All right, everyone. Welcome to the next session. It's titled Gene Therapy for Neurodegenerative Diseases. Are you there yet? It is my pleasure to introduce our next panel of speakers. From my far side, we have Zandy Forbes from Meira Gene Therapies. We have Matt Kapusta from uniQure, and we have Will Chou from Passage Bio. Thank you for joining us. Thank you, everyone, for being a part of our conference today. To get things started, if each of the panelists would spend a minute or two to introduce your companies and specifically talk about the program relevant for this panel, neurodegenerative disease, most recent milestone, and what we should be expecting next. Zandy, I'll start with you. Yeah, I'm the CEO of Meira, which is a genetic medicines company that was set up not just to treat inherited diseases, but much larger indications. Crucially, we only treat with very small doses locally delivered. We avoid systemic doses, which makes our drugs, A, much safer. In addition, our cost of goods is very low. We have our own manufacturing from process development, plasmid facility, to viral vector facilities, and the QC facility. All our manufacturing is in-house. We have two commercial licenses, and we're very well known by 15 regulatory agencies. For example, the FDA wrote to us during one of our phase II studies using our material for Aquaporin and told us that they considered it pivotal because we manufactured it. All of our late-stage studies, two pivotal and two awaiting BLA, are supported by our own manufacturing. That's one of the reasons we've been able to develop them so quickly. One of the reasons that all our programs have been successful is because of this approach of very local delivery of small doses that have a physiological change on a particular severe indication. Two of those are in the eye, one very rare disease, one more common disease. Those are awaiting BLA filing and European filing. One in Parkinson's, which does not address dopamine, is the only treatment to have three positive studies and two positive studies UPDRS against sham. That's never been shown for any gene therapy or cell therapy or growth factor therapy before. We've recently done a partnership in our Parkinson's program with an AI company out of UCL and Wellcome that's been developing over the last 10 years. Using that technology and 20 years of data from Queen's Square, the largest neuro-hospital in Europe, we were able for the first time to be able to show that in our phase II double-blind data, looking at the FDG PET scans, we had, in fact, physiologically changed the brain with our treatment that turns glutamate to GABA in specific nuclei of the brain. We changed the circuitry physiologically, and we had also changed the input to the Nigra in a way that was neuroprotective. Using this AI, which removes heterogeneity between patients and scans, we were for the first time able to see what we consider real disease modification in Parkinson's disease. I won't go on about the rest of the company. You go on. Yes, please. OK. Matt Kapusta, CEO of uniQure. uniQure is one of the early pioneers in genomic medicine. The company was actually founded, believe it or not, 27 years ago. Over that period of time, we've been fortunate enough to be part of sponsorship of two approved AAV gene therapies. The first was a product called Glybera in 2012 for lipoprotein lipase deficiency, the first ever approved AAV gene therapy in the world. The second was in 2022, a product called HEMGENIX, which was the first approved gene therapy for hemophilia B and has been partnered with CSL Behring. The relevance for this panel is our lead program in Huntington's disease. Huntington's has been a labor of love, love and sweat, I should say, for almost 15 years. I joined the company at the end of 2014. At that time, the program was three or four years in preclinical development. What we were trying to do was usher in a new period where most of the AAV gene therapies at that point in time were replacing something that was missing. That's similar to what we're doing for hemophilia. You have a missing factor IX gene. We're going to impregnate that cell nucleus with a construct that will produce that missing enzyme. We had developed a particular platform where we wanted to suppress something that was aberrant, something that potentially would lead to a gain of function. We developed a platform called miQURE, which has a proprietary scaffold associated with it that we can deliver a microRNA that binds to and degrades messenger RNA and suppresses an aberrant protein, which is precisely what you want to do in Huntington's disease, which is a CAG repeat disorder where you have a toxic gain of function associated with a protein called the Huntingtin protein. What we wanted to do, similar to what Zandy said, is it was very, very important to us that we would be able to make this drug bioavailable in the key areas of the brain. For us, that is the big problem that has vexed the industry for decades, is how do we deliver a drug into the brain that can be bioavailable in the key areas that are implicated in the pathology of disease? We really developed a precision targeted-based infusion deep within the structures of the brain where HD is known to manifest, where we can see under real-time contrast-enhanced MRI the filling of those key structures, and then really studied what happens once we deposit it there. Where does it project