Voyager Therapeutics, Inc. ($VYGR)

Hello, everyone. I'm Pete Stavropoulos, biotech analyst at Cantor, and welcome to our webinar series to discuss the blood-brain barrier processing technologies. We recently hosted Voyager to discuss their NeuroShuttle platform for delivering protein-based therapeutics across the blood-brain barrier. And today, we're continuing the series with a look at their tracer AAV capsid platform for CNS gene therapy. To sort of set the stage, the blood-brain barrier has a central role and has been a central bottleneck in CNS growth booming, especially for gene therapies, limiting the ability of systemically delivered AAVs to achieve broad efficient transduction in the brain. However, we have emerging gene therapy platforms with engineered AAV capsid optimized for blood-brain barrier crossing. We're also seeing growing strategic interest in the space over the past several years. There's been numerous partnerships and transactions centered around this AAV technology, capital technology and unsort the increasing recognition that the delivery and precision approaches are key to unlocking significant value for CNS gene therapy. We're excited to have with us today Voyager, joined by Todd Carter, the CSO; and Mathieu Nonnenmacher, I don't know if [indiscernible], sorry about that, Vice President of Gene Therapy. So welcome, thank you for taking the time. And let's start off with the introduction of yourselves and a snapshot of Voyager.

Right. Thanks, Pete. Very glad to speak with you again, and we had a great conversation a couple of weeks ago, and I'm happy to be talking about the capsids. I'm Todd Carter, CSO here at Voyager. I have been here not quite 10 years, have been in the field for a while and really excited to talk about our -- what we think are some really exciting work we're doing in the capsid space. Mathieu?

Hi, very glad to be here. Thank you for having us. My name is Mathieu Nonnenmacher. I'm the VP of Gene Therapy. I have been with Voyager for about 11 years now, where I've been leading mostly the novel capsid discovery program. I'm a virologist by training, and I spent my post grad before joining Voyager, [indiscernible] working on the cell biology of AAV vectors. Very glad to be here.

Thank you for that. Todd, do you want to give a resnap of the company?

Yes. So you're showing our pipeline slide here, but maybe just a general intro to Voyager. We're seeing us focused neurogenetics therapy company. And we're known, I think, for novel BBB-penetric capsids that we're going to be talking about today. Now these have a really broad delivery throughout the CNS after a systemic onetime intravenous injection. And one of the things that hopefully we'll talk about are the lower off-target tissue deliveries as we detarget the liver as well. In addition, and what we talked about a couple of weeks ago was all that work led us into the discovery receptors that we can use to get other things across the blood-brain barrier in a shuttle fashion. This year, in particular, is pretty important to us. We like to talk about it as the year of tau. So there are lots of readouts around tau in the field that have been coming out this year, and we have 2 programs in that area tracking in the clinic. One is a monoclonal antibody targeting tau. It's differentiated by where it targets and specificity for pathological to tau. We'll have tau pet data readouts in the second half of this year. And the second thing is the single dose gene therapy using one of Mathieu's tracer capsids. We'll be delivering a vectorized SiRNA to knockdown tau and RNA and protein, and that will be filing an IND and getting into the clinic in the second half of this year as well. And then finally, related to that aspect of things. We also have a partnership with Neurocrine for gene therapy for [indiscernible] also with another -- with a novel IV DBP penetrate capsid that will be moving to an IND this year.

All right. So your discovery platform enables sort of continuous capsid innovation, tail of properties for specific indications. We're going to get into a little bit of detail on that in a short while. However, how does that sort of translate into your partnerships? You did mention Neurocrine partnership strategy today. Can you just talk about your current partnerships and you did mention some of the Neurocrine milestones, but any others that we should expect?

So we've actually, I think, been pretty fortunate with our capsid partnerships. And the religamic from what I'll describe as maybe a pure licensing of our capsids. So we've got partners with Neurocrine, Alexion and Novartis. So a pure capsid license, it might be a case where they take our capsids and do their own work with them, advance their own programs, and we have multiple examples of that where they're working on CNS deliver the approaches. We have partnerships where it's fully collaborative. So for example, with Novartis, we've got Huntington's disease program, where we are working quite closely with them. We do work internally. All that should be inversed. So it really supports all of our de-therapy work. And we have partnerships with Neurocrine that are similar. I mentioned the FA program. There's a GBA1 program. And then finally, I'll just call out with Novartis. There's an example of spinal muscular atrophy, where they've licensed the program from us and they're working on additional work. And of course, Novartis is one of the most well-known gene therapies with their spinal muscular atrophy [indiscernible] gene therapy. So what we've been fortunate there is all of that has really allowed us to bring in close to $0.5 billion in revenue, mostly nondilutive over the past several years, across all of these programs in addition to which a lot of our work is just fully supported as what we're enhancing those. And all these programs are really enabled and by the novel BD penetrate capsids, the engineered capsids from the foundation of all of those relationships.

Okay. I probably won't disagree neuro indications are also in high-risk endeavors because of biological complexity and the tiles are delivering large therapeutics efficiently across the blood-brain barrier. What are some of the ways you at Voyager navigate through these risks to sort of increase or trying to increase the probability of success?

So there are multiple kinds of risks in the way we think about it. There's the target risk, there's the delivery risk. And then there's -- all of that also is included in, I'll call it, the clinical risk or the ability to get a readout in a cost- and time-effective fashion. And we try to address all of those. So on the delivery side, we think we're solving that these novel capsids in the gene therapy space really, as I mentioned, delivered broadly at low doses for AAV, and we think we're solving that delivery challenge. What that allows us then to do is take things that on the target side, and actually ask the question because I think one of the biggest issues in the field has been that we haven't really been able to answer the question of is a given target truly the appropriate therapeutic target or mechanism to treat a particular disease. We haven't been able to deliver to ask and answer those questions, call them the delivery problem should do that. So what we're doing are starting with the things that we think are the most validated on the neuro side, meaning they're genetically validated in the best cases or potentially clinically validated by other mechanisms. So we're able to take that risk, plan on the delivery side, where we're solving that and then take the lowest risk targets forward first. On the efficient path to clinical demonstration, right, the challenge -- one of the challenges in neuro has been, it takes a long time, particularly in their neurodegeneration to see whether or not the treatment could be working. And so to deal with that, we are taking paths where we can identify an effect, I'll call it a pharmacodynamic or pharmacological effect quickly. For example, for tau, we can look at tau knockdown, the gene therapy. To give us an idea of whether our treatment is working in a pharmacological sense. But then there are also clear ways to look at efficacy. So in a tow knockdown fashion, that is tau PET, where there are now pretty well understood in the effective ways and tracers for PET imaging of pathological tau. That's the approach we're taking to both derisk the earlier stage and to be able to get to that proof of biology and proof of concept in the plan as quickly as possible. So we can either identify that yes, we're working and we can identify the dose that we'll work moving forward. Or on the other side, we can cut our losses and move to something else as quickly as possible.

Okay. So how should we think about the blood-brain barrier as a bottleneck for CNS gene therapy. More specifically, can you just walk us through how the blood-brain barrier has constrained both the development and scalability of gene therapies?

