My name is Ted Tenthoff. I'm a senior biotech analyst at Piper Sandler. Before I begin, I'm required to point out certain disclosures regarding the relationship between Piper and our next presenting company, Gritstone Bio. Gritstone Bio is developing next-gen cancer and infectious disease vaccines. Gritstone Bio is conducting a phase II/III trial of a personal cancer vaccine, GRANITE, in microsatellite stable colorectal cancer, with preliminary data early next year. The company is also developing an off-the-shelf cancer vaccine called SLATE, and just received a $433 million BARDA contract to conduct a phase I study of next-gen COVID vaccine CORAL. Phase II. Phase II, sorry. Sorry. I wrote phase I, so I apologize. Phase II of CORAL. With us here from Gritstone is Co-founder, President, and CEO, Dr. Andrew Allen. Andrew, always nice to see you. Thank you, Ted. So maybe you can start off by describing your EDGE technology, because I really believe that's core to selecting the antigens that you're including in your vaccines. How does it work? How do you use it? Yeah, it's a rich question. So it is absolutely core to any neoantigen vaccine, is identifying the neoantigens correctly. And lots of people have taken different approaches here. We've always felt that it's important to make a product that can be used in the community by the average community oncologist. So there are lots of interesting science experiments out there about collecting T-cells and tumor cells from blood and live cells and so on and so forth, but none of those will ever, I think, make their way into a real product. So we constrained ourselves from the outset, saying that we have to be able to predict neoantigens based on a standard, routine, formal in- fixed paraffin-embedded tumor biopsy, a needle biopsy, the kind of thing you just send off to the path lab for regular histochemistry analysis. So when you do that, that means you're going to get DNA and RNA data out. So that was the constraint we put on ourselves, right? How can we predict just from DNA and RNA data? And what we did was to leverage a technique developed in the nineties, where what you're interested in, obviously, are the targets on the surface of the tumor cells. These are HLA peptide complexes, that is what are recognized by T cells. And to train a model, we decided to start collecting real human tumor samples at scale, so hundreds and now over 1,000 different samples from patients. And the technique we use is one whereby you take a fresh sample, you pulverize it, you immunoprecipitate all of the Class I HLA molecules, and then you identify the peptides on those Class I HLA using mass spec. So now you've got a library of peptides from a human tumor sample, all of which were processed and presented. That is incredibly valuable data because it tells you these are peptides that are indeed processed and presented. What you then do is take the sequence data from that same sample, the DNA and RNA data, and you let machine learning mathematics figure out the genomic features that predict whether a particular peptide will or will not make it to the tumor cell surface. We've now started using the large language models that obviously everybody's been reading about. This prediction technology works very well. We're now predicting with over 80% positive predictive value. Today, of course, that means that we don't have to do any of that live cell HLA work or anything. All we need to do is get a routine core needle biopsy, sequence it, plug the sequence data into our prediction model, and it will spit out the top 20 candidate neoantigens for that patient. We now use those in our personalized cancer vaccine. Yeah. So that's kind of how the whole system works. You validate it by showing that... We did some initial validation work. We published this in Nature Biotech a few years ago, using some Steven Rosenberg data from TIL, because that is a very good data set. What's good about it from a validation point of view is it's completely orthogonal, meaning he derives his data in a way that has nothing to do with prediction or vaccination or anything. It's all about T cells and expansion and proliferation of T cells. But of course, if we can show that we're able to essentially mirror his data using our prediction technique, that is really strong validation, and that's what got our manuscript into Nature Biotech. Then since then, of course, what we've shown in patients is the thing that really matters, that is the best validation, is if I vaccinate you with our predicted neoantigens, you make an immune response that leads to clinical benefit. Mm. At the end of the day, that's actually what really counts. Yeah. And so people ask us about how do I compare with you and Moderna and BioNTech, and the answer is you can't. Right. All you can do is look at clinical data. Agreed. Now, there is another big difference in your approach, which is you really deliver these selected neoantigens via a heterologous prime-boost vaccine, which is comprised of a chimpanzee adenovirus as well as a self-amplifying mRNA. We just saw, I think one of the first samRNAs approved maybe in Japan as a vaccine. Yep. Describe why—maybe first describe these