All right. Thanks very much. Happy to be here moderating a fireside chat with Al Sandrock, CEO of Voyager Therapeutics. Al has a tremendous amount of experience and wealth of knowledge in the neurology space, so I'll try to surprise him with some other questions too. Maybe to start, Al, do you want to just give an update and overview of the Voyager pipeline efforts in the tau space, and then we'll kind of dive deeper into some of the other data from Biogen and other key questions there?
Yeah. We're focused on neurotherapeutics. We were born as a gene therapy in the CNS company. I view ourselves as a multimodality neurotherapeutics company. I'd say that we have a sort of a catalyst-rich period coming up. I'd say it's related to the first one would be the tau antibody that we expect to read out actually this quarter. Then we have FDA clearance to introduce our gene therapy tau knockdown program, and we expect to dose the first patient in the coming months. Neurocrine also has one of our capsids going into Friedreich's ataxia, so two shots on goal to see whether we can validate our newly discovered blood-brain barrier penetrant capsids, and see if we can safely dose patients and get evidence of gene expression.
Yeah, so I think those are. When it comes to the gene therapy, we expect to see acute safety early. Capsid-related safety tends to declare itself early. It'll take probably about a year after we start dosing to see evidence of gene expression in the brain. In the case of our tau knockdown program, if we see tau going down, it must be because our vectorized siRNA is being expressed in cells. Then the third readout would be eventually to see whether or not our knockdown program affects tau PET imaging. There's a lot of data that's not only come out already, but will continue to come out that will help us understand the relationship between tau biomarkers and how much of an effect you need on tau before you get clinically significant effects.
We're starting to see some of that data this year with BIIB080, for example.
Yep. Okay. Maybe let's start with the BIIB080 readout. I mean, we thought it was promising proof of concept, but somewhat confounded by the dose response, that maybe it's not an inverse dose response, but certainly they weren't getting more benefit as they pushed the dose. What do you make of the totality of the data?
Yeah. I think there's several pieces of data that I think we should look at together in addition to BIIB080. So we have BIIB080, an intrathecal ASO knocking down the expression of tau. We have a Novartis intrathecal ASO that also knocks down the expression of tau that read out this year. Then we recently had the Arrowhead shuttled siRNA, where we've seen at least the high-level data in a press release. I think all three are very helpful as we think about the BIIB080 data. So we saw what was, the trial failed on the primary endpoint, which was a dose response statistic based on CDR Sum of Boxes. If anything, yeah, there was a slight downward trend. So that's going to destroy any attempt at seeing a dose response statistic.
Nevertheless, I think overall, as you said, I think it validates tau as a target. On ADAS-Cog, we see an effect size superior to the anti-amyloids. We saw that with MMSE, the mini-mental status as well, superior to the anti-amyloids. On CDR Sum of Boxes, at the lowest dose certainly, you saw in the same range, if you will, an effect size in the mid-20% range, similar to the anti-amyloids. So let's start with the biology. When you look at CSF or fluid-based tau biomarkers, from the lowest dose to the highest dose, there wasn't much change. At the lowest dose, it was a 40% reduction. At the highest dose, it was 50%. So you're at the flat part of the dose response curve in terms of tau knockdown. But I think what happened was that what we saw was a dose-dependent increase in a transient acute confusional state.
I think that, I think, is due to intrathecal ASOs having off-target effects. We've seen that with other intrathecal ASOs, several others. Ionis actually published this year papers on transient neuronal inhibition as well as neuronal activation phenomena that can occur with oligonucleotides regardless of sequence. So it's probably related to having a highly negatively charged oligonucleotide, high extracellular concentrations. Remember, you have to give relatively high doses intrathecally in the lumbar cistern if you want to get enough exposure up in the brain, because there's a pretty steep gradient from the lumbar cistern up to the cortex, and even more reduction as you get into deep cortical structures. So to protect those exposures up there, you got to give excessive amounts down in the lumbar cistern.