So some of this overlaps a bit with the last question, but the blood-brain barrier, I think we all understand it, it excludes 98% or more of small molecules and large therapeutics. Some small molecules can get across and you can do work to design and try to improve that delivery. Large therapeutics to date have been pretty challenging, the antibodies, et cetera. And of course, there are several approaches that people are taking to try to achieve that. But really, that blood-brain barrier has prevented us from tackling most of the targets that we want to be able to target, whether they're oligonucleotide-based antibody-based or proteins. By using gene therapy, we can actually deliver many of these things. If it's a gene therapy, we can express the protein or enzyme directly in the CNS. If it's a knockdown approach, like a tau knockdown program, we can directly express knockdown siRNA to target, in this case tau or whatever the particular mRNA it is. So really expect to that question of being able to deliver the right treatment for the right therapeutic area to the region. The other challenge is achieving broad or uniform distribution. So there are some approaches you can do intrathecal delivery or you're injecting into the effectively cerebral spinal fluid around the spinal column. You can do direct injection into the brain and some other similar routes. What those typically do those, they can give you a delivery, but they can do so in a way where you get a large gradient or a very substantial gradient. So delivering [indiscernible] into the intrathecal space, very high levels around final cord and then lesser amounts into the regions further into the brain, that could very well be appropriate for some diseases, but not so appropriate for other diseases where the particular regions that you want to hit don't really reflect that distribution. The same is true for [indiscernible] or a direct delivery into the brain, and Voyager has some experience because we spend time trying to optimize that. However, again, if you inject say, gene therapy into a very focal point in the brain, you get some distribution and some delivery, and you can change that a bit depending upon the serotype you use. But ultimately, you still end up with a very significant gradient, a very high amount of vector delivered at the point of delivery and lesser amounts as you move away from that. So one of the benefits of tackling this by using the vascular system is a vascular system is pretty widely distributed for good reason. And so by piggybacking on what we think are normal distribution methods or BBB crossing mechanisms we really hard is that ability. So maybe, Mathieu, I don't know if you had a couple of comments there.

Sure, sure. And I completely agree with what Todd just said. So you can look at the blood-brain barrier as the main implies, an obstacle to the diffusion of molecules, especially large biologics to the brain. But if you look at it as a gene therapy person, this is also an exceptionally good distribution network for the brain. And so the brain is a very energy-hungry organic requires -- it doesn't have its own energy storage. So it requires constantly nutrients, oxygen and everything it needs to function with very high energy demands. And in order to achieve this, it uses an extremely dense [indiscernible] network with a brain barrier, just 400 miles of blood vessels in a human brain. That's about 20 square meters of total vascular surface as an exchange service between the bloodstream and the brain cells. And it is generally believed that every cell in the brain is within 20 microns of a blood vessel. So if you can find a way to officially transport your biologics across these vascular network, you get an exceptionally good distribution -- distribution system for every region of the brain, including the deep brain regions. So this is really how the field has been looking at the BBB from the very beginning. It's an exceptionally good way of bringing things into the brain if you can find a way to cross it.

All right. There's sort of a clear shift towards engineering capsids, specifically for crossing this barrier rather than relying on naturally occuring [indiscernible]. Mathieu, you're sort of drawing on your experience advancing this technology, especially gene-based therapies have your role in developing a platform of Voyager as well. Can you give us a quick overview how you're leveraging the blood-brain barrier transport system to enable efficient delivery of gene therapies or large drugs into the CNS?

I'll start and then ask Mathieu to jump in. So really, Voyager started as using some of those delivery routes that I mentioned earlier. We started looking at [indiscernible] intrathecal delivery. And when we kept bumping up against these challenges, and these challenges I mentioned were the very large gradients, a very focal delivery, you can end up with a patchy and inconsistent or not homogeneous delivery. And that led us into the thought that if we could harness the vasculature and the way that Mathieu described, that could enable us to really deliver much more homogeneously and even at lower doses. Mathieu, do you want to talk them through kind of how we got where we are on this?

No, absolutely. So especially in the field of AAV capsid evolution. So that idea of randomly introducing sequences on the AAV capsid surface to impart some new properties has been around since 2003, 2006. In the case of the BBB precisely, it really came down to then government's early work when you would at Caltech and then at the [indiscernible] who came up with the first capsid that could truly traverse the BBB in a mouse model very efficiently. And that was such a paradigm shift to see what capsid directed evolution could do. So that capsid PHP was about 50x more efficient than AAV9, which arguably is the most efficient natural AAV when it comes to crossing the BBB. And that's really reignited the field very efficiently to see what capsids could do when they were properly engineered. And then Voyager jumped pretty much at the same time around 2016, 2017 with the tracer platform, and we were the first lab to demonstrate efficient BBB crossing in nonhuman primates with the [indiscernible], first of all, that we introduced in 2021. And then with the VTAP102 that came a little bit later with the identification of the receptor. But so all these different progress first in mouse than in nonhuman primates. We're done very rapidly after the technology for directed evolution was perfected to allow these discoveries. And so -- since then, we and others have come up with multiple versions of these [indiscernible] crossing capsids, multiple mechanisms and receptors. And we really believe that this is now an inflection point where the delivery part of the gene therapy is turning into something that has been sold or partially sold, and that really opens the gate for many high precision therapeutics to affect multiple brain regions and indications. So the years to come will be extremely exciting with multiple clinical validations of those capsids in patients and so on.

What do you believe are some of the core design principles that sort of may ultimately determine whether a blood-brain barrier across the gene therapy platform succeeds?

So, so far, it's been really the implementation of a function first discovery pipeline. Really, the difference between the work of [indiscernible] and then the work of Voyager and others and our predecessors was that there was a very strong and specific functional readout for those capsid screens that relies on gene expression and not just the physical accumulation of viral particles in a particular region of the brain. So I'm not going to go into the details just yet. We will get back to this later. But I think really that was the first principles of capsid library design and the enrichment, the screening method that was relying on gene expression that really made the difference between all these new generation of capsids and the previous ones.

All right. So one question I have is, why does AAV sort of remain the preferable modality for CNS gene therapy? And what are specific advantages of AAV versus others?

So there's multiple reasons for that. AAV has -- is extremely good for nondividing sales. You can provide lifelong expression of the transgene, providing that the cells are postmitotic, which is the case for neurons, for example, and mostly for astrocytes as well in the brain. It is viewed as a very low immunogenicity vector by comparison with, let's say, adenovirus, for example, that triggers a very strong immune response, AAV start to be much more healthy. It is a very simple virus to engineer. It has virtually no leftovers of viral genes in the final factor. It is completely devoid of potentially toxic viral genes. It only keeps the terminal repeats as the score of its previous life as of ours. And another fundamental advantage that AAV particles might have in the brain is because of their small size. They're only 25 nanometers in diameter, which makes them very capable of diffusing in the brain tissue where the extra similar states start to be constraining. Any particle larger than 7 nanometers in diameter is supposed to get stuck at some point in the [indiscernible] space, which severely limits the potential use of, let's say, LNPs or other larger LPs or viruses, but AAV because they're so tiny, they're among the smallest viruses in existence, they can actually [indiscernible] very efficiently throughout the brain tissue on several -- on relatively long distances. And they do have a very good track record of success from CNS diseases. If you look at [ Zolgensma ], this has been used, I think, in close to 4,000 patients to date, and this is close to -- as close as it gets to what we would call a miracle drug, most of these kids are still live today. They would certainly have died from their disease if it was not for the drug itself. And the viral vectors are still to date, I believe, the best options to deliver relatively large payloads, large genes, and they can be tunable for cell-specific disease specificity. They offer a lot of options and layers of specificity that can be used to control gene expression and eventually the safety and the efficacy of the therapeutics.

Pete, a little bit of that to mostly agree with Mathieu. The downsides can be that we're pretty limited on size. And I think that's probably the biggest challenge of AAV. But for those where we can deliver the payloads we want to, though very low pathogecity, the ability that we're able to achieve with these very blood-brain barrier for entry capsids to get that broad delivery is really excellent and that the proven durability of expression, I think, are very strong positives.

And yes, it is a onetime treatment. So that makes it also a pretty unique option.

Without integrating into the genome.

We have seen multiple improved AAV gene therapies. They've demonstrated meaningful efficacy. And as you mentioned, durability in patients [indiscernible] the example. As sort of the field has incurred, what have been -- what have this program sort of taught us about the relationship of our transaction efficiencies, clinical outcomes and safety. And as those lessons sort of inform the next gen CNS gene therapies that are being designed and developed today?