to? Really determining that it's covering the areas that ultimately spread with the advancement of Huntington's disease. All of this took about six, seven years. We started clinical testing about six years ago. We've treated thus far, what we've disclosed is 45 patients with AMT-130. We finally got to a point in time about a month ago where the majority of those patients had achieved three years of follow-up data. We saw really remarkable findings that were, I think, best case scenario internally and I think for the community, where we saw a 75% statistically significant slowing of disease progression based on the composite UHDRS. We saw also a statistically significant reduction or slowing based on total functional capacity. These are provable endpoints. We saw all the subdomain measures also with very favorable trends. Neurofilament light, which is an objective marker of neurodegeneration, was below baseline, which was unprecedented. We are very excited about that. We are expecting to hold a pre-BLA meeting in the fourth quarter and submit a BLA in the first quarter of next year and look forward to talking about that program in more detail. Excellent. Thank you, Matt. Will? Congratulations on the data, Matt. Thank you. It's great. Appreciate it. Will Chou from Passage Bio. We are a company formed out of technology from the Gene Therapy Program at the University of Pennsylvania. The key programs that are relevant for today, we have a lead clinical program in frontotemporal dementia patients with the GRN mutation. These patients are haploinsufficient in the GRN gene. Therefore, they have not enough progranulin. We deliver our AAV1 via the intracisterna magna route. It's about a one-hour procedure done in a CT scan by an interventional radiologist. The data we've shared to date looks pretty differentiated in a crowded, exciting space. We've shown very high durable levels of target engagement of CSF progranulin. We've shown the ability to stabilize plasma neurofilaments. We're looking forward to sharing more data next year. The other program that's relevant is a preclinical program. We have in Huntington's disease also. We have not yet disclosed the specific target, except we are going after the DNA damage response pathway. This one will be administered intraparenchymally, as Matt said. In Huntington's disease, you have to get into deep areas of the brain. It's a little different than FTD GRN. Excellent. Thank you, everyone, for the great overview. I think what you mentioned really highlights why, at least in your cases, the neurodegenerative diseases are a great target for gene therapy and highlight what could be potentially achieved with those. That brings me to the next question. What do you think are the biggest challenges specific to gene therapy for neurodegenerative diseases that still remain? Where do we stand in overcoming those challenges? We'll start with you here. Sure. One of them, as we mentioned, is getting it to the right spot. If you have something like a disease like Huntington’s disease where you really need to get a concentrated amount of vector into one spot, it is a procedure. A procedure is one option. Of course, cross-blood-brain barrier would be great. There’s still a long way to go in terms of making sure that that is safe and available for a lot of patients. I would say with any procedure, there’s time and complexity involved. Excellent. Got it. Matt? Yeah, I think for me, I completely agree that delivery is definitely challenge number one because you can have the most beautiful construct with the perfect target, and if you just can't get it there, it's all for naught. I will say one of the biggest challenges, because we've had to deal with this with HD and maybe some of you all as well, is that many neurodegenerative diseases are slow progressing. With a neurodegenerative disease, you can't, or it's very difficult to restore function, so you're looking at preventing decline. That is a very different thing to manifest in a slow progressing disease. This could be any degenerative disease, but in neurodegenerative diseases that are slow progressing, it's very difficult to design a clinical study that is going to detect changes on a slowly degenerative disease. I think that's been a big problem for things like Alzheimer's. I think it's presented problems in Parkinson's disease, where you can design a study that can actually manifest or suss those things out. I think the advantage that a gene therapy has is that it's one time administered and that you can, by default, even in early clinical settings or early clinical testing, you can follow patients for an extended period of time. If you're fortunate enough, like we are in Huntington's, where there's a substantial amount of natural history data, you can much more readily follow patients long term and elucidate out some of those effects in a way that is different for chronically administered treatments. I think that that's a challenge. In some respects, it can be better addressed with a one-time administered gene therapy. Got it. Thank you. Zandy? I agree. Both delivery and the long-term decline in some, if not all, neurodegenerative diseases are both challenges. However, it is possible to find very severe unmet needs in neurodegenerative disease where you can treat by changing the circuitry of the brain. If you look at the physiology of a number of different neurodegenerative diseases, rather than the