I think that what we've had was, you maxed out on tau reduction with the ASO because ASOs, they're okay for knockdown. siRNAs are a little bit better because they're catalytic, but you couldn't do much more than 50% knockdown. But the untoward effects, these transient neurological issues degraded, I think, the primary endpoint, which getting confusional state in a patient with Alzheimer's disease is bound to affect
Of course.
endpoints, right? Also, there's this peculiar thing where with tau-related drugs, it seems that you have a bigger effect on cognition than on function.
Yeah.
Activities of daily living did not look that great in BIIB080 either. Maybe that's related to the acute confusional state. What I wonder is it masking a potentially even bigger effect if you could achieve, say, 70% knockdown? You can't do that with an ASO because now you really have problems with higher doses, right?
Right.
I wanted to point out the Arrowhead data. Before we go to Arrowhead, let's go to Novartis. They have an intrathecal ASO, also injected into the lumbar cistern, that was positive in PSP, so much so that they are moving it into phase III. They also report encephalopathy, a transient encephalopathy, which I think is akin to this acute confusional state. So two intrathecal ASOs, and I do not think it is related to tau. I think it is related to these transient neurological effects that we have seen with many others. We even saw with Spinraza in the early days.
Yep.
Let me get back then. Arrowhead is a very different approach. It is siRNA, but it is not intrathecal. It is systemically delivered, shuttled, and they did not report an acute confusional state or anything like it, or encephalopathy. So I think that that is I think, I really do think it is intrathecal ASOs, and that gets back to VY1706. Our approach is more akin to Arrowhead. We get broad distribution. We do not have steep gradients because we have a BBB-penetrant capsid that we are using. Actually, we get no extracellular concentrations of siRNA because we make the siRNA in the cell. The gene therapy is an intracellular production of siRNA. So, I think I find all this kind of encouraging for our programs.
Yep. With the BIIB080 data, there is a much larger effect on CSF tau versus tau PET. I am wondering to what degree do you think that that is a byproduct of intrathecal administration inflating the CSF biomarker versus the possibility that tau PET is kind of rate-limited to how much of the protein the brain can clear, and that in itself is never going to be nearly as high as the knockdown you can get?
Yeah. It could be that there is more than one pool of pathological tau in a cell. It could be that there is a more reversible pool of that pathologic tau and a more irreversible pool. If you look at, for example, neurofibrillary tangles, it is hard to imagine that much of that is going to be reduced by just inhibiting the synthesis of new tau. They look like pretty darn permanent structures, if you will. On the other hand, I think pathological tau can exist in what I would call condensates before they become involved in NFTs. I think that probably is reversible. I suspect that with tau PET imaging, we are going to see pathological tau in either sort of phase, if you will, in either form, and that we are only going to be able to see the reduction at least by blocking synthesis.
Now, if there's a way to remove tau from the NFTs, I don't know of any right now, but I noticed that this morning there was a protein degrader company that is-
Yeah
that is targeting tau.
That just raised money, yeah.
I don't know if that's going to affect the NFTs, but anyway, so that's my interpretation.
Yep. Okay. As it relates to the gene therapy approach, first off, I guess talk about what makes you comfortable with irreversibly knocking down tau, that this will be safe.
Yeah, that's one of the reasons why we were waiting for BIIB080, right? In our animal studies, we can knock down tau even to 70% without seeing any untoward effects. There are genetic variants of tau where you get loss of function variants in tau that are tolerated in humans quite well. As you know, Paul, because you've written about this, tau is a microtubule-stabilizing protein, right? There's lots of microtubule-stabilizing proteins. There's redundancy in the system. Whereas I think tau plays a pretty important role during development, I think in the adult, the redundancy allows us to reduce tau by certainly 50%, maybe even 70% without any problems. I think BIIB080, with the patients were followed for 18 months, the transient confusional state notwithstanding, I didn't see any other long-term effects that are concerning, right? I think I'm pretty comfortable.
