Enabling On-Demand Therapies

Earlier this year, the company gained attention when it reported that developed a customized antisense oligonucleotide to treat a boy with an ultra-rare neurodevelopmental disorder in a year’s time. The company is leveraging AI to develop oligonucleotide medicines on demand. We spoke to Chris Hart, co-founder, president and CEO of Creyon Bio; about the proof-of-concept achieved with its recent N-of-1 therapy, the business model for Creyon, and the potential for its approach to reduce the time and cost of drug development.

Daniel Levine: Chris, thanks for joining us.

Chris Hart: Thanks for having me, Danny. I’m excited to be here.

Daniel Levine: We’re going to talk about RNA-based medicines, Creyon’s AI-based platform, and its recent success at delivering a customized antisense oligonucleotide therapy to a single patient diagnosed with a rare neurodevelopmental disorder. A patient advocate and mother of a boy who was one of only two known people with a specific neurodevelopmental condition was recently featured on the Rarecast. Creyon successfully developed an experimental therapy for her son Leo and treated him within a year of initiating the work. How did Creyon become engaged in this?

Chris Hart: Yeah, great starting point. We crossed paths through multiple mutual contacts actually, and as I’m sure she mentioned, she found us about seven months after she had received her diagnosis for Leo. And Leo was, and I’m sure Yiwei mentioned this as well, but he’s suffering from a very rare disorder, extremely rare genetic aberration that results in one of the two copies of his TNPO2 gene to harbor a single nucleotide change that ultimately distorts the function of the protein that’s expressed off of that genetic locus. And what we’ve been building, Creyon is really a platform that started at the beginning really to expediate the creation of a class of medicines, oligonucleotide-based medicines. These are antisense oligos, RNAs and the like. We’ll talk more about it I’m sure in a moment. But she found us just as we were putting the final touches on our first set of models that allow us to really transition from a place where traditional methods require you to screen and use trial and error methods to discover these types of drugs. And what we’ve pushed forward is a set of technologies allowing us to truly engineer compounds that will have a high chance of being safe and effective at controlling gene expression. And when Yiwei found us, Leo was just, I think, nine months old, seven months into his diagnosis, and the medics and molecular pathology around his disease was really well suited for something that we thought we could address. So, at that point, we leaned in and started working together.

Daniel Levine: How well understood is the condition?

Chris Hart: Like many rare diseases TNPO2 disease as a clinical syndrome or disorder is actually not particularly well understood. There’s not many people, we don’t really understand the full spectrum of the natural history or how this presents itself. But what is better understood is that the genetics and molecular biology around the TNPO2 gene and the protein it codes for is actually fairly well understood, or better understood than the disease [that] presents itself in the clinic. And what’s understood about it, and this is what’s important for what allowed us to move forward really, is that what’s understood is that the TNPO2 gene itself codes for a protein called Transport 2. Transport 2 is expressed in many cells, but it plays a very significant role in neurons. And the role it plays in neurons is to keep things in and out of the nucleus of the neuronal cell. And the things that it keeps in and out of the nucleus are largely signaling molecules. And what was pretty clear from the genetics of Leo’s disorder was that the variant that he had occurred in his genome and one of his alleles was actually causing one of those two copies to effectively be dysfunctional. And then things were going in and out of the nucleus causing those cells to be unable to control those signaling processes and ultimately leads to neuronal stress. So, it’s sort of like an analogy I sometimes like to use—if that gene was making tires, it’d be like a tire factory making 50 percent of the tires bad. And if you sent those tires off to a car manufacturer, every car coming off the assembly line would be broken. But if you could somehow do QC on those tires and stop the bad ones from going into the factory, you’d have a chance of making functional cars. And in the same way, what was understood and what’s understood about the pathology there is that the TNPO2 gene is critical for this control of signaling in the nucleus of these cells. And if we could get rid of the bad variant, the bad copy of this protein, we’d be able to allow the functional, the reference protein to do its job and likely have significant benefit in terms of making neurons healthier.

Daniel Levine: I suspect there are potentially thousands of diseases for which you could have been engaged to produce such a treatment. Why did you pursue this one and what are the range of things you need to be able to do to make a go-no-go decision?