Yes. So maybe again, I'll start and then ask Mathieu to get in to add. In terms of some of the key challenges that could be addressed, it really is the AAVs did not evolve to cross into the brain and do their thing, they evolve to go in -- to get into reasonably accessible tissues to replicate and then harness other viruses to provide all the machinery. I mean that's one of the things about AAV is why it can have such a particularly specific genome and why it's in and of itself, it doesn't have the pathogenicity seen with some others is because it's a natural [indiscernible] from and uses other viruses to do its journey work. But because of that, at least partially because of that, it doesn't need to target the brand. And so to target the brain, Mathieu and his team have really had to engineer that capability and now the benefit of doing that is we could start harnessing some of the endogenous mechanisms for doing so and get that product delivery. We also know that natural [indiscernible], can work in very young patients, presumably because of the blood-brain barrier isn't fully formed yet. The challenge there is it becomes much less effective in adults and not even adults, but as the children mature into older children than those -- the CNS areas aren't as available. And then the current narrow therapeutic window of CNS gene therapy. So as the field has tried to treat more and more diseases, we've been pushing the doses higher and higher to get the kind of delivery expression that we think we need to have efficacy. But as we do so, deliver, as an example, AADs can naturally go to the liver, but we've been pushing the delivery to even higher and higher, and we've seen some of the challenges around that with some of the safety outcomes. And so one of the things that we always need to do is to build in not only deliver it to the tissues we want to target, but to look for opportunities to de-target those tissues that -- like the liver that we don't want to deliver to. I don't know if, Mathieu, you've got any other comments.

No, that sums it up pretty well. I mean this is always a problem of drug development about finding the right therapeutic window, the delivery, the natural thing for most existing AAV capsids and AAV9 is no exception. And in Zolgensma, many patients have been manifesting some signs of liver toxicity shortly after dosing, which fortunately normalized after a while. But this is definitely one of the aspects of capsid engineering that can be built in to minimize lever exposure while maximizing the brain exposure. And we were fortunate enough to achieve such properties with some of our first-generation and second-generation capsids especially the ALPL binders that we have developed more recently at Voyager. So those capsids show not only a huge improvement of brain transduction between 50 and 200 fold above 89%, but also a very noticeable detargeting of the liver. We think that about 10 to 15 fold less AAV particles end up in the liver, thanks to those capsid modifications. So this is really, really an essential and a central aspect of capsid biology that engineering can address.

Okay. It means you just touched on a bunch of safety concerns for [indiscernible] drawback. But are there any other no safety concerns with AAV9 serial type [indiscernible]?

So there are definitely. The liver toxicity is the most common one. I mean -- you will notice on this table that the majority of these clinical approaches are using extremely high doses of AAV. So we consider anything that is higher than 114 viral genomes per kilo of body weight is considered a high dose. And those can come fortunately not often with noticeable liver impact and an elevation of liver enzymes and potential long-term toxicity to the liver. And other than that, there has been also, in some patients, unfortunately, this is a much less common in them, but there has been some best philosophies associated with AAV9 vectors especially, but also AAV8. These are not very well understood safety reactions against AAV vectors that are still being actively investigated. They can take multiple forms, TMA, HLH, capillary leak syndrome. They're fortunately extremely rare. They affect maybe less than 1% of patients in total, from all the patients that have been dosed with AVs, but the major source of toxicity is definitely liver toxicity. And this is usually seen only with doses of 513 or more barren genomes per kilo. This is something that doesn't typically happen at the doses were engineered capsids would be used in the clinic, fortunately.

I have to think -- sorry, Pete to interrupt. But we also have to comment on a little bit on capsid over last summer, where the -- that's the BBB penetrant capsid company that tragically had the death of the patients, and that was AAV9 capsid as well. So that's minimal work that's been done there. They share a little bit of information and data. We know that our capsids use different receptors than their capsids, so are very differentiated from them in that case. But it does speak to -- you need to know what we're doing and we need to pay attention to these things. I think on the liver side that Mathieu mentioned and some of these others, one of the things that we can do is build upon all the work that's been done to date to mitigate all those risks. So the immunosuppression that really was identified to, I think it effectively saw the issue with Zolgensma and the liver toxicity. And then the other ways we have in solving them, which are really make the therapeutic more effective, lower the dose and then de-target things like the liver.

Okay. Just to be clear that I heard you right, do you think sort of the window of taxes to be started at about [indiscernible].

So this is generally what is assumed from preclinical studies in nonhuman primates. And this has been seen every patient that has manifested liver toxicity, for example, with those with a dose of 5e13 or higher, if I'm not mistaken, which is the majority of the clinical trial currently happening because they are using natural AAV capsids. And in order to achieve the pharmacology, you need to ramp up the dose to those levels. So there is definitely -- I mean the dose is the poison when it comes to toxicity and safety. And there's definitely a dose-dependent effect of AAV vectors, particularly when it comes to liver toxicity.

And I think we're going to touch on your capsid issue. So just quickly, briefly, the primary routes of AAV administration to the CNS that's being used in clinical development today. And sort of how do they -- how do they in respect of biodistribution profiles sort of differ and ultimately clinical implications?

You might take that, Mathieu?

Sure. Absolutely. So right now, in terms of the patient number, the most common dosing route has been IV with Zolgensma, which has proven to be effective in very young patients, so 2 years -- 2 years old or younger. And these typically start to achieve a very margin as very, very bran-wide distribution in both the superficial, the cortical and the deep brain regions. Now among the other dosing paradigms, and many of those have been actually tested at Voyager in the early days of the company, intrathecal has been used pretty widely, especially when it comes to diseases of the final cohort and the brainstem. So this can be quite successful despite the limitation of the PR membrane that limits the transduction of the spinal neurons. It's an actual physical membrane between the spinal cord itself and the CSF. And also, it sometimes translates into higher exposure of the dose with ganglia, which can lead also to some potential toxicities. But overall, that's a method that can work fairly well for spinal cord diseases. After that, this is really indication-specific. So cisterna magna can be used as well for relatively wide distribution in the brain. So this is a relatively invasive method. This is not something benign. And usually, dose direct injection methods not only are a little bit more invasive, but also translating very steep gradients of transgene expression around the injection site that quickly taper off when you go a little bit deeper in the tissue because the brain tissue is relatively solid and compact and even AAV vectors do not diffuse that far across the parenchyma from the injection site. In the case of Huntington's disease, for example, from [ UniQure ] in order to achieve a wide enough distribution, you need to perform 8 burr holes in the skull and 8 separate injection sites, which is quite an invasive method. And even then, it seems that there might be some speed gradients around the injection site where you will have a lot of gene expression immediately around the needle track that quickly tapers off in just a few centimeters away. We see another potential issue with local delivery methods, which are the fact that the rate of diffusion of at particles in the brain tissue do not really scale up linearly with the dose on the brain size. There's physical constraints on the distance that a particle can be used through a solid medium. And a human brain is about 15x bigger than a cat brain. And so whatever you learn from preclinical studies [indiscernible] might not linearly translate into your patients just for a matter of scale and volumetric scale of the brain tissue. There are a few cases though where local delivery can be highly desirable and Voyager was a very good example of that in our very early years with the Parkinson's disease program with delivery of an ADC and coating vector. So the idea here was to turn some neurons present in the [indiscernible] neurons like expressing ADC, which is an enzyme that can metabolize [indiscernible] to dopamine. And for the success of this therapy, it was absolutely essential that the expression of the charging be limited exclusively to the prepayment neuron. So in that case, that's one of the few instances where local delivery is not a constraint, but it's actually an asset. It's a feature. You cannot deliver those vectors throughout the entire brain, because god knows what would happen if every neuron becomes a dopaminergic neuron. And so that's one example where local delivery is actually almost the only way to perform this gene therapy efficiently. So ID can solve a lot of problems, but this is not necessarily the kind of [indiscernible] for every single indication and every single disease. Some of them still require some more localized intervention.

[indiscernible] have little bit of a difficulty with the slide deck, but we'll just continue until it's solved. And also, which sort of disease areas of neuroscience are most likely to sort of benefit from the blood-brain barrier crossing AAV capsid, sort of like what is the low-hanging fruit?

