medicines that of course have a very Dark History. Ben Ebert from the Farber recommended this book, Dark Remedy. I know he's spoken at this group meeting before. Perhaps everybody has read it, but I do highly recommend this reflection on the way that the loose chemical analogy of thalidomide to phenobarbital led to the regrettable consideration without any preclinical evidence or suggestion that these medicines, these glutaramides with thalidomide conjugation might be sedatives. And given to women in post-war Europe, it was learned that these are potent teratogens and first trimester. This led to many important FDA reforms and later the reconsideration of this drug class through decades of phenotypic reconsideration. They emerged as modulators of immune cell signaling. It's still imperfectly understood, I think, how that works. But it was true that thalidomide in cultivated monocytes would decrease the elaboration of certain pro-inflammatory cytokines, establishing a rationale for its use in patients in inflammatory conditions like leprosy, with erythema nodosum leprosum as a one hallmark dermatologic manifestation. And the medicine was even approved for this use. But still, the mechanism of action of these imids was not well understood. But science soldiered on, and in the Folkman Lab nearby to the Dana-Farber Cancer Institute at Children's Hospital, it was learned that thalidomide, might retard another interesting phenotype, the proliferation of retinal vessels, as is commonly observed in acute macular degeneration and other tissue injury to the retina. And so there was a putative activity for angiogenesis, which had been connected by Folkman to cancer, leading to the consideration of thalidomide as a multiple myeloma drug, brought to Bart Barlogie at the University of Arkansas by Beth Jacobson, married to an attorney under the care of Folkman, a 35-year-old cardiologist with myeloma. Barlogie there in Little Rock bravely tried thalidomide in this patient. And though there was no clinical benefit to that man, he stuck with the idea, ultimately demonstrating through larger clinical trials with Celgene remarkable activity. And now these IMiDs have become a standard of care in that disease. An amazing progression of human phenotypic screening. We have been very curious as to know how thalidomide works. A handful of us in training over at the Broad Institute, and with the advent of SILAC, ligand affinity mass spectrometry, or target identification technology innovated by Xiaolin Ong and Steve Carr, Ralph Mazicek and I fashioned some patient-discarded Revlimid onto polystyrene beads and went fishing in the human proteome, and we didn't find anything. But we were just delighted to read that Hiroshi Honda, using an alkylation chemistry in a different intermediate versus the arylamide coupling that we had pursued, was able to identify the biophysical target of fulidomide, published in Science magazine in 2010, as the cereblon DDB1 ubiquitin ligase complex. And this got us very excited, because we imagined at the time that this molecule might be inhibiting ubiquitin ligase complex, might be blocking its transfer of ubiquitin to other substrates. But we weren't so concerned about that at the time, we were very taken in my lab with the idea that it bound cereblon at all, meaning that perhaps by binding cereblon on a polystyrene bead with a linker, we could drag this ubiquitin ligase complex around cells, degrading protein targets of interest, using heterobifunctional small molecule degraders, so as to redirect cereblime to a physiologic or even therapeutically relevant substrates. And there's a strong rationale to protein degradation that attendees to this lecture series certainly don't need me to recite, but I've listed some of them here. And why I remain so excited about protein degraders is the potential access to intractable proteins, difficult-to-drug proteins. You see, in most of medicinal chemistry, our monosynaptic strategy for inhibition is to bind to and disrupt active sites, like so successful has been our and other innovation of selective inhibitors for kinases. Sometimes the druggable part of a protein isn't the active site. Then what do you do? So there's this possibility maybe to bind and degrade in spaces where you can't easily inhibit, and in the space of molecular glues, Perhaps even you don't even have to bind the protein at all through your small molecule. Rather, the small molecule completes an interface created as an ensemble between the small molecule and the surrounding protein surfaces, recruiting proteins that would just be purely undruggable through any other consideration. And that's a molecular glue, again, one that doesn't have saturable kinetics. The other reason we're really excited about degraders is that they might be stronger than inhibitors to take out all the multiple domains of a protein. And then third, the pharmacology might be different, that the durable impact of degradation can only be overcome by cellular signaling with reconstitution of the protein itself, therefore uncoupling pharmacokinetics with durable pharmacodynamics. Well, I think it's also, as an industry scientist now, worth reflecting that this idea was first conceived of in industry by two discovery scientists, Kenton