Chris Hart: Yeah, I think there are thousands of genes. I think Francis Collins wrote this in one of his op-eds about a decade after having led the public sector’s genome project. The hope of the genome revolution was really the ability to unravel the molecular and genetic underlying aces for pathogenesis of disease. He might’ve said it with a little bit more eloquence, but the bottom line is there’s a lot of opportunities out there. And where is there a gap? The gap is how do we translate that growing understanding of disease biology to new drugs? That’s what Creyon has been focused on building. And in this particular case, what were the go-no-go decisions? Well, as I said, we understood the genetics around Leo’s particular variant that was driving his disease biology. We understood the potential mechanisms that we could use to correct that aberration, and it was compatible with the types of drugs that we could make with the platform we developed. The second thing that made us comfortable moving forward is that we knew that the phenotypes that were emerging in Leo were consistent with largely neuronal dysfunction. And we also know that our drug can get into neurons when intrathecally the delivered into the patient. So, with that sense, we knew we could get a drug that can control the genetics and the molecular biology of the pathogenesis. We knew we could get the drug into the right cells. And then the last aspect of it is that we also knew that doing this would likely be well tolerated in terms of controlling the gene expression at the target level because there are people in the world, and we know this from the population genetic databases that exist that only have one copy of TNPO2. So, the likelihood that by getting rid of just one bad copy, the good copy would be inadequate to maintain health was low. We knew that that would probably be an okay situation to get to because there are people that only have one functional copy. And then the platform allows us to have confidence that we’ll make a safe drug. And with those things lined up, the last thing you want to ask yourself before you pursue a new drug indication like this is if we did make a safe and effective drug and we deliver it to the patient, would we be able to tell that we were having benefit? And in Leo’s case, he was starting to, and it got more severe over the time course in which we were developing the drug. He was having seizures that were pretty significant. Often they were apneic seizures where he was stopping breathing and he was starting to lose developmental milestones. Both of those sets of observations were things that we thought we could show benefit for if we had an effective drug. So, when those all lined up, we said, yes, this is a feasible project for us to push forward on.

Daniel Levine: Well, walk us through the process. How do you go from having a gene that’s responsible for a condition identified to actually producing a therapy that you can dose the patient with?

Chris Hart: Yeah, so what I’ll say is this process is—I’ll walk through it at a high level. I’m happy to dive into as much details and you may have follow-up questions. But the general process that we followed for Leo is actually not that different than it would be for any other drug development program. Candidly, whether it’s an ultra-rare indication like what we’ve developed for Leo or for a common disorder, at the end of the day, what do we need to know? First, we need to understand what is the underlying molecular hypothesis in terms of how are we going to correct the pathology that the disease is causing? In Leo’s case, we knew exactly what this was. He was suffering from literally a single nucleotide change in his genome that he carried on one of his two copies of the genome because we have two copies, one from your mom and one from your dad. So he only has one mistake out of 6 billion letters, but we knew exactly which one it was, and we knew that we needed to control the expression of the RNA that harbored that variant that would then ultimately lead to a misfunctioning protein. So, once we had that sort of clarity in terms of the molecular and genetic basis of this disease, we then could move forward and say, how do we test that? We needed to make sure we had in vitro systems to ensure that we could. In vivo systems would be reasonable to do, but creating an in vivo model for an ultra-rare disease like this would take years. We didn’t have years, so we said, what’s the next best thing? Let’s use these cells from Leo where we know that the genetics that are going to recapitulate in the cells and then we can demonstrate the drugs are going to work there, and then we had to make the drug. When we made the drug, we were able to leverage the technology that we developed to go very quickly, and we can go a little bit more into this in a moment, but effectively what we leveraged was a platform that allowed us to look at the variant that we wanted to target to make the RNA get degraded. We took that sequence and then we ran it through our platform to then engineer around that what would be the optimal chemical modifications, given that sequence, that would assure that we could still engage with the enzyme systems that exist in the cell to degrade the RNA. This happens to be RNase H1. What are the chemical modifications that we can put in there that would retain the safety of that drug? And then once we had that design set, we then pushed them forward and started to test those cells to ensure that they were doing what they were supposed to be doing. We then put them through their paces in terms of the safety, pharmacology. We wanted to make sure that the drug was going to be well-tolerated both in in vitro systems but also in in vivo systems. We used animal models to ensure safety of these compounds. Once we had taken the first set of compounds that we engineered to advance forward, we identified out of that first batch a lead that was exceptionally well tolerated and exceptionally allele selective, we then could push forward and we then followed the regulatory process of scaling that compound up for what’s called good manufacturing practice. We got GMP product. We then put it into good laboratory practice studies to ensure the safety of the compounds. And then that allowed us to file ultimately for an investigation on the drug that we could then …