two components. Why does it make sense to combine these to generate a, an optimal response? Yeah. So there's a very interesting kind of strategy, question embedded here, which is around... If you step back, in 2014, in November, online in the New England Journal, was published a seminal bit of work from Tim Chan at Memorial Sloan Kettering, one of the Gritstone co-founders. He wasn't at the time.... became one for that very reason. And what Tim showed is that in humans responding to checkpoint inhibitors, the biology that underlies that clinical response appeared to be neoantigen-specific T cells, particularly CD8 T cells. Those were data that Tim surfaced at that early time point, and that was the first time in humans that we've really seen evidence that these tumor neoantigens were acting as critical targets for checkpoint responsive patients. So that was, I think, the innovation catalyst. And two types of company then glom onto that catalyst. There are companies that have an existing technology that they think they can apply to the problem. Enter Moderna and BioNTech, because they had mRNA vaccines that could obviously be made quickly, and speed obviously is important for what we're talking about here. You have to make these patient-specific products. So they entered the field using their mRNA technologies. It's important to note, they weren't choosing their mRNAs because they were great at driving CD8 T cells. They were choosing them because that's what they did. Essentially, you know, they're hammers and everything's a nail, so we'll try our mRNA technology. It's new. Let's try it here. So it makes sense. Then there are different companies like us that said, "That's a problem that we can solve. Let's build the best possible set of tools to solve that problem." We framed the problem as: how do you create neoantigen-specific CD8 T cells in humans? And therefore, we looked at the entire literature of vaccines, all of the data, and spoke to a lot of people because we had no pre-specified platform. Right. We could choose the best, and interestingly, the best in humans is adenovirus for priming CD8 T cells. And I'm using those words carefully. In humans, priming and CD8- CD8, yeah. 'Cause lots of stuff works in mouse. Doesn't matter, 'cause they don't replicate in humans. Priming is immunologically hard to do, and CD8s are really hard to prime. Right. So this is not a property that most vaccines have. Peptides do not do this. Right. That's why a lot of the literature filled with peptide data from academics over the decades, ultimately, I think will never work- For sure. ... because it just doesn't prime a strong enough T cell response. So we chose adenovirus as the prime, and of course, it's not cool these days, and it has this rare toxicity of 1 in 100,000 or something to cause the vaccine-induced thrombotic thrombocytopenia. But in the context of cancer, that obviously is not a concern. These are people who are going to die 100%, and so nobody really wants to hear the answer, adenovirus. So the populist in me always wants to focus on the RNA, the self-amplifying mRNA- Yeah. ... 'cause it's trendy and everybody's excited about it. But the truth is that the adenovirus is actually doing the heavy lifting- Yeah. ... that primes the CD8. But our belief is you want to boost those, CD8 T cells, and the challenge with adenovirus is you can't use them close together. Right. You can't boost consecutively because you make good antibodies to the virus itself that will neutralize the second dose. Sure. So if you prime and then you quickly boost with the same adenovirus, the second dose doesn't do anything because it's functionally utilized by adenovirus-specific antibodies, but the first dose induced. Therefore, you need to switch to a different vector. So you're gonna deliver the same set of antigens, but using a different vaccine vector. So we had choice over which boosting vector to use. We looked at mRNA because it, obviously, it was interesting and it was new, but the data suggested that linear mRNA, first-gen mRNA, was not very good at priming CD8s, and I think the data borne that out. It's not it doesn't work, it's just not very potent. And Moderna's data suggests maybe you need, you know, nine doses. Remember, they give nine shots every three weeks. You may need that kind of number of vaccinations to elicit a good immune response. So we chose self-amplifying mRNA based on some preclinical data suggesting it might have superior properties. I think that, in retrospect, was a good choice, not only because it does seem to boost T cells very effectively, but as we've learned in our COVID work, it has beautiful and very distinctive properties around antibody generation as well. Not why we chose it, let me be clear, that was a serendipitous finding, but we chose it for the T cells. So when you switch the vectors, that's called heterologous prime-boost, and that's what we use. Yeah. It seems to work very well. And then I thank you for going into that level of detail because I really want to build on that- Yep. Firstly, with GRANITE. So tell us a little bit about this vaccine. Remind us of the early, data that you've shown, the encouraging data