and Roberts, who even patented the concept of making molecules that bridge the ubiquitin ligase system with targets of interest, but they never did get it working. And there, the seminal work of Crews and Deschez using peptidyl fragments of HIF2α and engineered solutions sometimes in xenopus extracts established the their seminal contribution to PNAS, this idea of PROTACs, bifunctional molecules that can degrade proteins on command. But there was something holding the field back, even from this conceptual framework and early suggestive evidence that it could work, and, and that is that peptides are hard. They're big, they don't penetrate cells, they're not drug-like. I've been down that road with stapled helices, and what was really needed were chemical solutions. That would lead to facile cell penetration and use in vivo, where so many— so much interesting biology happens these days. And thanks to the discovery of Honda, a group of scientists in my lab then at the Dana-Farber, Georg Winter on the left, Dennis Buckley on the right, created a first chemical solution to protein degradation that we and others have found so enabling. At the time, there were these many technological challenges, that it was engineering, it was peptides, there wasn't much in the way of characterization biochemically, low potency in the mid to high micromolar range. We didn't even really know if these things were selective or not. And so, I mean, a study that I'm sure most have some familiarity with, so I'll just browse through this quickly and get to the good stuff. We ticked off a lot of these boxes. We created bifunctional molecules to target the BRD4 bromodomain-containing protein that we were so curious about mechanistically in transcriptional elongation and proximal promoter pause release. And with this bifunctional molecule and the alkylated thalamide that you see here, we were able to achieve potent recruitment of both the Cerblon-DDB1 complex as well as Brd4, solve the crystal structure of this bifunctional degrader in Brd4, and you could see we lost atomic resolution in this linker domain purposefully created to to bridge and recruit the ligase, leading to really unprecedented potent and selective degradation of BRD2, 3, and 4, the intentional targets among 7,000 proteins queried. And this molecule worked in vivo, which we thought was quite exciting because we could test its putative utility as an early proof of concept molecule for anti-leukemia therapy. And the reviewers of Science were tough. It was in review for a while, and they said, you know, This is amazing, but do it again, which is kind of harsh. And so, we did do it again. We went back, and as BRD4 is principally a nuclear protein, we degraded the cytosolic protein FKBP12 with a series of ligands of increasing potency that allowed us to establish rough guidance as to the thresholds necessary for this new brand of pharmacology at about 150 nanomolar cellular binding energy at equilibrium. But as the bromodomain lab, we couldn't help but try this chemistry out on other bromodomain-containing proteins, and the field of chemical biology and SGC in particular had created a number of small molecule binders to bromodomains, but most of these were not terribly active in cellular assays, that the bromodomain perceived as being relevant for its capacity to bind acetylated lysine in nuclear euchromatin maybe didn't have as much relevance to cell biology as say, the acetyltransferase domain of CBP and p300, or the ATPase domain of Brd1 and Brm. And so scientists in the group took ineffectual or low-potency bromodomain binders and turned them into quite potent degraders. We did this with Brd9, taking initially a non-selective Brd9 binder and turning it into quite a selective BRD9 degrader, validating that target in leukemias, myeloid leukemias. With TRIM24, we took a peptide fragment from Les O'Toole's work, from Dennis Buckley's PhD thesis work, and appended it to a TRIM24 bromodomain binder, creating a nice degrader. And GSK did the same thing. I thought it a beautiful study of GCN5, where the bromodomain binder does nothing at all,. But as a degrader to both pCAF and GCN5, a selective degrader therein, they could block inflammatory signaling. So the field's moving really fast, and I thought I might just share some emerging themes or truisms that we're starting to see replicated in the literature. First, this is just like a surgical strike. For some reason, we by and large degrade the protein target of the small molecule, but we don't degrade all the other proteins associated with that target protein. There are some exceptions, but they're very few, and I find all of them very interesting. Panagis Philippakopoulos and colleagues show that BRD4 might bind to 60 proteins in the cell, but only 3 proteins are degraded, and those are the direct biophysical targets of the recruiting ligand. Sugal Kadoch at the Farber showed that there are, you know, 30 or more components to the non-canonical BAF complex that contains BRD9, but we only degrade Brd9 with our Brd9-directed degrader. And perhaps there's some insight there about the geometrical— geometric restrictions with ubiquitin transfer, as we brilliantly discussed at the last TPD seminar series here, and as Alessio Ciulli well discussed