Yes. So maybe there are a few. As I mentioned at the beginning, we tried to derisk on the target side since we're exploring delivery side to begin with. And by doing that, we'd like to look at the genetic diseases, there's a clear through line. There's genetic -- significant genetic evidence [indiscernible], right? Things like [indiscernible]. We've had -- we've had SOD1 programs, Huntington's disease, you can look at [indiscernible] and others. And another example of that is GBA1 for Parkinson's given that GBA1 is a common -- the most common mutation found in Parkinson's disease mutations. Something like Tau, there's -- while -- while tau is not genetically linked to Alzheimer's disease, it is genetically linked to neuro degeneration, their mutations in tau that now the cost neuro degenerative disorders. And of course, tau itself is heavily implicated in things like product traumatic encephalopathy in addition to Alzheimer's disease. And so there are sporadic diseases that have a very clear through line particular targets. Oncology is another example where often the blood-brain barrier is a literal barrier that prevents the therapeutic from acting in the brain. And so you might clear in the periphery that you could have metastasis to occur in the brain at all and then the predicted area. So if you can deliver across -- you've been taking something that has been shown to work it gets a particular cancer, and just provide that in the compartment where that therapy has been previously prevented from region. And so those are some of the areas that we think are really the -- I don't know, low-hanging fruit is kind of a front term, it's neuroscience, so there isn't a lot of true low hanging fruit, but I think there are really opportunities in all of those.

Okay. Next slide, sort of from an investment perspective, there's a lot of focus on capsid innovation as a key differentiator. So how do you sort of think about the key design principles when developing capsid that can cross the blood-brain barrier, what properties actually matter most for clinical translation?

You can start, Mathieu.

So I mean, in this case, really the key differentiators are really the transaction efficiency in the brain. So we think that hitting about 50% of brain sales at a dose that is way below the expected toxicity for a hold of AAV would be the pharmacology criteria that we're going after. And seeing some significant targeting of peripheral tissue at this dose is a plus. I mean it's definitely something that we really like to see. In addition to that, there are some indication-specific features that can be more like which cell types in the brain does your capsid target the most, some capsids appear to be mostly neurotrophic. Some of them are more widespread in terms of their similar preference. The advantage of gene therapy and viral vectors, especially the modularity of the approach. You don't have to do all the selection and the work at the capsid level. You can also work on the payload level to complement the capsid profit. At the end of the day, the distribution of the gene therapy is a the diagram of capsid trophism route of injection and cassette control. So these really gives you a lot of options to tailor your gene therapy vector to the disease. But really mostly having favorable pharmacology and if possible, the targeting from very [indiscernible] tissues. And then we can come back later to the translatability aspect, which we think is absolutely fundamental to new capsid design as well.

I think that's worth commenting here as well, sorry to interrupt Pete, but that translatability gets to something that we built in. I mean, in a sense, we learned a little bit the hard way, but we built it in very early, and I think it's been really critical to developing our current client capsid platform. And that's ultimately the requirement that we needed to see multi-cross DCs activity across multiple species. And that stems from some of what Mathieu was describing earlier, whereas the amazing work that came out of Caltech with the Deverman in these mice that showed you can deliver something into the brain after an intravenous injection, but it was limited. It did not cross in other species and in fact, it's not even crossing the multiple strains of mice. And so pretty early on, we discovered that you can identify using that functional readout, the expression-based readout that Mathieu described that we had to get the technology to do, something that worked really well in one species, but the chance that, that would translate into humans is pretty low. So we built it in even though it took a little extra work and more experiments to do so. But I think that's been critical to getting to the kind of translatability that we expect to have in humans.

A little bit of more work, and was that for you? Or was it for Mathieu? [indiscernible] you increase sort of derisk a bit more, the translability.

What it is, and I think we've got evidence that it actually did so. And hopefully, we'll have some time to talk about receptor discovery because I think that's the key output. It isn't just the hypothesis, it's that we're seeing it play out in real time.

Okay. I guess sort of talking about the receptor. When you think about engineering, the next gen capsids, sort of how do you approach identifying and validating the right receptor, sort of what characteristics matter most in terms of expression and trapping and ultimately translational relevance?

That's a Mathieu question.

That's it. Thank you. So that has really become an absolutely key aspect of capsid engineering. Once again, this was pioneered by directly the work of Ben Deverman. I mean, when he identified the receptor for the [indiscernible] capsid, which turned out to be [indiscernible] that was really the first evidence that a direct capsid receptor interaction was the main mechanism for crossing the BBB. And that really started this receptor chase that every capsid developer guarding to because this just gives you so much confidence about the potential translatability of your capsid in patients or allows you to exclude capsids based on their lack of interaction with a human receptor. So because AAV capsids are biological entities, and they do rely on putting interaction to accomplish their work. This, in turn, triggers a lot of specificity. There's a lot of species specifically about protein-protein interaction. People who've been working for a long time on transferring receptor in the shuttle field know very well what I'm talking about because a single amino acid in the transferring receptor can prevent your therapeutics to buy into the monkey eyes form, for example, whereas it can bind very well to the humanize form. So we're talking about that level of specificity. And in the case of AAV capsids, the translatability across species and humans across preclinical species and humans has been a major block in the advancement of the field. [indiscernible] is an example. And so the identification of the receptor that is mediating the properties of the capsid is absolutely key to this day to the success and the translation of these gene therapies between preclinical models and humans. So that doesn't mean you have to design your capsids with a specific receptor in mind. Most of the existing screens, including tracer or empirical screeners that rely on the biology and the function of the capsids first. So you put your library in a monkey and you wait for the monkey to do the work and to determine which capsids are going to reach the brain and which ones are going to get excluded. And the receptor identification comes up posterior. An interesting part of this is that the receptors that come out of those empirical screens typically are very new and unexpected. These are not the ones that you would expect from expression patterns, abundance in the BBB and so on. They usually are quite surprising in terms of their biology. And -- but you cannot argue with the function. I mean these receptors work. They can carry a giant bioparticles or 4 mega daltons across the BBB. And so that means they are very high-capacity transport systems. So that's one aspect of it. And to recall, evolution screen of the capsids followed by the identification of the mechanism of the receptor, which is absolutely essential for the translatability to humans. And then you have the mirror image of this now that is becoming more and more common in the field is to specifically tailor capsids to force them to engineer with a particular receptor, a priori. And this has been done very successfully with the transferring receptor. For example, they are extremely potent capsids these days that have been purposely designed to interact with TFR, and they do transport themselves across the BBB into the brain with very high efficiency, and there's multiple other receptors that are being used right now to evolve the capsids in a very design first way and function second, if you will. So this has really opened an entire field instead of working with the black box, which was the case until recently where nobody knew what mechanism those engineered capsids were using. We were just happy that they were doing what they were doing. Now this has morphed into a much more controlled, much more scientific, if you will, way to see how the captive interacts with the cells, how they perform their transcytosis across the BBB and maybe also the targeting of the wholesale at the end to mediate their eventual transaction.

And just again, maybe double down on a little bit of that. It's hard to overstate the importance of knowing what the receptor is and being able to show that our capsid binds not only the nonhuman primate receptor and the rodent receptor form of the receptor, but it also binds the human ortholog. And that gives us just a very high level of confidence that it should translate into humans. And then we can look at what the capsids are doing in the mouse, doing in the monkeys, we can do dose response studies and show the equivalent in those species, all of which just really enables the programs as we move to the human.

Okay. So is there any trade-offs between, let's say, broad brain distribution versus cell type specificity and sort of how do you optimize for that? Even though you sort of alluded to it a little bit earlier in terms of expression in particular cells?

So we don't really believe so. If I want to focus on the most recent capsids that Voyager developed, they appear to be very broad in terms of sales preference. They can target exercise very efficiently as well as neurons, various types of neurons and even oligodendrocytes to a very significant extent. And these are aspects that you can control later on by changing your construct and your promoter or enhancer within your payload. We think that it would be better to have a versatile capsid to begin with that you can later tailor to specific cell preference and indication rather than being constrained by your capsid preference and then not being able to evolve it into targeting different cell types. So there are some excellent capsids out there that appear to have mostly a neuronal preference, and we have identified such capsids as well at Voyager, but there's really something with a broad specificity capsid with a universal capsid [indiscernible] blank canvas for you to keep evolving those gene therapy vectors in whatever way you want and still working with the same capsid. I don't want to downplay the importance of having a capsid that will be soon validated in human patients because getting this out of the way, basically allows you to really start plug-in different payloads into a capsid that has been essentially derisked. And so that could lead to much faster development of gene therapy drugs provided at least the transport part has been validated in humans.