Daniel Levine: When you know the protein, it’s not too hard to generate an oligonucleotide. You made reference to the chemical modifications, and I take it with these therapies, the chemical modifications are really an essential part. Can you talk about what those modifications do and how you use an AI-based platform to determine the proper ones?

Chris Hart: Yeah, great. So, one of the amazing things about oligos as a drug substrate and really oli nucleutide based medicines have emerged as this super modality. We have antisense oligos, RNAs, guide RNAs, aptamers. These are all effectively just polymeric nucleic acids that we leverage to impart some sort of biological effect as drugs. I’m going to focus on ASOs or antisense oligos first. Antisense oligos is the type of drug that we made for Leo. This is a modality that was first described in the late seventies actually. So, we’re now 40 years plus in the history of development of antisense. And there are now multiple approved products that are antisense oligos. And what’s clear is if you’re going to make an antisense oligo, you have to turn these short DNA polymers into something that have reasonable drug-like properties. That’s what the chemical modifications do, and those chemical modifications without having a whiteboard and a graph of the molecule. But you can see that the DNA has effectively, it’s a polymer that’s composed of each one of these nucleotide monomers. On each one of the monomers, effectively, there’s a backbone that connects it to the next monomer. There’s a sugar modification that is really there just to connect to the backbone—a base. And the base is what allows us to integrate with the genetic information of a cell. In this case, we’re using RNA molecules, but the basis for the Watson Crick Franklin hybrids-based diseases. When we chemically modify these drugs, we will change the backbone sometimes. We’ll place some of the oxygens on the phosphodiester backbone to sulfur. Sometimes we’ll change the sugars where we have a ribose or a deoxyribose on a DNA molecule. We’ll put different types of chemical modifications in there and occasionally we’ll even modify bases. What emerged over the last couple of decades in trying to make these drugs is that drugs, you can have the same chemical modification pattern and put them on multiple different sequences, and some of those drugs will be exceptionally well-tolerated, whereas other of those drugs will be terribly toxic. So, the conundrum was always, how do I get a better chemistry that doesn’t have this distribution? And what we recognized was that it’s not a question about how do I get a better chemistry? Generally, the question is how do I ensure I choose the right chemistry for any given sequence? Because the bases themselves are so close to these chemical modifications from a physical point of view that they actually interact. And depending on the bases that are on the oligo, which is what you use to target for particular sequences, you’ll impart very different properties on the molecules by changing the different chemical aspects of it. And by assuming that you can separate them completely, you’ll never win because it’s a gamble. And what we’ve realized is that you have to optimize the chemistry with any given sequence, but that ends up being a really hard problem. So, what we recognized was that’s a hard problem, and it’s hard because it’s a combinatorial problem. There’s a lot of different combinations. You take any 16 mer sequence, that’s just A, C, Ts and Gs. You’ve got 4 billion, you add in the different chemical modifications, you get to astronomical numbers, 10 to the 20, 10 to the 30 possible permutations. And when you start thinking about these combinatorial problems, you really end up in a place where you have to learn how to navigate this space efficiently if you’re going to understand how interactions happen. And that’s what we did at Creyon. The first thing we innovated on was how to create the right data set. And then we were able to then use that data to train AI and machine learning models that would connect and allow us to have this engineering clarity, if you will. So given a particular sequence, what would be the optimal chemistry to leverage to ensure it’s safe. What’s interesting though is that by solving this problem of safety in terms of what chemistries you put on which sequences, we’ve actually done something that I think is pretty unique in the industry at large. Most drug development efforts, whether it’s an oligo or small molecule, an antibody, the first thing you start with is how do I find something that’s active and you find an active molecule. And that’s usually a process. Serendipity, throughput, laborious trial and error SAR processes typically to a place where you have something that’s safe and well enough to go into the clinic. Oligoes, you don’t really have this serendipitous problem with finding things that’ll interact with your RNA molecule, as we talked about, it’s just Watson Crick Franklin. But what has been the stumbling block in industries, how do we turn that thing that binds to an RNA molecule, and how do we put chemical modifications on it so that it retains its safety as we push it forward? And by solving that, and that’s really where AI has come to help in this program also help across different programs, is that expedience. How do you go from a target of interest to a drug that’s going to be safe and effective moving forward into the clinic?