etiology, what protein is causing the initiation versus what causes the degeneration, in many cases, hyperactivation or the loss of repression of glutamate-producing neurons drives faster and faster degeneration. You can locally, in different diseases, as we've done in Parkinson's, convert in tiny nodules glutamate to GABA. GABA is a neurotransmitter that calms the brain and, in fact, slows degeneration. It gives you the opportunity to identify loci that allow you to change circuitry in neurodegenerative diseases, such as Parkinson's, that allow you to slow degeneration at the same time as restoring function. I've described Parkinson's. It turns out that neuropathic pain is very similar, hyperactivation from inflammation of glutamate. We can use our gene therapy to locally deliver to a ganglion that is causing trigeminal neuralgia, for example. It converts glutamate to GABA. Rather than using systemic Lyrica or systemic gabapentin, we locally produce out of the hyperactive glutamate GABA and calm down that nucleus. Just with that one enzyme, there's a huge number of different neurodegenerative diseases that we can treat in a slightly different way. With our AI partnership, we can start to identify exactly the nuclei of hyperactivity in very severe diseases, including OCD, by the way, incredibly severe in a very small number of patients. There are certain nuclei that you can convert from glutamatergic to GABAergic and relieve the symptoms of OCD. It's a different approach. It's using gene therapy like an intervention rather than a drug. It's a one-time treatment. It's often used when ablation is the only alternative and can result in paralysis and a whole load of other problems. It physiologically changes that hyperactivation to something calming and resolves whatever symptom it may be. There are ways of getting around both delivery and endpoints, because these are quite rapid endpoints that you see when you change physiology. That allows you to treat multiple diseases in different ways that haven't really been addressable locally before. Got it. Zandy, to follow up on that and that partnership with Halogen, is that something that you are actively pursuing, actively investigating as part of your pivotal program that's about to enter the clinic? Is that something that you plan on implementing further down the road? No, we've used their technology to look at our phase II blinded data. We are using their technology to increase the robustness of our primary endpoint in our pivotal. That's UPDRS. Essentially, it's a slightly more advanced type of AI from ChatGPT, everything OpenAI does, which is essentially it looks for identity. If you have the current large language models, they look for a perfect match for the thing that you've presented. You present them with an apple. They look in the data to find a perfect apple. In life, no living entities are identical to each other. You can't be identical in life because monocultures die, bam, you're obliterated in time if the environment changes. There is heterogeneity in life. These large medicine models are the first models that don't treat heterogeneity as noise, like the large language models do. They're able to model the natural heterogeneity in data from life and remove it from noise, allowing you to see real physiological changes. We looked at double-blind data from our phase II and saw these physiological changes. We have used those models to, when you model what should happen in your study, create things called supercovariates rather than multiple covariates. We ran this by the FDA. They seem to think what has been proposed is acceptable. We have also run by the FDA the use of AI in analysis of imaging, in removing bias from large sets of imaging data. The FDA seemed to think that that's also a reasonable approach. Both of those things, both of those aspects of the large medicine models, I'll call them, are being used in our pivotal to analyze the pivotal. What we don't do is what people think you use AI for in a pivotal study, which is choose the one or two patients it's going to work in. That's not helpful because it's not helpful to go and use your current large language models to pick the five patients that are going to respond. What you want to do is remove all the noise so you can see if your drug works or not. That's what we do. Got it. OK. The question of heterogeneity in neurodegenerative diseases is the big one, right, for FTD, for Huntington's, for Parkinson's, for other indications as well. Different patients have different rates of progression and so on. Given that complexity that stems from heterogeneity, what are the fair expectations for gene therapies in those indications? Maybe how should we look at different patient groups at earlier stages, more advanced stages? What can we expect gene therapies to achieve in those patients? Let us start with you. Sure. I think an important thing in terms of a barometer of effectiveness, as Matt said, in adult neurodegenerative disease, it's a long lead time to see clinical changes. To the extent that natural history datasets, whether it's in Huntington's or in FTD, can build up good evidence around biomarkers of progression, be they MRI of whole brain or specific brain regions or neurofilament levels, I think using those early indicators of the product is going to be effective or is showing signs of effectiveness, that's going to be really important for the future bar. I do think