Okay. And next slide, just go a little bit deeper into it. Once the therapeutic is in the CNS, how should we be thinking about the role of the construct, design, promoters and enhancers, et cetera?

Sorry, it's fundamental, and it's really guided by the indication itself. Some indications require a very specific gene dosage. For example, if you think about red syndrome, where too much of the [indiscernible] will be detrimental, too little will be decremental. And so in this case, we have options for a slightly more sophisticated construct than what is illustrated here. But even working with cell-specific or disease-specific promoters or enhancers can be extremely good, that can also really help to target the peripheral tissues. If you work with the neuron specific promoter, let's say, human synapsin 1 to site a very, very well-known one, you simultaneously present transgene expression in the liver, the heart, the muscle, all the tissues outside of the brain in addition to focusing your expression in neurons, which is extremely precious for some indications. And there's multiple ways to do this. You can use microRNA targets. I think this is going to be addressed a little bit later. You can use logic circuits. You can use very sophisticated regulatable promoter to control the expression of your transgene. So once again, I want to reiterate that once the capsid does the job of bringing the payload where it belongs, that really opens multiple options to engineer very sophisticated control and regulation mechanisms if needed. And then you have other indications for which this doesn't appear to be so important. Diseases, for example, that can be addressed by cross correction with secreted genes such as lysosomal storage diseases, for example, or therapeutic antibodies who can work not only on the cells that are being made in, but also the neighboring cells. In this case, it's generally admitted that the level of expression is not that important. Those transgenes and those proteins are not start to be very toxic even at high level. And so in this case, just a very strong promoter that would provide this therapeutic protein across the entire body and across the entire brain might be completely adequate. Genome editing, for example, you don't really need to be cell-specific or tissue specific. If you have a gene mutation, it is present throughout your entire body. So it shouldn't be a major issue to modify it in other cells than the ones that you're already targeting. Same thing for RNA interference. If your target RNA is not present in the cell, then it doesn't truly matter. If it is present, your micro RNA will not get down and everybody is happy. So it's really modality specific, indication specific. But again, having a capsid that goes everywhere or at least goes everywhere in the brain offers a lot of interesting possibilities to play around with the payload and the genome construct.

Okay. And just, I guess, one of the take-home messages was the promoter-driven control is sort of a way to help mitigate some of the toxicities like liver and DRG versus remind just on the capsid alone?

Absolutely. So promoter strength and self-specific can be leveraged to really, really tailor and control the expression of your transgene.

Okay. And next slide, sort of building on that theme of controlling beyond promoter enhancers, other modifications that shape both efficacy and safety...

This is a very interesting approach. I mean that was actually pioneered by [indiscernible] Brown, I think when he was in [indiscernible] lab, and the idea was to prevent the expression of transgene from [ lentiviral ] vectors in immune cells. So the very first iteration of this system, which relies on endogenously expressed microRNA to prevent transient expression, so it is Bryan Brown, not [indiscernible]. So that was initially used to prevent expression of a foreign protein in antigen presenting cells. So that was a microRNA called [indiscernible] that was used to prevent expression. But since that system has been expanded to multiple other modalities, it's extremely elegant because you don't need to introduce much into your bio vector. You just introduce a few sequences that will be targeted by a microRNA that is abundant in the cell type where you want to shut down expression. For example, miR-122 is very well known in the liver. It's a very highly expressed micron in the liver and only in the liver. So if your genome cassettes contains a miR-122 binding sequence at the end, you will completely shut down the transgene expression in the liver in the hepatocytes. You can use miR1 for the muscle, for example, that is -- there's been some very good work from [indiscernible] in humans to prevent unwanted expression of the transgene in the heart and the skeletal muscle by using miR1 target sequences. And more recently, the lab of [indiscernible] has come up with a way to minimize transgene expression in the [indiscernible] 83, which appears to be highly expressed in the DRG neurons, but not so much in brain neurons. So that's a very subtle, an elegant way to prevent and add the specificity layer to your expression cassette. It's a very interesting approach that Voyager has been using as well.

All right. Next slide. We did touch on this toxicities sort of related to AAV that sort of repeatedly emerge. When you do look across the data and just briefly because we did touch on it, what are the sort of the largest toxicity risk? And what are some of the rare ones that you actually look out for or should be looking out for?

So again, like I mentioned a bit earlier, the most common form of AAV toxicity, and this is especially visible and high dose is the liver tox. There's multiple hypotheses to what drives it. There's a combination of direct toxicity to the liver vasculature. So there's a lot of interesting work on this that has been pioneered by [indiscernible] who was working with Jim Wilson. And she has performed some very elegant experiments in nonhuman primates suggesting that there's a direct injury to the liver vasculature that can be caused by high dose AAV that translates into multiple pathogenic evolutions of the liver. And in addition to that, there is also potential capsids antigen presentation by immune cells in the liver and elsewhere that can lead to a CD8 cytotoxic response. So typically, those are managed by steroid, sirolimus injections. There's also more recently the emergence of a role of the complement activation in AAV toxicity, and this is true for both liver tox and vascular toxicity also, which is more rare. And so that is currently managed in the clinic using complement inhibitors. And so this is really the most common manifestation of AAV tox that usually happens at 5e13, 1e14 or even more [indiscernible] and this is not something to be overlooked. I mean this has resulted in patients death in some instances. So this is something that we really need to keep an eye on. But again, something that capsid engineering should help a lot. And it should help in 2 ways. One way is by directly verdetargeting, which we and others have been observing repeatedly in our capsids and also the fact that those capsids will be effective at lower doses. So the goal really is to reduce the doses if possible, to 5e13 and below to really minimize the probability of liver tox occurrence. And the second major tox that has been really emerging from clinical trials. The reason why I'm saying that it's because these are not toxicities that are observed in preclinical models, including nonhuman primates, which is quite annoying, should I say. So those are typically more vascular toxicities, vasculopathies. They do have different manifestations that they can all be grouped under the guys of basically damage to the blood vessels that are caused by high dose AAV. Fortunately, those are much, much more rare. These are still freak accidents or incidents in AAV trials, but we're spending a lot of work and efforts trying to understand those pathologies a bit more. Again, these are most likely pathology that could be minimized by lowering the dose. So this is really the key aspect of capsid engineering. And the more engineering is happening to more first generation, second generation, third generation capsids will be made, the lower the dose that will be necessary to achieve the right pharmacology, like the [indiscernible] for the brain, we think, is an achievable goal. We have had some very encouraging pharmacology with some of our vectors at doses of 1e13 viral genome per kilo in the brain in our VY-1706 to program. So we do believe that with more efforts in capsid engineering and optimization, we should be able to reach satisfactory brain pharmacology at 2030 or less. This is going to take a few more years. But I think this is something that is achievable. And we're already seeing this in preclinical models for muscle capsids, for example. There are multiple examples out there of engineered muscle trophic capsids that show efficacy at 5e12 biogenomes per kilo, at least sub-13 bio genomes per kilo, which is extremely encouraging for the future of the field. And overall, I must also note that even though the field is extremely worried about some of those findings, the SAEs that are observed in patients, so far, grouping all the clinical use of AAV vectors, the rate of SAEs has been about 5% and patient death, which are always extremely unfortunate, are about 0.5% in terms of AAV treatment in the clinic, which could [indiscernible] I think on par with many, many other drugs outside of the gene therapy field.

One other AAV sort of to touch on is DRG on the next slide. Just how is monitoring clinical trials and what's the clinical significance of DRG toxicity?