Daniel Levine: You’re not the first to do an N-of-1, but you’re certainly in a rarefied group of people that have succeeded at doing this. How much of a challenge was FDA in the process and did the folks who went before you help in any way?

Chris Hart: By all means, I think we stand on the shoulder of giants in terms of what we’re building in the field. As I mentioned before, oligo medicines is something that’s been building up over the last four decades and N-of-1 treatments have come forward. I think Tim Yu with his first case with milasen really proved the point that it was possible. And in response to that, the agency has been very forward thinking on it, and they put forward the FDA guidance on N-of-1 ASO treatments, which really actually made it pretty straightforward to go from: we have a compound, what does the path look like to go forward in terms of making a drug that we can dose into patients? And in that sense, by all means, having others [that] have gone through this path before certainly was helpful. What I’ll say is that that N-of-1 guidance, although quite helpful in clarifying, does have some limitations that manifest themselves, especially for commercial sponsors like Creyon. So, in the way that it’s drafted, really the new drug applications, the investigational new drug has to be filed as an investigator initiated IND. That makes it a little challenging because now we’re fighting both wit—I shouldn’t say fighting. We’re working both with the FDA to get this drug approved for dosing, but we only can do that after we find a clinical sponsor investigator who’s willing to take on the responsibility and the burden of paperwork and the burden of treating this family and this individual for the rest of lives, potentially a significant ask to push on people. And additionally, the guidance right now really applies on a single drug at a time basis. And there’s no real room right now to expediate aspects of the development process that may be consistent from one drug to the next. And now I’ll point to something that’s obvious, which is, for instance, ASOs, when we think about the stability of these molecules in test tubes or in dosing syringes on the shelf, it’s a pretty well understood thing. We know how these molecules generally behave, and the agency still requires for us to do stability testing of these compounds. And that requires tus to make extra—we have to manufacture extra drug, we have to make extra. We have to set up the studies and we have to maintain a stability study going forward for each one of these drugs, even though there’s probably very little variance across the different types of ASOs you might make, this is something that we understand. So, we think there’s places where the agency could probably make some efficiencies build up, and we think there’s probably even broader things in the long term that you could think about building off on. Imagine our modeling gets even better, where when I engineer a compound, I predict out its safety as well as animals do, right? If I can tell you before I put in an animal if that drug’s going to be safe or toxic, how many times do I have to do that? If I’ve created the right data and validated the right data for the bound of compounds that I’m making before I can push forward and foundationally reduce some other purposes of bringing these drugs, we think there’s a path there. We think we have to create the right data work, and I think there’s probably a movement afoot associated with the platform guidances that come before and otherwise that will in the long term allow us to make things even more efficient.

Daniel Levine: So, you did this from initiation of the project to dosing in a year. How much do you think you can improve on that given these potential efficiencies?

Chris Hart: Long-term visions? Short-term, what I’ll say is that we went from whiteboard to a lead compound in about five months, the remaining time from five months to a dosing time about a year later, right? Well, 13 months in total with all the regulatory requirements, both from GMP and GLP studies and all these things. And some of those are by all means, I think, completely reasonable. Some of them are, but we can certainly think about shortening that quite a bit. I think we could probably hasten the timeframe in which we get it. So, in the short term, I think we have paths without any regulatory shifts to accelerate things by months. As we look forward though, as we think about ways in which we can work with regulators and create the right data to build confidence both in ourselves and in regulators and treating physicians and families, that the platform we’re developing is capable of engineering with clarity that the drugs we’re building are safe and efficacious at a molecular level and have proven out to be unlikely to cause any harm. I think there’s ways in which we can think about accelerating that even and even more to, effectively, manufacturing. But that’s a longer term. I don’t want to promise that’s tomorrow, but that’s a forward thinking vision. ,

Daniel Levine: Leosen, which is named in honor of Leo, is an antisense oligonucleotide. As we mentioned earlier, you alluded to the fact that Creyon works across oligonucleotide based medicines. Why an antisense therapy rather than some other type of oligo approach, and how do you determine the best approach for a given disease?