Alzheimer's patients arguably have declared that they have too much tau, and it's a fatal disease. We forget sometimes that Alzheimer's disease is a terrible disease, and it's fatal. There are patients who still die from Alzheimer's disease. Thousands of patients with Alzheimer's are dying every day. Even though we have the anti-amyloids, a lot of people can't take them. If you're an APOE4 homozygote, you're probably not going to be prescribed anti-amyloid because of the risk of ARIA. By the way, when we talk about verubecestat and we say the CDR is on the same range as anti-amyloids, they achieve that without any risk of ARIA. Even heterozygotes with APOE4 carrier, which there are a lot of heterozygotes who have Alzheimer's disease. Many people are not being treated because of the risk of ARIA. It's higher in even heterozygotes, right?
Anyway, I think tau is the next big thing for treating Alzheimer's patients with a bad disease, and I think we should try to figure out a way to help those patients.
Yeah. With the gene therapy approach, how are you thinking about the therapeutic index for capsid-related safety issues? Maybe if there's any context you can kind of give around the dose range that you are planning on studying and how that kind of compares to where we saw tox for ELEVIDYS or some of these other AAVs that ran into liver or HLH or things like that.
Yeah. There's two things that give us hope with our capsids. One is that we're using a lower dose. Most, all to my knowledge, of the systemically given AAV treatments are in the 1E14 vg/kg range. Our highest dose is not going to be higher than 5E13 vg/kg. We are at a lower dose. Second is we have a capsid that detargets the liver 30-fold relative to AAV9. As you know, patients have died from the Sarepta gene therapy from liver failure. The fact that we detarget the liver is also helpful. Now, we will use immunosuppressive treatments during the time of administration, but our tox studies were quite promising on the safety front, and we had a very robust discussion with FDA.
We had a pre-IND meeting that was more than a year before we submitted the IND. I went back, we decided to go back to FDA after the capsid patient died to say, "What did FDA know about what happened to that patient?" So we can incorporate their feedback into our GLP tox studies. We did so, and we made additional measurements, and then we submitted our IND. We had a robust discussion with FDA. I'm pretty confident that we are doing the very best we can in the interest of patient safety.
Yep. Okay. Makes sense. Maybe talk about the planned timing of the disclosures from the gene therapy data. You said obviously early data insight into safety, and then maybe the next two sort of layers of PD insights that you'll get.
Yeah. The first is, we think, Q1, we still think acute safety. Much of it is going to be capsid-related, right? I mean, because capsid issues occur early. Then about a year after we start dosing, we will measure CSF tau PD, and then sometime after that, and we haven't even guided to when we would get the tau PET image. That would tell us whether we have an asset that can help Alzheimer's patients. Does that answer your question?
Yeah, it does. I think, maybe to just drill down on that, what have we learned from other programs in the degree to which PET lags knockdown?
Yeah. You-
It feels like that's driven by the actual brain versus the actual modality. Is that fair?
Yeah. Well, yeah.
Like whatever the endogenous thing.
Well, listen, I mean, when we look at messenger RNA, we knock it down rapidly, but tau has a slow turnover, so we have to wait for tau. But in our studies, for example, at three months after we give our gene therapy, we see tau going down in non-human primates, then we see further knockdown in six months. But then I think for tau PET imaging, I think you need at least a year. What's responsible for that lag? Probably due to sensitivity of tau PET imaging, but also the PET signal grows. So there's accumulation in the region at baseline. When the patients come in, they have certain regions of the brain that already have tau, and the PET signal grows within that region, then it spreads to other regions, and that's a cell-to-cell spread likely.
That takes time to occur, therefore it takes time to see an effect on those.
Yep. Okay. Makes sense. Maybe last question on the gene therapy side, and we can talk about your antibody as well, but as you think about getting a product like this to market, most innovation we've seen in the gene therapy space has happened in rare diseases where there's greater flexibility around the size of the safety database or the trials you need to do. Is your assumption that for the gene therapy you're going to kind of have to operate and run more of a lecanemab-like program, or is there an in-between potential here?