there, I know we're going to get into regulatory discussions. I do think there is an increased appetite in understanding from regulatory agencies like the FDA in giving latitude for things like biomarkers. What we've seen from the current agency is an emphasis on giving patients the opportunity to try when there's no other opportunities out there, when there's no other therapies. I think the bar is going to change as more natural history data is developed. Got it. Matt? Yeah, I mean, for me, the biggest thing as it relates to heterogeneity is having access to natural history data. As an example, we enrolled 45 patients in our phase I two program. We have an inclusion criteria. Everybody's got it. Even within that inclusion criteria, there is heterogeneity in those patients. What we have tried to do, by the way, even if we had done a randomized controlled blinded study, you're just not going to be able to enroll the large number of patients that can help reduce that noise. That's really the other way to do it, you enroll a large enough number of patients where that noise gets neutralized or canceled out. When you have a relatively small number of patients, you need to be able to somehow match them or compare them against something. For HD, we looked at a number of factors when it came to trying to generate an external control. We did a statistical methodology called propensity-score matching, where we developed eight prognostic factors that are known to be correlated with how disease progresses within a particular individual. Then we're able to assign all of the patients in a natural history database a propensity score and then match them individually to a number of patients that are highly similar to them. You're essentially creating a digital set of twins through a natural history that helps reduce that heterogeneity bias. It's not perfect, but it's an excellent tool that regulatory agencies are beginning to become more flexible in assessing, particularly for diseases of high unmet need. OK. Zandy? This problem of heterogeneity is exactly the problem that we found so compelling to have been solved by this AI, which came out of Wellcome and UCL over the last 10 years. What made us so excited is that we'd done a double-blind phase II study, 40 patients, approximately 20 in each arm. UPDRS was positive. We looked at scans, and there was some sort of difference in the pattern. You could tell a computer to find a difference in the pattern, and there was some difference. We gave the AI that double-blind data. This is trained on all the data over 20 years, every scan, every blood test, every patient visit from Queen's Square, largest neurohospital in Europe, and longitudinal, everything. Massive, massive data. That is like looking for your twins, but millions of pieces of data. We gave them it, and back came double-blind data. They showed actually the physiology of the brain had changed. The nuclei's function was changing. The input to the Nigra changed. Completely clear, massive statistical significance. Based on that huge database of every neurodegenerative disease of all Parkinson's patients that were in that database, they could see what was heterogeneity, like millions of twins, versus what was an actual real change in the physiology of the brain that was shown on the scans. To us, that was a super powerful technology, not just for the brain. They initiated this technology using the brain because it's the most invisible thing, but you can use it on blood, for immunology, for cancer, to really take out heterogeneity between patients, between diseases. It's very, very powerful for any data that comes from life or the clinic. OK. Got it. Thank you. I wanted to talk about safety too. We can't talk about gene therapy and not mention safety. Obviously, that's always a concern. We had a recent set of patient deaths, one from a capsular synergy trial in neurological disorder. What read-through do these safety signals have to your programs? Since you are approaching your indications through a local delivery, do you think you're largely immune from those systemic safety effects? Matt, let's start with you. I think I speak for everybody. I think any patient death in any study is a patient death too many. We do have to recognize that these are absolutely terrible diseases that we're battling and that the prognosis for many of these patients is also equally terrible. I do think that for any disease, there's a risk-benefit equation. The reality is that when you're, for us here, maybe not every one of our programs, but for us here, most of our programs are viral in nature. Those programs are going to interact with the immune systems in a way that are hard to predict. There's also millions of particles being administered, and they can be inefficient. Not every single one of them is getting into the target structure. I explained to you why we went direct. We went direct predominantly because of bioavailability. It is an advantage, I think, when we talk about read-through. It is an advantage because the systemic exposure, I mean, we're administering 3 ml of product. The amount of systemic exposure associated with that kind of administration is substantially lower. The truth is most of the acute organ toxicity events that have led to death have been systemically administered therapies that can lead to these complications. I do think we have to remember that AAVs come a long, long way, but there's