in his elegant talk. It's fair to say we're learning more and more about the geometry of ubiquitin transfer through these small molecules, but they appear to be quite selective to the protein engaged. Second is that these are like little molecular catalysts. We showed for the first time using paired cell lines, TALL cell lines that did or didn't have the ubiquitin ligase Cereblon, that these degraders without Cereblon behave like bromodomain inhibitors, but in the presence of Cereblon, there's a 2 to 3 log shift towards increased potency. This is, of course, because the small molecule is not degraded, but the protein target is, and has led to even sub-nanomolar degraders for BRD4, so potent that many out there are now considering them as substitutes for natural products in antibody drug conjugate design. This substoichiometric shift is really exciting, but also gives me the chills a little bit. Because once you're in substoichiometric degradation space, you have to really worry about off-target effects that would therefore also be potentially substoichiometric. More on that in a minute. We know now that degradation is very different than inhibition. We should have known that, because there are a lot of things that hairpins do that small molecules directed at protein targets don't. But the JQ1 bromodomain inhibitor causes a cytostatic response, while the degrader causes a cytotoxic response. The first clue. But as we characterize in a Molecular Cell paper a couple of years ago, the transcriptional consequence of inhibition versus degradation of Brd4, we learned something quite fundamental about Brd4 for very much the first time. That if inhibition of Brd4 downregulates these short half-life or super-enhancer-associated genes, Brd4 degradation wipes out all of productive transcriptional elongation. Establishing Brd4, as we did in this study with Sterling Churchman, as a master elongation factor licensing the CDK9-PTEFB complex to license RNA polymerase for proximal promoter pause release. So these small molecules can be really elegant temporal probes for mechanistic cell biology. We've learned that ligand specificity does not often direct degrader selectivity. And we keep learning this over and again. The first insight came from our work with BRD9, where the BRD9 binder we were using would also bind to BRD4 and BRD7, but the degrader would only degrade BRD9, which was interesting. We followed up on this with Nathaniel Gray, in transition from the Farber to Stanford, and hello Nathaniel and colleagues, where we took some of the most non-selective kinase inhibitors we could find that literally light up the kinome by kinome scan binding assays, but that degrade only a small number of kinases. Perhaps there are geometric reasons for this, accessibilities of lysines, accessibilities of proteins, turnover rates, probably many factors, but we have been inspired by this to reconsider old chemical equity where we failed to engender adequate selectivity and, and give these projects new life as degraders. We have a handful of those in our pipeline now. Protein degradation chemistry can be good for target identification. Target identification can be really challenging. If ligand affinity chromatography doesn't work, it can be really challenging. We had a sunitinib derivative that we made to degrade FLT3, and it just killed these FLT3-mutated leukemia cells. We were so excited about this, but when we looked for the protein consequences of this small molecule, using kinase-enriched proteomics, we didn't find any kinases degraded. And this was kind of a bummer, so we did the project again. We didn't find any kinases degraded, and so we did shotgun proteomics and found that this intended kinase degrader potently degrades GSPT1. And this would be a sort of a toxic consequence of this molecule, contributed not, as it turns out, by sunitinib at all, but by the ensemble of the cereblon binding feature and its appended linker. And so one might take a small molecule of phenotypic interest and consider it as a degrader, and maybe you'd find a new protein of interest mediating that phenotype, and one such paper was brilliantly contributed just last week in the chemical biology literature. What do you do if you don't have a ligand for a protein of interest? What if you kind of want to ask the question, is the juice worth the squeeze? Or, or you want to validate a target before undertaking discovery. And so, but on Nabet in my lab now, completing his research with Nathaniel Gray, created a really cool technology I'm sure most on this call know all about called the dTAG technology. We had the idea that we would leverage this bump-hole relationship with the AP1497 chemical series that has a bump, and then build the established hole in the FKBP12, the F36 valine mutant, that accommodates uniquely this halogenated derivative of AP1497 from Klaxton's work at Ariad Pharmaceuticals. And then this little 12-kilodalton open reading frame can be engineered as a chimeric, the amino or carboxy terminus, leading to selective degradation of only that protein of interest. And this has been very helpful to our work to validate targets pharmacologically, and to understand the fast biology downstream of fast-acting proteins. As reported, so I won't go too