So there has been observations mostly in nonhuman primates of [indiscernible] caused by high virus doses. I think in patients, if I'm not mistaken, there has been 1 occurrence of measurable clinical pathology that was due to the DRG tox with an RH10 vector, if I'm not mistaken. So it appears that [indiscernible] in particular, might be oversensitive to the [indiscernible] toxicity at doses that are currently being used in the clinic and that appear to be perfectly safe. Now what Voyager -- with Voyager Therapeutics is implementing in all our preclinical experiments is a very careful follow-up of neurofilament levels in the CSF and in the blood to make sure that we keep a very close eye on the RG pathology. And at every first sign of NFL elevation, we consider that these doses are starting to get a little bit high. So we keep a very, very close eye on potential neurotoxic and especially DRG pathology by closely monitoring NFL levels in the CSS serum. And typically, in order to advance our programs further to the clinic, we like to define our therapeutic window as a region where we have satisfactory brain pharmacology with little or no NFL elevation for following injection of our capsids at the mill. So we keep a very close eye on it.

And maybe I'll add to it. I think in the clinic, it really is almost observational in terms of monitoring for it. I think Mathieu is entirely correct that when you look in the nonhuman primates, there has been [indiscernible] findings as you push the dose. And do you think that's probably because ERG. You see this in intravenous. You see this in intrathecal delivery. It's basically these neurons kind of sit on both sides, both -- on both sides of the blood-brain barrier, so they're just susceptible. They did hit or targeted by the gene therapy treatments. But the small immunochemical findings that can be seen in the monkeys rarely, actually convert to something that looks observational or clinically meaningful. And so that's been one of the things that the field has been learning. But as we push the dose into the high E13 per kilogram, that's when you start to manifest. And so when a great way of dealing with that, in addition to the kinds of things that Mathieu described, where you're going to have the mirror binding site, et cetera, is just to lower your dose because if you go on a low enough dose, then PAUSE aren't seeing the DRG is particularly hard and you can avoid some of the issues. And in particular, avoid any issues where you're actually pushing with this high enough to have the clinical manifestations, which appear to be extremely rare in humans.

All right. I guess one major innovation is the development of a family of capsids that you -- that you've innovated and you have the ability to officially cross the blood brain barrier and trans do cells across diverse brain regions. Just walk us through the discovery process for these capsids?

Mathieu, that's yours.

Sure. Well, I mean so for the tracer platform, it's a system that is conceptually extremely simple. So we started like many of our predecessors by engineering peptide display and capsid library. So there's many ways to randomize the surface of AAV capsids. The one we chose was to introduce short stretches of completely randomized I mean, as said, 6 or 7 amino acids in key locations on the capsid surface that are very tolerant to mutations and that are very, very exposed. And the idea really was to create a de novo binding domain for BBB receptor or anything else by introducing these completely randomized sequence. And due to the recent innovations in AB library design and all this is an extremely powerful system. We can test more than 100 million variants in every single biopaing experiment in our animals. And really, the innovation of the tracer platform was to use an RNA-based recovery system that restricts the enrichment of the capsids to neurons. So we have a neuron-specific expression of the capsid library RNA. And so in order to see an enrichment of these particular mutants, you needed to transduce neurons efficiently. So I'm not sure I'm doing a very good job at explaining. But that's really the innovation of tracer versus previous approaches is that in order for your capsid to pop out of the screen, it needs to transduce neurons, absolutely. And this is really what we want to achieve at the end of the day. So this is really a function first system where you focus exclusively under the biological function of your variants. And so following a few rounds of enrichment like this, followed by deep sequencing analysis for which the peptide display approach is particularly appropriate, you start seeing some sequences emerging out of your monkeys and your mice. And this is really the crux of the tracer platform is when you start seeing these sequences rising above the genetic noise, you know that you're on to something. And after this, we have some other systems that are being implemented to keep multiple single capsids and testing relatively large pools. And in the case of the capsids that are pursued the more seriously now by Voyager, the [indiscernible] series, that system allowed us to actually identify a whole family of those. There was more than 20 capsids that seem to have similar performances, and they all have the same motive that is here on the deep panel with the figure on the left. They had this SPH motive, which is absolutely indispensable for the interaction with ALPL of these capsids. And really, that was the first time we saw such a large family of capsids. And in addition, they were capable of identical performance in mice and nonhuman primates, which is exceedingly rare in AV capsid evolution. Typically, you go towards a single species. And in that case, having the spend species properties was extremely interesting to us because that immediately derisked a lot of the clinical development for these capsids. And so the identification of the receptor came afterwards, took us about a year to get to the receptor and it turned out to be a very interesting protein called ALPL, Alkaline phosphatase that is very abundant on the brain vasculature. You can see that on the right, these outstandings of ALPL in human brain sections. And they also expressed at high levels in nonhuman primates and in mouse. Those are very, very highly conserved proteins. They are present all the way from bacteria to mammals. And so they are very basic bricks of leading beams. And so we think this is one of the reasons why those capsids are working so well across species because they can bind to these very, very highly conserved alkaline phosphatase receptor. So that was the read the story of the tracer platform and how that led us to discover some of these first nonhuman primate capsids. This one was not the first, but it's clearly one of the most interesting that we have in stock at this point.

All right. Next slide, and you've got to find 2 of the capsids. What are some of the key characteristics and attributes you're sort of focusing on when selecting the capsids from both safety and efficacy, but also when you think about commercialization in the [indiscernible]?

Absolutely. So this is really what is represented on the Venn diagram here. You want to maximize the pharmacology, so you want to capsid that is even better in the brain than the first generation. If possible, minimize the peripheral tissue exposure, especially the liver. And ideally retain the good manufacturability of the capsule. And so we were extremely lucky with the second-gen capsids that came out of the [indiscernible] because they actually showed those 3 characteristics in one single package. They were showing another 4, 5-fold increase in brain transduction in nonhuman primate. They were retaining their properties in mouse, which was essential. They were showing even further decrease in liver exposure compared to the first-generation capsids, and they were still extremely, extremely favorable in terms of manufacturability. It is hard to make AAV. This is probably one of the most complex drugs to make at this point. It is a very complex process. And those capsids are quite equivalent to AAV9 in terms of manufacturing process and efficiency. So this is something that is absolutely crucial. Gene therapies are very expensive drugs. The cost of goods for gene therapy vector production is extremely high. And so if you deal with a capsid that is difficult to make, that is definitely going to impede your capacity to manage large conditions and to evolve it to the clinic. And so in order to move towards clinical development, we think a good capsid has to fulfill at least those 3 properties, and of course, the translation.

And then I'll add, we can use the methods that have been already developed for efficient and relatively cost-effective ways of producing things like AAV9. So we require that all of our capsids have those key manufacturability criteria. So we can use similar techniques. We have a tech ops group that knows what they're doing, and we can take advantage of these economies of scale when we produce our material.

Okay. Next slide, please. So how do you view ALPL sort of mechanistically -- similar to transparent in terms of transcytosis or are they meaningful differences that you'd like to highlight. But also just curious of whether the transformers there will pop up on your screens?