to date. Yeah, thank you. So if you step back and say, "What do I want to see?" If you believe in the scientific hypothesis, as stated when we started the company in 2015, our goal is to take cold tumors that do not respond to checkpoint inhibitor therapy because they don't, those patients don't have neoantigen-reactive CD8 T cells, and we're gonna give them neoantigen-reactive CD8 T cells, and then they're gonna respond to checkpoints. That was the fundamental hypothesis. So the schedule that we use is to take a biopsy from the patient, as I mentioned. We sequence it, we predict neoantigens, and we make the personalized vaccines, and we then deliver them in a heterologous prime boost fashion. And the, the optimized regimen we now use, we give basically 6 injections spaced out over 12 months, and we get a quick response. So in the phase I/ II data in advanced solid tumors, we published this in Nature Medicine last year. When we started, and I'm gonna focus my comments on colorectal because it's where we're focused now, and it's the cleanest canvas to paint on because it, they're cold tumors and checkpoints don't work. So we've got a nice sort of clear base here.... We started treating patients in third line. So these are pretty tough patients who have a median survival of about six or seven months. We were treating them in third line because although there are some third-line drugs approved, they're really not very effective, and the agency was comfortable that we essentially substituted for those. When we started the treatments, the biopsy showed that these were cold tumors, so no T cells, no PD-L1 expression. They're low tumor mutational burden, typically less than 10. When you look at the blood, you cannot find neoantigen reactive T cells using standard assays. So these are very cold, classic tumors. We vaccinate. We then look in the blood. Are we eliciting neoantigen reactive T cells? Yes, we are, in everybody. Are they recognizing multiple neoantigens? Yes. So when we were able to disaggregate the data, we showed that typically over 10 of the 20 administered neoantigens were driving T cell responses, primarily CD8s. Okay, so that's an important box to check, that we're now driving good, strong CD8 responses, probably being primed, because they weren't there at the beginning, in all of our patients. Now, next question: Do they get into tumors? So to answer that, you have to do an on-study biopsy and go looking for T cells, and you can do that now using TCR sequencing technologies. And we showed that, yes, these T cells traffic into the tumors and they proliferate. So that's the second check. They're doing what you want them to do. The critical third question is: Are they killing tumor cells? And this is where it gets a little challenging for the orthodox-minded folks, because people are used to looking at RECIST response as a proxy for efficacy. A RECIST response is great if you're a chemotherapy agent or a targeted therapy agent. It is miserable if you're doing what we're trying to do. We are driving T cells into lesions where they proliferate. We've shown that. Therefore, expecting lesions to shrink makes no sense. Belatedly, they might, but they might not. They might just turn into lymph nodes, right? And that would be a success as well. At some level, lesion size no longer has any utility to us as a biomarker of efficacy. I want a direct metric of tumor kill. Now, some patients with colorectal will have an increased CEA or CA 19-9. These are classic protein biomarkers, and we saw that if patients did have an increased CEA, it went down half the time, so 50% response rate there. Fortunately, now we have circulating tumor DNA as a plasma-based metric of tumor mass, and more and more folks now are measuring ctDNA, and the rules of the game are very simple. There is prognostic value in your absolute level of ctDNA. The higher it is, the sicker you are, the more tumor you have, the faster you're likely to die from your cancer. It has prognostic value, but it also has predictive value. If you're on treatment, if your ctDNA goes up, that is bad. Your tumor is still growing. If it goes down, that is good. Your tumor is now shrinking. And so we measured ctDNA, and happily, we saw ctDNA responses, meaning at least a 30% reduction in half of our treated patients. So now we've got direct evidence that our T cells are being induced, they're trafficking into tumors, and they're killing tumor cells. Don't ask me why the other half don't respond. I don't know yet. Yeah. We're working on that. Good. The key issue, of course, over time, was, does that lead to other evidence of clinical benefit? We look at survival because that's what you care about. The answer was yes. Of our colorectal patients, the half that had ctDNA responses had a median survival, not reached, but in excess of 22 months. The half that didn't have molecular responses, who looked identical at baseline, had a median survival of 7 months, as expected. That's the typical outcome in third-line colorectal. Yeah. So not randomized, obviously, but what we're seeing is clear evidence that everything is lining up and behaving as it should. Yep. We launched our randomized study, which