far into it, we became interested in the Yates domains that that also bind acetylated lysine, structurally dissimilar from the 4-alpha helical bundle of the bromodomains, and whether any of them might be involved in transcriptional pause release, and if any of them might be interesting as blood cancer targets. And we focused on ENL. And using dTAG, we could selectively degrade ENL. I hope you can see my cursor as shown in the upper right. And the bump hole was working, as we did not degrade cytosolic FKBP12. And now with this chemical tool to only interfere with one protein in the cell, we could ask questions like, do cancers depend on ENL for their viability, as this IC50 curve shows? And then what are the downstream transcriptional consequences of losing a master elongation factor like ENL? And this pointed us in the direction of a core circuitry of cell-specifying and state-defining transcriptional regulators. It could be handy to have versions of dTAG that work not just with cereblon, but with other ubiquitin ligases, and so we've recently, just recently reported Benham's work to develop a VHL-dependent dTAG system, and now with a host of small molecules that bind VHL as miniaturized peptidyl derivatives of the HIF2 alpha, you can see the vestigial hydroxyproline in these structures. Benham's elegant work has created a series of molecules that are themselves stereochemical pairs, active and inactive, to allow the study of target-directed degradation now via cereblon or VHL, and these probes importantly, as shown in the lower right, work in vivo, leading to the fast, profound, efficient, and durable degradation of proteins of interest, and Benham has remarkably made these available to scientists around the world. Back in 2014, we got excited about this, this dTAG technology as maybe being therapeutically relevant. At this time, CAR T-cell therapy was emerging as a very active cellular therapy for blood cancers, in particular B-cell leukemias and lymphomas. And we were caring for some of these patients at the Dana-Farber and following the work of Carl June and others with incredible admiration. But many of these patients in Seattle and Juneau studies and others were having severe neurotoxicity. And so we imagined a capacity to use dTAG to turn off the protein payload in a way that would abrogate this toxicology, but wouldn't actually kill the CAR T cells, so as not to interfere with the persistence of the CAR T, so important we believed at the time and now know to contribute meaningful logarithmic tumor reduction and durable clinical responses. So we cloned two CAR-T cells targeting vectors directed at CD19, but with two different flavors of intracellular CD3 activating domains, one built on NKG2D, and, and the other is the traditional UPenn-Novartis chimera. And what Bonhomme showed is that if you put the dTAG on the extracellular surface, of course, you don't degrade, but if you put it intracellularly, then you actually can't quite potently degrade here with these cereblon degraders. And this work, I believe, was licensed to C4 Therapeutics, which we started out of our lab in 2015. In truth, at Novartis, now Innovating CAR-T Cell Therapies, we haven't, for the blood cancer therapies, at least as we use our constructs, needed post-translational degradation capacity. But perhaps bringing CAR T-cell therapy more broadly into solid tumors, these types of approaches could be really useful. And so I think it's an area to keep your eye on. We surely do. But Benam has been very generous to follow the example of chemical probes from the Dana-Farber chemical biology groups, ours included, and make this broadly available, even wrote a how-to document. And I learned the other day that there are 800 now distributions of these vectors and many, many of the small molecules. And so we hope that the number of studies that we see across biology using the dTAG system to grow— and it's super cool. I hope this works well over this medium, but here's a dTAG on a GFP but not on an RFP. And you can see the way that this allows the study of really, really rapid biology in vitro and in vivo. We've learned through all of this that ligases have some substrate preferences. We tried to degrade the TRIM24 protein with a TRIM24 bromodomain-directed small molecule, and initially we worked off of cereblon and we could not degrade TRIM24, but by pivoting to a VHL-recruiting bifunctional molecule, we could, and that was good learning. There are now binders to probably 12-ish ubiquitin ligase complexes that discovery chemists and chemical biologists can work with. I think there's still a need for, for others, but here the pending VHL binding ligand was really important to create a bioactive degrader of TRIM24, which we use to invalidate the target in breast cancer, but to suggest the target in blood cancers. We've learned that ligases have conformational preferences, and as you have Eric Fischer there and had Alessio out recently, I won't go deep on this, but suffice it to say that through linkerology, we together have been able to achieve remarkable selectivity for only degrading BRD4 with molecules that have been always very non-selective between BRDs 2, 3, and 4. And we worked for many years with a bunch of graduate students and postdocs to try