So ALPL is quite different from transferring receptor. And as I mentioned before, this is not necessarily something that would have popped out from a list that would have gone after immediately transfer and there was very good reasons for people to do after transferring. It was known for having a very high capacity to transcytose [indiscernible] transferring across blood vessels. It has a very high recycling rate. ALPL doesn't seem to have any of this. ALPL steady state is at the cell surface. It doesn't seem to have a particularly fast and endocytosis rhythm. And it is -- we think it is induced by the binding of the virus itself. So the interesting thing about ALPL is that it tells us that you can start looking outside receptors that appear to have the right profile for transcytosis. This is not a transmembrane coating. This is a GPI-anchored protein, which is also a different category. And yes, it works. So that actually expanded our horizon a little bit as to what constituted a good transcytosis receptor in the BBB because this protein was not supposed to. It is not famous for having a very high recycling rate or any such thing. And it turns out that it doesn't seem to be indefensible, which is really where the function-first approach really shines because regardless of the biology that we know about a given receptor, the virus tells you what works and what doesn't work. And so this is really a fundamentally different way to identify good BBB shuttles. By looking first at what can carry this giant 25-nanometer nanoparticle across the blood vessels and ALPL was one of those. The distribution of ALPL is quite different from transferring receptor. It's mostly expressed on arterial vessels and capillaries, whereas transferring is mostly on the venous side of things. So it's a different subpopulation of blood vessels that seem to express it. The biology of ALPL with relation to age and disease is also a little bit different. It appears that ALPL is expressed at higher and higher levels when people age. Transferrin receptor is quite the opposite. It seems to be expressed more into younger people, and then it goes down with age. So they don't have much in common really, except that they both work in carrying AAV across the BBB. And to answer your second question, I am not aware that anybody in the field has ever identified TFR as a capsid receptor in an empirical screen. We know that if you engineer an AAV capsid to bind to TFR, it is going to become a very successful brain capsid. But so far, this has worked only with the human TFR. And I don't know if it's a matter of species. I don't know if the monkey TFR or the mouse TFR is not very prone to interacting with AAVs. But as far as I know, nobody has empirically discovered the TFR binding capsid in a library screen. And that's still a mystery for the field that nobody has really an answer for at this point. So yes, that's -- there's another fundamental difference for that almost between TFR and ALPL, which is transferring receptor is the textbook receptor for class remediated endocytosis and trafficking whereas ALPL just like most of GPI proteins, traffic with lipid rafts using a completely different trafficking route that goes to the golgioparatis by retort transport. At least most of the lipid [indiscernible] proteins are supposed to do that. So completely different.

And so if I heard you correctly, in general, ALPL doesn't sort of move, but once you actually engage it, I guess, on that site where the [indiscernible] then it's sort of [indiscernible]?

Yes, which is quite common. I mean this is what we know from our own in vitro experiments and also from what is known about ALPL function, it's not a transporter. It's not something that catches things and bring them into cells. That's not its primary function. The primary function of the ALPL is to defer things on the outside of the cell. But it's very common in trafficking and endocytosis that when you start binding multiple receptors on the cell surface and you cluster them, especially big cluster around a spherical particle like an AAV capsid, they can start curbing the platform membrane and that in turn triggers endocytosis. This is a pretty common mechanism of endocytosis, especially in the lipid wraps and especially with GPI proteins because they do not have a cytoplastic tail. And so that's the only way they know that they need to form an endosome. If you start clustering them around a particle outside the cell. This is not fundamentally surprising from a trafficking point of view.

Okay. All right. Sort of moving on to next slide. Just walk us through [indiscernible] I would emerge from the broader capsid family? And are there -- in some of the preclinical data that increase our confidence in this capsid that made it the one to advance?

So the 102 emerged because of two things. It was one of the tough capsids in terms of rain transaction in primates and in mice. And you also started with a noticeable level of liver detargeting. And that -- it kind of stood out the other family of ALPL binders in a screen because of this. At baseline, for first-generation capsid already has some very interesting properties. I mean had a very, very favorable biodistribution in the brain like 10- to 20-fold more viral genomes per sale compared to 89. And 16 to 50 to 200-fold higher transgene expression compared to AAV9 depending on the brain regions. So -- and these properties were pretty much identical between nonhuman primates and mice. So that capsid already cannot had it all to begin with. It had hybrid transaction, liver-detargeting and mirror image of behavior or actually identical behavior across rodents and primates. So it already had pretty much everything we liked. The only thing we did by performing the step-wise evolution into our Gen 2 capsid, which is the one that is mostly being advanced to clinical programs these days, was that we introduced another level of brain targeting enhancement by 3, 4-fold and we further reduced the liver exposure. So the Gen 2 capsids that we're using now is a really optimized version of BKAP-102 that is just pushing everything to the max and that is being even more effective and we hope even safer than the 102. The mechanism is the same. They do interact with ALPL the same way. We have not really been able to identify and to pinpoint the mechanistic reason why the second-gen capsule was better than the first gen. This is probably a very subtle difference. They are proceeding pretty much the same way except the Gen 2 is even better. It's a big 1 or 2 on steroids, if you want.

All right. So you are telling if you look at your broad CNS transduction multiple brain regions with the dose of [indiscernible]. What do you think about those as a key differentiator for this next gen sort of capsid?

Yes, it's absolutely key. It's absolutely key. Depending on the CapEx and the program, we really believe that low E13 doses might be sufficient at least from what we gathered from our NHP studies, including GLP tox studies. Those capsids are extremely effective at 1, 2 E13 viral genomes per kilo with certain payloads. The efficiency of the payload can vary across the programs, the promoters and so on. But the capsids themselves, we think are extremely effective. And most importantly, at those doses, we have almost no signal at all coming from the liver. So we always keep an eye on AST and ALT, the deliver tox markers or bilirubin, there is absolutely no change in the level of liver enzymes at the doses that give us the pharmacology that we want in the brain. So this is absolutely key. Your therapeutic window is massively enhanced by using capsids. To give you an idea, the second gen capsid was the first time where we saw that we could transduce more cells in the brain than in the liver with a ubiquitous promoter and an IV dosing. We could actually reach more neurons and astrocytes than we hit hepatocytes. So that was a first even for us, for [indiscernible] capsid. And this is what we're working with right now, mostly.

Okay. Sort of how do you think about the consistency of transduction across the different brain regions?

No, absolutely. I mean those typically give a very margins distribution, especially across the gray matter. Typically, you have much less transient expression in the white matter. But across the gray matter, it seems to be extremely margins. There's always regions that are hit a little bit less than others. The heapecampus is a very well-known example among capsid engineers. The thalamus is usually transduced at very high levels. The dentate nuclei, the lateral geniculate nuclei are regions that gets targeted very, very, very efficiently. We don't know why. There's no obvious difference in vascular density or any such thing. But there's always some nuances across brain regions. But compared to local delivery, this is day and night. If you look at cell staining, like with an example on the [indiscernible] the left, it turns out that the expression of the transgene across single cells is much more margins than you would get with more local delivery methods. We really believe the vasculature in the brain works like a sponge with this really dense network and distribute the capsids in a very homogenous way across the entire brain tissue. And so these participate in the margin as expression that you see. Of course, you will never or at least not before a very long time, you will never have exactly one copy per cell across every cell in the brain. That would be just like a dream for gene therapy, but these are things that you can work with later with controlled assets and so on. Of course, the ideal gene therapy would be 2 copies per cell in every cell, just to mining what happens with chromosomes. But this is not yet something that we can do with our existing cases, but working on it.

I will say, and I'll add to the slide that you're showing, Pete. If you look on the upper the left, there is a measure of the number of vector genomes per cell out of these different brain regions. And what you can see is it's a logarithmic scale. But we're on the order of 1 to 2, and that's generally even the [indiscernible], on the order of a few vector genomes per cell. Contrast that with direct injection where you might have hundreds of copies per cell around the region where it's injected and then make gradient away from there. So that homogeneity across the brain is just an order of magnitude better if you're trying to get broad delivery than any of the local delivery methods.

And I guess that goes into the next question, which is basically how important is deep parenchymal penetration for therapeutic targets that you're pursuing?

For some...

I mean I think -- so yes, I mean we could talk a lot about the intraparenchymal delivery. I think there are some indications that Mathieu pointed out earlier, where it could be particularly useful. But generally, for most of the diseases that we want to target, we want to see broader delivery. And through a combination of the broad delivery and if you need more cell-specific tweaking it by promoters or other methods that Mathieu described as the way we tackling.

All right. Next slide. So we did talk about liver tox and also DRG. And so here, we have an example of and some data from your platform in terms of how be targeted both the liver and DRG. How does that -- how do you think about the dose levels and sort of therapeutic window?