I think there's an in-between potential. On the other hand here, first of all, Alzheimer's disease is so common. If you can do a once-and-done IV treatment that can lower tau, and let's say, and we've seen that 50% lowering should be sufficient, wouldn't that be great? Because if you look at SMA, every infant gets ZOLGENSMA in this country, right, shortly after birth, right? That's like the baseline treatment. Then if patients need more of an SMN protein-inducing treatment, they get either Spinraza or risdiplam. I'd say a lot of people are getting additional treatments beyond the gene therapy, but the gene therapy provides sort of the baseline.
Yeah.
I could easily see that happening in Alzheimer's, because to treat millions of people with intrathecal ASOs or even subcu-delivered siRNAs is-
It's a lot.
It is a lot.
Yeah.
I do think it is going to have to be larger than the rare disease trials.
Of course.
It is going to have to be powered to see an effect on probably CDR-SB, so it is not going to be small.
Right.
We're not also, I would say that there's improvements in cost of goods for AAV therapies. Every year as companies scale, the cost of goods is going down, and we're not going to have to charge the kind of, I mean, look at the prices people are charging for rare disease because you have to make ends meet, right? With Alzheimer's disease, there's volume.
Right.
We can be commercially successful without charging a high price.
Yep. Okay.
So-
Makes sense. On the antibody side, obviously that's an approach with tau and to some extent synuclein, where there's been this overarching question around attacking an intracellular target. There's been a number of failures for each, a little bit of promising data, but just maybe in the context of that, talk about your antibody, why it's differentiated, and why you think that approach still has some merit.
Yeah. I call it the dark horse because nobody thinks it's going to work, but I think there's a decent chance. Otherwise, it wouldn't be in our pipeline. It's differentiated because, first of all, our antibody's specific for pathological forms of tau. In the anti-amyloid space, that turned out to be pretty darn important, partly because you don't want antibodies binding to normal tau and essentially be a sink for the intended target, which is pathological tau. Second is that it worked in an animal model that I have relied heavily on. We had literally hundreds of antibodies against tau. How do you choose the right one? We relied heavily on an animal model that Steve Paul, who was the original CEO of Voyager, brought in from Cornell.
This is an animal, this is a mouse that expresses human tau, the P301S mouse, and you inject paired helical filaments from Alzheimer's brain, human Alzheimer's brain, into one region of the brain, usually the hippocampus. Then you look for accumulation in the ipsilateral hippocampus of pathological tau, and then you see it spread to the contralateral hippocampus transsynaptically. The belief there is that the biology that underlies that accumulation and spread is similar to what's going on in humans, because in humans, that's the characteristic of Alzheimer's. When we talk about Braak staging of Alzheimer's, it's based on where tau has spread, not where amyloid is.
Yep. Nope.
This model correctly predicted that the N-terminal antibodies would fail. The N-terminal antibodies that failed in the clinic failed in this model. It correctly predicted that bepranemab would block the spread of tau, did so by about 50%. You'll recall, bepranemab was positive on ADAS-Cog in the whole population, a pre-specified secondary endpoint. It was not a post-hoc analysis. It's failed on the CDR-SB, but there was a trend. By the way, UCB is moving forward-
Right
with bepranemab. It also correctly predicted the failure of the J&J antibody, which we learned about roughly a year ago. We actually had an antibody very close to this epitope that J&J targeted. We did not pursue that because it didn't work as well as our antibody in that model. So, so far, that animal model is four for four in predicting the human clinical, three on the negative side and one on the positive side. So hopefully we're five for five, when we announce our results.
Yep. Do you want to talk about study and data timing there?
Yeah. It's, we've said Q4, so any day now. And it's the study's powered to show an effect on tau PET imaging. It's a multiple ascending dose study where we are dosing, and we haven't disclosed the actual doses, but we did a single ascending dose study previously. So we know the dose that we need to give to achieve the exposure we need to get in the CNS based on CSF sampling. And so we know we're well above the EC50, or I guess here it's the IC50, for affecting tau spread. So, hey, you know what? I'm an eternal optimist, as you know, Paul, and I've been there before when everybody says it's not going to work, and sometimes things still work anyway.
Right.