to get selective BRD4-only antagonists, couldn't couldn't do it. But with the geometry perhaps of this degradation chemistry, we arrived at ZXH3-26, which is, I think, really the first truly selective degrader of BRD4 across a several logarithmic range consistently in many different cell types. And that's a really useful chemical tool, because other than with dTAG and hairpins, we haven't really been able to ask the question, what's BRD4-specific biology versus, say, BRD2 and 3. We learned that ligases can degrade ligases. This hybrid molecule was able to degrade MDM2, its self-ubiquitin ligase, through Cereblon recruitment. Jan Kronke made this really cool suicidal molecule that causes Cereblon to digest itself, and Kronke then made the molecule where VHL goes over and degrades Cereblon and not the other way around, if that's a window perhaps into the pleiotropic degradation capacity of VHL. And that's cool because there are a lot of really interesting ubiquitin ligases implicated in human disease. We've learned that resistance is quite polygenic. Data from our own Constantine Mitsiatis reported this week with some help from folks from our lab that there are any number of opportunities for evasive resistance to small molecule —bifunctional degraders used as anti-cancer drugs, and Garrett Vinter had published just earlier that— a really gorgeous study looking at multiple different targets and multiple different ligases, that the CRL-UPP system just lights up in forward chemical genetic studies of pathway-directed, in this case CRISPR hairpins, selection for cellular clones that can grow through the pressure of these ligase-directed small molecules is quite easy to achieve. And this has been the clinical experience as well. Though the IMiDs are profoundly important for multiple myeloma, as in the literature recently, there was a gorgeous study of somatic alterations to cereblon that are associated clinically with progression on lenalidomide. And this is just a cautionary tale that even epigenetic downregulation of some of these pathway members could lead to evasive resistance. This work has proven quite extensible and feels super real right now. Arvenis has made a cereblon-directed BRD4 degrader. Alessio, as hopefully he shared, has made a BRD4 degrader off VHL. Craig has recently contributed a degrader for BRD4 that uses recruitment of MDM2. And BRD4 has kind of come out as sort of like the testing or the proving ground for a lot of different degraders, which cracks me up because, as I was so new to chemistry when I started the lab, people always called BRD4 a target so easy to inhibit even an oncologist can do it, and maybe now it's the target so easy to degrade even Brenner could do it. Together with Dan Nomura at Berkeley, we've created a center for covalent chemoproteomics, where we're mapping the ligandable cysteines in the human proteome, so as to build a reference for starting points for ligand discovery across the 8 or more disease areas where we develop medicines at Novartis. This work has been super fruitful also to find novel irreversible binders to ligases, and Ben Cravatt and many others are doing this as well, but recently reported we've now found small molecules that bind RNF114 as the nimbulide natural product does, and I know Dan's had a chance to come. It's just a growth industry, you guys. You know, there's so many great contributions to literature. It's really thrilling to see this chemistry take on, and there's just publications everywhere. There's, you know, like, um, 2 dozen drug companies, billions of dollars being invested, and all of this makes us very hopeful that there will be really new therapeutics rising. So back to Gluz and Novartis, we started when I joined in 2016 a targeted protein degradation initiative that would be really nimble and inviting. It's sort of like an open sandbox to experiment with this chemistry, and it's recruited the attention of a lot of scientists from different disease areas. And we build a toolbox of assays, methodologies, proteomic capacities, and importantly, small molecules that scientists can access to, you know, really quickly get to second base and have a molecule that works proof of concept in their cellular system. But we need to go into a drug-like direction and always trying to look around the next corner. In 2016, we put a really asymmetric emphasis and resource into molecular glues, and we actually now have 4 molecular glues in human clinical investigation. 2 are what one might call intermolecular glues and 2 are intramolecular glues, and I'll refer to a couple of them in the closing moments here. It's been great to have Nico Tomei at Novartis at the Friedrich Miescher Institute, is one of our research institutes at the Novartis Institutes. And I know Nico and Eric have lectured in this series, and insights from Phil have supported this idea that orally bioavailable small molecules can recruit proteins like Ikaros transcription factors that nobody would have seen coming. And I want to acknowledge the contributions of Bill Caelan to this work at the Farber as well, with whom we reported in 2014 the surprising finding that small molecule binders of cereblon don't inhibit, but they actually confer a gain-of-function capacity to degrade Ikaros. And