So the liver the targeting gives you a lot more freedom about what you want to do. But in gene therapy in general and in every drug development, we're still going to -- we're going to get as close as possible to the minimal dose that provides us with our wanted pharmacology. In terms of liver toxicity, we know that we could go much higher with those capsids if we wanted to, but there's really no point whatsoever to keep ramping up the capsid dose provided that you reach your desired pharmacology in the brain. The DRG -- the DRG detargeting, sorry, is a little bit less spectacular with 102, there was some significantly lower DRG expression and better genomes. This is something that we anticipate to solve or this is something that will take advantage of the lower dose that we can use and reach our desired pharmacology. This is still the same answer. And I'm sorry if I give that answer over and over, but really, the lower dose is the key with capsid engineering. If you can reach your brain pharmacology with 2e13 viral genomes per kilo, then that automatically solves a lot of the other programs that could arise at much higher doses. If you inject those patients with 10x less vector than what is being used with [indiscernible] with or without the DRG targeting with or without liver detargeting, you would anticipate much less adverse events than what has been observed with AAV9 capsids. So this is really the absolute key is the right pharmacology at much lower dose and the translatability between preclinical models and patients. That is the absolute key to capsid engineering for us.

And I'll just add, I don't know, I mean we're probably running up against time here. But the second-gen capsid that we're using for our town knockdown program. So you're showing the first gen capsid data, but we're 30-fold detargeted from the liver, for example, with the second gen capsid. So we have a tenfold lower dose and then on top of that, we have 30-fold reduction in delivery to the liver and early targeting of the liver. And so it's really something that we lean into and think is important for the novel case.

Yes, we are running over time, [indiscernible] if you guys have it. It's okay, it's good to go through the program thoroughly. But next slide, just, I guess, quickly walk us through the rationale for going after to with the vectorized siRNA?

Sure. So I'll jump in and Mathieu can jump in where you feel appropriate. But -- the -- so for Alzheimer's disease, there is the kind of 2 process model where amyloid beta builds up is plaques that traders misfolding and spread of tau this misfolded tau throughout the brain in this very stereo type fashion. The spreads -- that stereotype spread has led to the prion-like spread hypothesis. So based on that, this misfolded tau accumulates in neurons and there's probably some toxicity as it's accumulating and when it begins aggregating. But over neurodegeneration really begins to aggregates forms [indiscernible] plaques and kills the cells. So the therapeutic hypothesis is if we can reduce the expression of tau, both starting an mRNA and then with protein, we can do two things. We can reduce the burden on a cell that is already having an issue with misfold protein, this [indiscernible] tau. But we also reduced that sales contribution to the next cell down the line, and then we reduce tau in the recipient cell, allowing us to have less of an impact. And so we're seeing this potentially play out in a non-gene therapy fashion with [indiscernible] delivered. It needs to be delivered every 3 or 6 months. But it appears and some data they released a couple of years ago in a preliminary fashion be having a pretty meaningful impact on pathological tau in Alzheimer's patients and also a surprisingly strong impact on cognitive and other clinical endpoints. So fundamental hypothesis is that we're reducing tau protein ultimately starting with the mRNA and that reduces both the burden on sort of donor and recipient cells and then treating Alzheimer's in that fashion.

The next slide, just an overview of some of the preclinical data that you've generated to date?

So one of the key models that we like to use is a mouse model. This mouse expresses human to tau. This is a case where the mouse is a P301 mouse expresses a mutant form of tau that accumulates pathological tau. What you can see in the middle, our data showing that we had expressions we got vector genome delivered in orange. And in that middle, you can see a pretty remarkable reduction in tau mRNA. This is all after a single dose of our gene therapy, tau updown gene therapy, where we're up to 97% reduction in its first tau talent in this particular case. And then we see both mRNA and then tau protein. And ultimately, that results in a 97% to 98% reduction of pathological [indiscernible] in that mouse. So that therapeutic hypothesis I mentioned, tau reduction resulting in a decrease in pathological forms of tau accumulating, that's what this model shows.

All right. And next slide, in nonhuman primates treated with is a single dose of [indiscernible] how do you interpret the level of tau mRNA reduction here? Is it sort of within the range that you believe is either for clinical efficacy?

Yes. So it is. And we can see a lot of that, not only from a lot of work we've done in the mice, but more importantly, data that are coming out of the clinic in humans with BV80 where 50% plus a little more appears to be showing efficacy as well as safety in those repeat dosing intrathecal trials. So again, what I showed you before what Pete showed you before were the mouse data. These are nonhemoprimate data. It's a single dose of our tau knockdown gene therapy, and we're looking at 5 or 11 weeks to almost 3 months after dosing at single dose. And in red, in the middle, you can see we're getting 50% to 75% knockdown broadly through the internal cortex, the HIPAA campus and the [indiscernible] frontal cortices of the monkey after that single dose. So look, we're getting a broad delivery, a single dose and reduction well within the realm of where BBB showing these results in the tumor. On the far right, we've touched on this, but that's showing that 30-fold targeting in the liver. So at this dose, and that's an important point. This is a 1.313 per telegram dose. That's well within the range that Mathieu was describing earlier, tenfold lower than Zolgensma in the clinic, and we're seeing great knockdown and this 30-fold targeting in the liver in these animals.

How -- so the data look like at 5 and 11 weeks of full [indiscernible]? How does quickly do the mRNA reduction sort of translate into protein changes? And I guess, theoretically pathological changes?

So the protein itself, tau protein in monkeys and humans has a fairly long half-life to tell me order of 3 weeks or so. And so you need to give it 5 or so half-life to begin to approach an equilibrium. That's why it was important to go out to at least 11 weeks here. We expect this to continue, if not improve in terms of reduction on the protein side as we go out to even longer time points.

And how are you thinking about the dose response relationship in terms of vector exposure, biodistribution and sort of down [indiscernible]?

So all that work is ongoing. We hope to say more about that in some upcoming discussions from our GLP and our dose finding studies. Again, this is sort of a remarkably low dose. You can see here there's a 3e12 and 1e13. We have a dose response here between those 2 doses. Given what we've seen in the mouse, one of the benefits of our ALPL capsid is its remarkable consistency between the mouse and the monkey, which gives us this relationship that we can translate between those 2, giving us greater confidence that it should translate into humans as well. So we think this is likely to hold up.

Okay. [indiscernible] ask you the last question, anything else you'd like to sort of highlight about the program or overall the platform?

Yes. So the platform is exciting. Mathieu's talked through a lot of the details. The fact that the empirical process has led us to these capsids with receptors that we wouldn't have known to go forward in the first place is really exciting. The process that Mathieu devised to look for things that across PCs played out, not only in where we see it working in both mice and monkeys for this particular capsid family, but also when we look at the receptor responsible and being able to translate that to the human form of the receptor. And for Voyager, this year, is an important year. We were getting the tau antibody data, but we will be going into the clinic with this tau knockdown and hopefully demonstrating in humans that this will translate and work well and be safe. And so it's a really important year for us, and we're really excited for the back half of this year.

Okay. Last question, the next key milestones for the patent platform over the next sort of 2 years, let's say? And when should we expect to hear about some of the other receptor candidates?

So I think the most important milestone is in the clinic, getting the capsids in the clinic. I mentioned that for this program. I also mentioned earlier, the neurocrine program, which will be moving into the clinic in the second half of this year as well. And I think those are the most important. I think, Mathieu, you might have a few comments on other milestones.

Sure. I mean I agree with you that nothing is more important at this point than validating our capsids in patients. In the meantime, the team keeps working on further improvement of capsids. As I said, we are always striving for this 13 dose in our capsids in the brain. Voyager has been doing some work with a little bit less involvement and intensity, maybe on non-CNS issues, some of which will be disclosed at the upcoming ASGCT conference in May in Boston. And we keep working on some other aspects of capsids like stealth capsids to avoid preexisting antibodies in a fraction of the patients and other aspects as well. So it's a continuous evolution. It's a process. We keep identifying new capsids, new receptors as well, some of which are feeding directly also into the neuro shadow platform because every new BBB receptor is potentially extremely important. And so the work is carrying on, no problem. There's still a lot to do. But again, 2026 is going to be absolutely fundamental for us.

Great to hear. So thank you very, very much for your time. I know that we went over a little bit, but not a big deal. I love these conversations. So looking forward to the progress. I'm looking forward to some of the catalysts, a little bit of clinical data with this platform, but as well as the other [indiscernible] blood-brain barrier protein technology. Really looking forward to that. And thank you, and thank your audience.

Thank you, Pete. I appreciate the discussion as always.

Thank you for having us.