Hopefully, we'll be lucky again. Patients need something, you know?
Yep.
Patients need better treatments for Alzheimer's. We are far away from a cure.
Yep. As you think about tau PET, are you hoping to see a reduction or is even like slowing the accumulation
Yeah
good enough?
I do not think we will see a reduction, because I think we are going to slow the accumulation and spread. We have bepranemab as a benchmark. We need to be at least as good as bepranemab or better, which is a pretty tall order. They had about a 50% effect, so we want to be north of that.
Yep. Okay. If that is positive, what are the next steps there? Do you see yourself developing that independently?
Well, we've always said we need a partner.
Yep.
For Alzheimer's, we need a partner at some point.
Right.
Right? Because it's too big even to do late-stage trials.
Could you take it any farther yourself, or do you need a partner?
Yeah, I think we probably could. Well, look, as you know, I'm stepping down as CEO in the next few months, and Robin Swartz, who's my successor, is sitting over there. That's going to be her decision. I'll help her. I'll be on the board with her decision. I think there's a lot of opportunities here. I think we could potentially move to a phase II trial. We could shuttle it with our ALPL shuttle. I think our ALPL shuttle would be perfect for that drug, actually. We always talk to potential partners. As you know, we've done a lot of deals, Paul, and I'm always open for business. There's some people who've said, "When we get data on that, come back.
Right.
We'll talk to people, and so there's lots of options there.
Makes sense. On the ALPL shuttle, it's just super interesting to have an alternative to the TFR1 approach. How does ALPL compare? For your shuttle, I guess this was sort of discovered, right, where you sort of screen capsids, found out they got across the BBB, and then kind of backed into how they're getting there. Are you comfortable that it's just ALPL that was sort of explaining that finding, and that you can really isolate it to that and then leverage that pathway?
Yeah. If you take out ALPL, our capsids don't cross.
They don't.
Yeah.
Okay.
We've made shuttles against ALPL, and they do cross. We plan to share a lot of this data. We have done in vivo pharmacology in mice and in non-human primates, and we're going to be sharing that next month at the Antibody Engineering meeting in San Diego. I can tell you, when you think about shuttles, there's three things to think about: distribution in the body, pharmacokinetics, and safety. Each shuttle is going to have it be distinguished on those three fronts. ALPL has a much more limited distribution in the body. As you know, TFRs will shuttle things into muscle as well as brain and to several other tissues as well.
Right.
With ALPL, it's much more restricted.
It is more biased to the blood-brain barrier.
Yeah. That allows it to have a much longer half-life. So if you have it in the blood for longer, there is more opportunity to keep getting into the brain, if you will. Third is that it has no effect on any of the hematologic issues that we have seen with TFR. Now, patients who have complete loss of function mutations in ALPL, or severe ones, have bone mineralization problems. So you can bet we have looked very, very carefully at some of those parameters. Luckily, there is very good ways to assess that in animals, so we will be sharing much of that in December.
With the TFR1 space, companies have found ways to at least better bias their antibodies towards receptor expressed in a given tissue type. Can you do that with ALPL, avoid bone?
We have not explored that fully yet. I bet there-
Do you feel like you need to do that?
I don't think we do
Okay
right now, so far.
Okay.
Because as I said, in humans, you have to have pretty severe loss of function before you see any problems with bone. And we're going to share some data that suggests that maybe we don't need to.
Yeah. Okay. In the last minute here, Al, anything else you want to add about Voyager and the prospects ahead?
Catalyst-rich period coming up. After all, when I came, we had just discovered these capsids. We are finally at the point where Neurocrine is taking one of these capsids into FA.
Yeah.
We are taking one of these capsids into Alzheimer's with tau. Boy, if that works, as a neuro drug developer, to be able to cross the BBB with a capsid or with a shuttle is, boy, I am salivating at the prospect. Do not forget the dark horse, this antibody.
Nobody expects it to work, but sometimes things still work anyway.
All right. Well, any day. We look forward to it.
Yeah.
All right. Thanks, Al. Appreciate it.
Thank you, Paul.