that Ben Ebert's lab concurrently reported and markedly extended these findings, establishing even a primary structural rationale to that mode of action. I want to acknowledge Eric Fisher's really important contributions that build on the suggestive evidence that SAL4, the causative human developmental gene associated with Oka Hiro syndrome, a syndrome that looks an awful lot like focal milia attributable to thalidomide toxicity, might be the, the teratogenic consequence of thalidomide, and we and so many others believe that it is, and build molecules now so as not to have this capacity. We recently reported a new sort of dTAG system called the HILD TAG system that uses IMiDs and then a much, much smaller dTAG than even the 12-kilodalton FQBP12, one based on Ikaros, and showed how this dTAG might be useful to build a safer CAR T-cell, as I've already shared. And wanna acknowledge the brilliant work of Max and others who presented at this series, and that was also so exciting to hear how both an on and an off switch have been considered with this biochemistry now in the Ebert lab, collaborating with Marcel at MGH. Really cool stuff. At C4, surely independent of any contribution of mine, this is not a collaboration with C4, but C4 has created an A-tag, an Achilles tag as they call it, here using MTH1 and direct-acting small molecule binders that in the lone animal range lead to degradation capacity, considered in this presentation from Rami Zeid, a graduate of our lab, as another potential solution to open up the therapeutic index of CAR-T. I actually am really excited about all of these technologies because we're heavily invested in AAV gene therapy., and in cell progenitor, stem progenitor cell therapy at Novartis. And having pharmacologic control over protein payloads in gene therapy, it's actually a really big deal. So I think we should see, probably not in CAR-T cells, but in other applications of gene therapy, the use of post-translational switches just like these, if not these. Our studies at Novartis are maturing really quite nicely, as I mentioned, with the 4 clinical programs. It's not all we do, and so we've had to be really choiceful about where we make research investments. And in total, I would say we have probably about 8 projects trending towards clinical investigation and maybe a dozen sort of exploratory projects. And so I'd like to think we're really leaders in certain aspects of it, like the glues, but we are also very interested to partner with other groups that are making great contributions in this widely democratized area of TPD and molecular glues. Recently, we had a chance to report, Dirk, Jonathan, and Josh, and others from our lab, and NIBR, the structural basis for the surprising glue-like activity of the sulfonamides in DUSALAM. And here, as Eric Fisher has brilliantly reported as well, we find these sulfonamide small molecules really don't have demonstrable binding constants to this machinery, but like a molecular glue should, facilitate the assembly of DCAF15 with RRM proteins like RBM39 and RBM23. We haven't been as lucky to really deduce a structural degron as we have with Ikaros for this work, but we do think that it's just another sign that molecular glues are everywhere. You just have to look for them. We really wanted to find them in an intentional way though, and so Greg Michaud in our chemical biology and therapeutics group, and this paper has just been submitted for consideration for publication, has created a series of molecules that bind VHL and may or may not have glue. So instead of finding these sort of accidentally, can we find these, you know, purposefully? And so with a library of molecules that bind VHL that have a chemical diversification pre-built in, And with a series of protein arrays that present baculovirus-prepared proteins or protein folds, and they can be incubated with small molecules and the VHL-purified minimal machinery, Greg and colleagues have been able to find glue-like activity, potent glue-like activity of multiple small molecules for several different protein targets, and most notably, and best validated, is cysteine deoxygenase 1, that lit up on this array with a handful of these small molecules. We find, as one would expect for a degrading glue, that there is no binding constant, really, to CD01, but that this molecule profoundly influences the capacity of CD01 to bind VHL, and thus no hook-like activity. This molecule degrades in a manner that depends on the VHL ubiquitin ligase that's blocked by proteasome inhibition, as well as inhibition of nedylation, as so brilliantly considered at the last presentation. We've solved the high-resolution crystal structure of CDL1 bound into VHL, and it's just gorgeous. It's sort of like shades of cyclosporine A again, where this now extended peptidyl molecule is able to create a host of principally hydrophobic invagination compatibilities with CD01, but there are also important electrostatic bridging interactions as well, both with CD01 and VHL. And so in this way, this compound 5, in this case, it needs a cooler name, is really itself sort of a non-covalent post-translational modification of VHL that leads to altered specificity and recruitment. This ternary complex formation, we think, is pretty cool. It