Chris Hart: Yeah, for Leo, the choice was pretty simple. We needed to have an allele selective compound that we could dose into Leo intrathecally to get into neurons. There’s really only one modality that can do that right now. That’s antisense oligos. Nothing else has anything to stand on. So that is the only path that we could think about going forward with Leo and is the perfect fit for what we needed there. For other disorders, as we look at it, this is something we get asked quite frequently, which is the variant that I’m trying to treat amenable to oligo therapies and which and what would be the right way of going forward. And there’s unfortunately not a sort of simple answer because biology is complicated and depending on the gene locus and exactly how different variants are manifesting themselves into dysregulation of the gene or functional aberration of the ultimate protein product, different approaches are appropriate. And it’s kind of a case by case basis in many instances where you have to look at the gene, the disease, the regulatory landscape around it, and then we can think about how do we intervene with an oligo approach where we can have an impact to correct whatever it’s driving, the underlying pathogenesis.

Daniel Levine: And what’s known about the safety and efficacy of leosen from what you’ve seen to date?

Chris Hart: So obviously Yiwei and family are in the best position to speak to how Leo’s doing, but we’re excited by the reports that we see. What I can say is that the drug that’s been dosed, our first drug administration was July of last year. So, he’s been on the drug for over a year now, and it remains perhaps the safest intrathecal developed drug that we’ve seen. We see no signs of hydrocephalus or any sign of white cells or protein in the CSF when we look and critically, we’re also quite excited because he seems to be having positive responses to the drug. We’ve seen a precipitous drop in seizure frequencies. Often those seizures are milder, and it’s exciting to see the milestones that he’s regained. He’s able to sit upright, which he wasn’t able to do, still assisted, but sit upright. He’s starting to be able to reach the thing, have volition. What’s really quite clear is that he’s able to engage in a world that he was previously starting to slip away from. And I think that’s a testament to, really, the power of treating the underlying molecular pathology of disease, and all of those are a tremendous way of being able to go in and have these sorts of really amazing, amazing results.

Daniel Levine: This is a big proof of concept for what you’ve been building at Creyon. How do you scale this now?

Chris Hart: Yeah, that’s something we think about all the time. And we started Creyon really with a mission, which is to make gene centric medicines available to patients globally. And the way that you scale to meet that demand is, we think, to systematically look at where there’s inefficiencies in the drug development process and chip away at them. And the first piece that we looked at when we started Creyon was, and oligos were a perfect substrate to start with in terms of building out in this realm, is that we understand for many diseases, both rare and common, what is the genetic pathogenesis that we could correct either by controlling gene expression or changing the way genes are regulated, splicing or otherwise. But transforming that understanding of disease pathology to new drugs is still this herculean process, and it often fails. There’s really two reasons that drugs fail. One is that you had the wrong biological hypothesis. It just turns out that you’re barking up the wrong tree for whatever reason you thought this gene would be important in disease biology, you control it, and it doesn’t work. Well, I guess there’s another reason. You created a drug that actually doesn’t test that hypothesis or your drug is toxic. Those three things need to be resolved. So, all the issues around getting something that works in terms of targeting a particular, right? We can test very exquisitely whether or not an oligo is actually having the molecular impact intended to do so, and we can make sure that it gets into the right cells and tissues through a variety of mechanisms. That’s another aspect of what Creyon has been working on. And then the safety aspect, though, is something that was an unsolved problem in the field and was being solved by effective laborious trial and error screening. So, our first thing was how do we engineer that safety problem because we can push everything else forward as that problem is resolved, that will ultimately lower the cost and time and allow us to approach more and more indications across the spectrum. And that’s really what we’re focused on, is how do we continue to make progress solving this engineering problem so we systematically road the inefficiencies that exist in the drug development process.

Daniel Levine: You’ve shown you can do an N-of-1 medicine, but that’s not necessarily the business model. You’ve said in the past that you’re agnostic to the size of a patient population. Talk about what Creyon’s business model is.