we'll come to in a second. Yeah. But the one point to make additionally is we have a SLATE program, which is an off-the-shelf KRAS product, which we took into advanced lung cancer, and we saw the same biology. Yeah. ... with a 40% molecular response rate and a doubling of overall survival in those patients- Yeah. - compared to non-responders. So when you see the same biology and the same manifestations twice- Yep. In different contexts with different products, but leveraging the same biology, to me, that's like, this is real. Yeah. This is the product doing exactly what it's designed to do. We entered our randomized phase II/III study with high confidence. So let's talk about that now. Walk us through the design of the II/III trial in microsatellite stable colorectal. What should we expect from data in the new year? Yep. So the study is a frontline study, newly diagnosed metastatic colorectal cancer, MSS type, and patients receive induction chemotherapy standard, which is usually FOLFOX plus bevacizumab, and they're randomized at the beginning of that treatment, and both arms of the trial receive the same induction chemo. Mm-hmm. Then, typically after about 5 months, patients stop oxaliplatin because of neuropathy, and they continue in the maintenance phase, which is 5-FU plus bevacizumab. So the control arm gets 5-FU Bev. The test arm gets 5-FU Bev plus our immunotherapy. The primary endpoint is molecular response, as I've said. And so what we'll have in Q1 of next year, in approximately half of the 100 randomized subjects, will be ctDNA response data. What we expect to see, conservatively, in the test arm is a 50% molecular response rate, because that's what we saw previously in third line. These are healthier patients, and we've got a slightly optimized regimen. We've learned, I think, how to use our product a little bit more, more effectively... So I think 50% is probably a conservative estimate, but let's go with that. The control arm, these are patients who may have had some reduction in ctDNA as they started chemotherapy, but by the time they enter the maintenance phase, their ctDNA will be flat and/or rising, and there is no reason to believe that there will be responses at that time point as they start their maintenance phase. The ctDNA response rate in the control arm should be no greater than 10%. We're looking for a 50 versus 10 delta. Even with 50 subjects, that is, we're highly powered to see that with statistical significance. That's the primary endpoint. Of course, over time, the OS data will mature, and then you'll start to be able to say, "Are we seeing a relationship between ctDNA response and overall survival, as we expect? Yeah. So important readouts there. We'll also get progress from the SLATE program. In the time that we have left, though, I do want to talk about your COVID program. And again, the congratulations on this $433 million contract. Tell us, you know, tell us about your COVID program first. What does this contract enable you to do? This is a contract to run a 10,000-subject randomized phase IIb study, comparing our product, our next gen Spike Plus, as we call it, COVID vaccine, versus an approved vaccine. The 1:1 randomization, 5,000 subjects each. It's a clinical endpoint study, so the primary efficacy endpoint is clinical disease, so PCR positive symptomatic disease, using now the standard FDA definition. And it is a 12-month study, because as we all know, the current products are good in the short term, but their effects are transient. And one of the key objectives of Project NextGen, there are three really objectives of that program when it comes to next gen vaccines: durable immunity, variant-proof immunity, meaning that you retain your clinical protection even if Spike keeps mutating. Thirdly, and not relevant to us, mucosal immunity, which there are a couple of interesting companies working against. We obviously focus on these first two. So we've been working with NIAID, you know, the division of the NIH that works on infectious disease, since 2021, and they were doing a phase I study in healthy volunteers of our Spike Plus vaccine. So they know the product, and obviously, BARDA, when they put out the call for submissions for this funding for a randomized phase IIb, you know, we, I think we're in a good position because, you know, we're working with the government, they know our product, and here we are, the beneficiaries of the contract, which obviously we're grateful for. And it's excellent validation of what we're doing. So the question obviously is: How do we differentiate? Yeah. So if you look at the first-gen mRNA vaccines, beautiful antibody responses, but transient. So after six months, you typically, the antibodies have faded and you're below clinically protective levels. So that's challenge number one. Challenge number two is that as Spike mutates, those antibodies that you're making using an older variant of Spike have reduced potency against new variants. So that's challenge number two. Challenge number three, you have to make quite a lot of RNA, right? 