may or may not be a structural degron, but it has encouraged us to think of VHL as a framework for additional molecular glues that we might more intentionally build and direct. I use this funny term serendipity. I'm desperate for this not to catch on in the scientific world, but really, You know, the clinical validation of targeted protein degradation is so firmly established by medicines like fulvestrant, here called ICI, that for whatever reason degrade estrogen receptor alpha, arsenic trioxide that for whatever reason degrades PML-RR alpha, and of course the IMiDs that for a very clear structural reason degrade. And we're seeing serendipity everywhere. LEAD series that now instead of optimized for inhibitory capacities and binding in homogeneous assays, are considered as maybe degraders. And we're getting really systematic at Novartis now, for certain programs, to always be sure to ask the question periodically in med chem, do any molecules in this series degrade? And I'd bet dollars to donuts that, that fulvestrin is a glue degrader to some ubiquitin ligase. Recently, there are 3 just gorgeous findings, again, of serendipity. Here, the amazing simultaneous discovery from 3 groups globally, many friends on these papers, that cyclin K via CDK12 engagement can be degraded, even without the Cereblon adapter, but through DDB1 itself. Serendipity is everywhere. So when to degrade? This community hopefully is a community that will reach for the highest-hanging fruit and not, like, you know, get distracted by the opportunistic kinase this, kinase that degrading. I think especially in these pandemic times, I think about degrading coronavirus proteins and those with ligands, we'd be very motivated to help build degraders if you lack that capacity. I like the idea of degrading multi-domain proteins as shared. But I really like extending this into undruggable space, and the proteins that we have made degraders for into human clinical investigations so far are really proteins that I think you'd have a hard time inhibiting or degrading without this new mode of molecular glue chemistry. There are a lot of challenges, and I phrase these as opportunities, because in an academic audience, I think if you can reduce any of these to facile practice, It's just like the rising tide that makes everything better. In particular, and Eric, I hope you're on this call, Radek, please take it back. We need better multiplexing capacity in proteomics, like yesterday. Frequently biology is the slow step in drug hunting. Sometimes it's chemistry, but here it's protein sciences. In closing, I want to give you a couple of snapshots. Intermolecular glues go well beyond protein degradation. For spinal muscular atrophy, just a horrific neurodegenerative disease that is lethal in children with SMA Type 1, for which at the time there was no available therapy. Scientists at NIBR considered that there could be small molecules that would lead to exon 7 retention in SMN2, unaltered in spinal muscular atrophy, that would compensate with more productive protein expression from an exon 7 retained mRNA, compensating for SMN1 loss. And using a series of really elegant signal-on phenotypic screening assays, discovered in millions of compounds studied a small number of small molecules with consistent core pharmacophores that led to SMN2 enhanced exon retention and protein production. And this ultimately led to the just elegant medicinal chemistry. Believe me, each of these arrows have thousands of of associate hours embedded to LMIO70, the first ever small molecule that in a sequence-specific manner corrects exon 7 splicing. And we've characterized now this molecule exhaustively. And this molecule, believe it or not, does not bind to a protein in a demonstrable way. Ligand affinity chromatography was dead negative, but rather has a sequence-specific stabilization of the pre-spliced RNA on the U1 snurp. It's a molecular glue that only binds its protein target in the presence of this RNA, stabilizing the RNA. This is now a structural model leading to this restorative protein consequence. This molecule is in human clinical investigation. Molecules inspired by this by PTC and Roche are now seeking regulatory approval for SMA and have it in some jurisdictions. This medicine has reversed the trend in early clinical studies of decline of the hallmark clinical characteristics of SMA, actually leading to milestones achieved by these children. We've recently characterized the consequences of LMA's action across all of the mRNAs and splicing events in human biology that we can. And it's exquisitely selective. We have found remarkably, and this is unpublished data submitted for publication, but I've not previously shared it, that LMIO70 in the way that it stabilizes the pre-spliced mRNA of SMN2, it actually destabilizes mutant Huntington transcript., and in model cell lines that have been workhorses for us for Huntington's disease drug discovery, and in patient-derived fibroblasts, we have observed a decrease in mutant Huntington RNA stability and expression. We have now validated this in patient-derived samples from the spinal muscular atrophy study, and this medicine is entering into advanced clinical trials as potentially oral therapy for Huntington suppression in that