Chris Hart: Yeah, so I think the short end of our business model is pretty straightforward. We built a platform where we think we could create best in class oligonucleotide based medicines, and we want to sell them so that we can make money making drugs that address unmet medical needs and have true patient benefit. We see this as a spectrum where our case with Leo demonstrates that the platform works all the way down to literally N-of-1. And then we also have efforts both internally and through partnerships where we’re developing drugs for common disorders or what we sometimes refer to as larger or commercial rare programs that fit within the standard models of biopharma. So, our ultimate goal is to get to a place where we can develop these drugs with such efficiencies that we can make all the drugs that doesn’t matter what patient population size is, but the goal is to make sure that at the end of the day they can be prescribed and then they can be paid for by payers or whatever. The healthcare system that we live with can afford to do so for common diseases and commercial rare indications, a straightforward tried and true path. As we think about these ultra-rare indications where we’re developing out novel therapeutics that need to be created, we need to work forward and think, how do we get to a place where those compounds can be effectively paid for and prescribed within the healthcare system that we all live with today as standard of care? Because we know they can have huge benefit. And we think there’s a path of doing that by ensuring that we can demonstrate that we’re creating safe and effective compounds and creating the right data to ensure that we have an expediated path towards effective approval and payer support.

Daniel Levine: You mentioned partnerships. How does partnering fit into your greater strategy?

Chris Hart: Yeah, so when we first started Creyon, one of our lead VCs, DCVC Bio and their partner at DCVC, I was talking to Matt, the founder of DCVC, and he described our platform as fitting in with his thesis around creating a radical abundance. He invests in things that—I should back up and say he invests in things in the StarTrek future versus the Star Wars future. And the StarTrek future is really across many things really marked by a plethora of radical abundance as a solution to many of the world’s problems. And where we think what we’re foundationally working on is building out a radical abundance of gene centric medicines in that sense. But what that puts us in right now is in a place where we can create more compounds than we can possibly develop. So, partnering is a way for us to ensure that we can use the platform to address unmet need in an efficient manner across more than we could do ourselves. Additionally, we’re excited to work with outside teams where they have novel target ideas, novel approaches to clinical development, and, candidly, expertise that we can learn from as we grow Creyon. So, partnering is a critical way for us to expanding the reach of Creyon, working with teams where they have ways in which they can leverage our technology to address unmet need that we wouldn’t be able to do on our own, bring in that additional expertise. And of course, when we’re successful, we’re excited to share in the revenue that would support us as well.

Daniel Levine: And is there a pipeline strategy or are you building your own pipeline?

Chris Hart: We are building a balanced pipeline between both internal assets and partnered programs. What I haven’t discussed is that we’ve also developed out an ancillary set of technologies that build on top of our core competency in really engineering polymer nucleic acids around aptamer technology. Aptamers are effectively just longer nucleic acids that fold and then bind to proteins. What we recognized is that we’re not the only ones that recognize it, but there’s two problems in the oligo world. One is, I think I talked about it. One is safety. The second one, which we haven’t talked much about, is getting them into the right tissues. You can get away with—for severe neurological conditions like Leo has—with intrathecal delivery into the brain. They can do these local administrations. We have very good technologies to get all of those into the liver. We don’t have technologies to get into any cell we want. They just don’t penetrate deeply into a lot of tissue beds. So, the second challenge that we’ve been working on solving, and we have quite a bit of progress on, is around using aptamers to bind to receptors on cells that will then allow us to drag our oligos into various different tissue beds. So a lot of our pipeline is developing around places where we have unique opportunities to leverage our capacity to create best in class oligos and couple that to a differentiated delivery platform that allows us to get into tissues that others can’t practice pharmacology in. That provides us an immediate place to go where we can attack well understood biology, but it still sits in a realm of truly unmet need. So, we can go very quickly to creating meaningful drugs for patients across the board.

Daniel Levine: Chris Hart, co-founder, CEO, and president of Creyon Bio. Chris, thanks as always.

Chris Hart: Thanks so much, Danny. Really enjoyed the conversation.

This transcript has been lightly edited for clarity and readability.

The RARECast podcast is made possible through support from the Global Genes’ Corporate Alliance. The members of the Corporate Alliance support Global Genes’ mission and programs, work to meet the vital needs of people with rare diseases, and address inequities they face. To learn more about the Corporate Alliance or how your organization can become a member, click here.