100 mcg is the, is the standard Moderna dose, as you know, for priming. With self-amplifying mRNA, we potentially solve those three problems. So first of all, when we make antibodies using our vaccine, the antibodies last for at least 12 months, maybe a lot longer. We don't have data beyond 12 months yet, but we just showed the 12-month data. It's really quite remarkable. The antibodies come up and they stay flat. They don't fade, they stay flat. I think that's because with self-amplifying mRNA, the RNA makes copies of itself and it persists in the lymphoid system much longer than first gen mRNA does, and the antigen therefore persists. There's some relationship between antigen persistence in the lymphoid system and creation of immunological memory. Great. It all makes sense, right? So anyway, antibody persistence is virtue number one. Virtue number two, current mRNA vaccines just encode Spike, and Spike changes. We have more payload capacity, so we encode Spike plus other fragments of genes from elsewhere in the SARS-CoV-2 genome that encode for proteins which don't mutate. So we've done a big analysis looking at conserved genes in the SARS-CoV-2 genome, you know, things that are basically unchanged from SARS-CoV-1 as you move to SARS-CoV-2. So early on in the pandemic, we didn't have all the variants we now do. We basically had to compare CoV1 versus CoV2, and you could see that some genes just didn't change. Okay. So we were focused on those, and obviously, as Spike continues mutating and as the coronavirus mutates, you have more information showing which regions are conserved. And some of those regions create good T-cell antigens. And we now use our prediction technology, the EDGE platform we started talking about, to predict which of the regions of SARS-CoV-2 don't change and are good antigenic targets for T-cells. And so we create the Spike Plus vaccine. We have full Spike, for which is whole protein gene for antibodies, and then we have these TCEs, as we call them, T-cell epitope regions, that contain the conserved fragments of other genes. So it's that Spike Plus construct that generates T-cell immunity that potentially will drive durable clinical protection, because those T-cells that you create, that recognize conserved antigens, should provide you with continuing clinical protection against severe disease. Even if Spike mutates beyond all recognition, that T-cell immunity should persist. And there's lots of evidence of the importance of T-cell immunity to protect you against severe and fatal SARS-CoV-2 infections. So that's point of differentiation number two. Thirdly, all of this is done using low dose administration. So we've shown that as low as 3 mcg gives you good levels of antibody, and that obviously has a huge cost of goods advantage, which is important to pharma and to people, obviously, who have to pay for this in countries with lower incomes. It also creates some potential for combination products, where you combine things like influenza and SARS-CoV-2. Because when you think about combining different pathogen products, you start to get into the issue of blending different products into a single syringe. So it's a single shot, but you're actually giving effectively two different vaccines, one against SARS-CoV-2 and one against influenza, for example. And now, because of our payload capacity opportunities and because of the low cost of goods, you could think about blending three different 5 mcg products. Total dose of 15 mcg, well-tolerated, cheap to make, but gives you good, durable antibodies and T cells against three different pathogens. That platform flexibility, I think, is gonna be very important in the field as we're starting to move to this notion of multiplexed products. Nobody wants three different shots to protect against influenza, SARS-CoV-2, and RSV, for example. So multiplexing, I think, will become the goal for the field, and I think samRNA positions itself well for that opportunity. We're out of time, but I do want to ask, maybe brief answer: What do you see as sort of the registrational path, the path towards licensure for CORAL? Yeah, I think the big 10,000 patient study should be the key clinical data set, showing that the platform has superiority over first gen products. Then you'll need to do a little bridging study. This is hopefully something that we'll be doing somewhere in the summer of 2025, right? If this study starts in Q1 next year, it's a 12-month study, so we'll have data as we enter mid 2025. Then in mid 2025, presumably, the ACIP, the CDC committees will sit down and make their recommendation for the strain of virus to be included for the fall vaccines. We'll make that product. We'll vaccinate, I don't know, 100 healthy volunteers, show that you're making good antibodies, as you'd expect. This is the established paradigm, obviously. Then in principle, you could file with your big clinical data package using an XBB.1.5 variant and your current immunobridging study in a smaller group of healthy volunteers. That package, in principle, could be sufficient. We've not spoken to the FDA- Yeah. Let me be clear about that, but that's the concept that I think we're driving towards. Very helpful. Well, a lot going on. We're looking forward to a busy year. Very good. Thanks. Thanks, Andrew. Appreciate it. Thank you.