horrible disease. I didn't have time to talk about intramolecular glues, but our best example I did share at a grand rounds at the Farber previously. Here, protein phosphatases have never been successfully approached by discovery chemistry, and we have reported and innovated the very first ever bona fide phosphatase inhibitor as a therapeutic product. Here we target SHP2 because it is a a requisite signaling feature, a tyrosine phosphatase, and so many oncogenic tyrosine kinase pathways. And it's also an important facilitator collaborator of KRAS signaling, in particular KRAS G12C. We and others have made very potent inhibitors of phosphatases by looking at only the singular phosphatase domain. The problem is the anionic nature of those molecules binding to the cationic pocket that recognizes phosphates don't make for great medicines. And so we reconsidered this with the full ensemble of domains in a high-throughput screen that threw out all phosphatase-directed inhibitors, but retained small molecules that bound— that worked only in the context of all three domains, the protein tyrosine phosphatase domain, as well as the two SH2 regulatory domains. Together with Mike Eck and our own internal scientists, we've characterized these molecules They are intramolecular glues that function by bridging a coordinated lattice of electrostatic interactions at an allosteric site on the phosphatase domain, but binds only in the presence of the two SH2 domains. And through really elegant ligand-efficient lead optimization, we have matured this aminopyrimidine core to optimize pi-stacking interactions, in one case, that are critical to this chlorophenyl ring, This eastern papyrazine was thought not to be pivotal. We, as you can see, added this extra cyclic basic amine to pick up an order of magnitude of activity, and ultimately spun the nitrogens around the heterocycle to optimize a pi-cationic interaction with arginine-111. That's a long-winded way of saying, really first-world med chem that led to a very potent and active lead molecule. How does it work? This is a simulation. From a morph structure, but the molecule binds directly into this central channel and ultimately bridges all three domains. It's non-saturable. This is an intramolecular glue. It just happens to be the proteins recruited are within a singular open reading frame. Its binding leads to invasion of this green loop into the active site that ultimately blocks catalytic function. And here you can see those bridging electrostatic interactions. It's because of this exquisite pharmacology that it's so selective for SHIP-2 and not SHIP-1, where a more open pore, as shown in the lower middle of this panel, leads to water just flushing this molecule out of the active site. We completed medicinal chemistry. This is just a masterclass paper, if you should read it in J. Med. Chem., ultimately realizing TNO-155 that is advancing alone and in combination clinical trials. Data Farber is a very important site for this clinical investigation. I want to give you one sneak peek of data I've not presented previously of where this molecule might have huge clinical impact. You might know that first covalent molecules have been developed now for the G12C allele of KRAS by at least 6 companies, ours and others, and that these molecules are heading into clinical investigation as single agents, And that Amgen has recently approached a point where they might receive full regulatory consideration with a high objective response rate as a single agent. Unfortunately, these patients are not cured. And so for this, as for other targeted therapies, we need a mechanistic basis for combination agents. And published studies and unpublished studies have converged on SHIP2 as being important for G12C cycling and an opportunity for combination therapy that could anticipate evasive resistance. This study is open and enrolling, but recently collaborators of ours at Mirati, who've innovated a G12C inhibitor, presented this clinical experience of a patient, a 53-year-old smoker who was diagnosed with non-small cell lung carcinoma, who received the Amgen KRAS G12C inhibitor, progressed, was enrolled on the Mirati G12c inhibitor in combination with our SHIP-2 drug. And at active dosing levels of both medicines, this patient has had an extraordinary response. It's a maturing story. We're so hopeful that this medicine will be well tolerated and it will be consistently active in this disease. And we appreciate the collaboration with the Dana-Farber to move this concept of G12c shift to therapy along. In closing, we may be at the end of undruggable. I hate that word. Maybe intractable is a little bit better. But listen, if you can drug RNA in a sequence-specific way with small molecule glues that recruit the U1 snurp, both to stabilize and destabilize transcripts, then nothing's undruggable. It's just on us. To reimagine what it means to be a therapeutic and its likeness. And anyway, thanks for the invitation to join. I'll take any questions if you have them. It's wonderful to be back here at Dana-Farber. Believe it or not, 5 years now, my anniversary at NIBR. And thanks to all of you for your heroic work taking care of cancer patients during the impossible times of this pandemic. Take any questions. Thank you.