still unknowns about what are the factors associated with certain patients that lead to these inflammatory or immunological events. There is an elegance to precision-targeted delivery in knowing that you're getting the drug to the place that matters and to other healthy tissue, you are sparing them. Yeah, there are certainly unique aspects of every patient that you can't predict. What we can say about AAV safety is that there is a level of dose responsiveness in terms of safety. The total genome copies have been in the 10^15 range, whether administered directly to the CNS or systemically. If total genome copies are higher, one of the advantages of going directly to the CNS, like we're doing with ICM, is we can give much lower doses. Our doses are 50x to 100x lower in total vector. Yes, you are still going to get an immune response. Yes, it will be idiosyncratic, patient to patient. You're giving almost 100x lower dose. The opportunity for major adverse events from an inflammatory response is far lower when you're giving lower doses. I would not just read through those safety events from some gene therapies to all. Got it. Makes sense. Yeah, I agree. I mean, safety is arguably, when you start studies, the most important thing you're looking for. We've done dose-ranging studies. We've done two sham-controlled studies. We deliver a very low dose. For example, total viral dose in 50 μL is in the E10s. It's AAV2, which we use because it's really sticky. It doesn't move. Both of those two things maybe are reflected in our incredibly good safety. In fact, in our phase II, our treated arm was overall safer than the untreated who just got a hole drilled in their head. The reason being is the safety issues were worsening Parkinson's. The dose, the site of delivery, the type of capsid make a big difference. Looking really carefully at safety is super important. I agree that smaller doses, more locally delivered, tend to be much safer. AAV2 doesn't move. That makes it even safer. OK. Makes sense. All right, let's shift gears and talk a little bit about regulatory aspects. We touched on this a little bit. We talked about the complexity. We talked about the progressive nature and variability within the disease, which make some of those indications either insufficient or inappropriate for traditional trial design. We have to think outside the box, so to speak, and implement novel trial designs, novel endpoints, biomarkers, and so on. As I mentioned, we talked a little bit about a few of those. Maybe let's spend a couple of minutes and summarize some of the changes that the FDA is open to adopting and the sponsors begin using. Let's start with you. Sure. I'm going to let Matt talk about the great news from Gary from uniQure. That was my best example. You should talk about that. I would say the other really great thing we heard right away from this new FDA was wanting to continue, I guess, the Peter Marks ethos of looking at rare diseases. We heard that in words, and then we saw last month the rare disease evidence principles that they laid out. If it is a rare disease with high unmet need that is very rapidly progressing and you have a therapy that actually treats an easy, there's a low protein, you're replacing that protein, you have an opportunity to ask for something like a single-arm study with supportive evidence. That has always been something that we all thought. They went and they put it on paper. They gave us bullet points. They put it out there last month. It's a very strong signal. We've also heard that in our interactions with the agency, which is, hey, we know you're in a rare disease. We know this is gene therapy. We don't look at you in the same way as we would, say, a monoclonal antibody in a large therapy area. I think those are all very positive, tangible responses that we've seen for the unique situation we're in. I mean, Matt's situation is the best. This was a great model. Your interaction with the FDA was a great model for all of us. Huntington's isn't even as rare a disease as frontotemporal dementia with GRN mutations is. Yeah, no, it is very, very difficult. We're tackling rare diseases with novel medicines, with novel clinical development structures in a highly regulated space. It is not easy. Sometimes when you have your back against the wall, you really feel like, what do I have to lose here? I think we have felt that doing a proper randomized, Roche did a 700-patient randomized controlled phase III study. We just said, look, we're dead in the water. Who can do a study like this? It was very clear to us that that was not going to be a viable pathway. It really takes a lot of courage from regulators to recognize the substantial unmet needs in these diseases and the need for flexibility. Every single one of us would be honored to have a chance to take advantage of that flexibility and be committed to providing confirmatory evidence to support therapeutic benefit. If we're not able to justify that, the products are taken off the market. When you have slow progressing diseases, neurodegenerative diseases, where in many cases there are very difficult biomarkers, many of which are not validated with subjective endpoints, you just cannot apply a classical gold standard randomized controlled double-blinded framework. I mean, you can do it. There are going to be feasibility implications and ethical implications. If you enroll an HD patient and you follow them for three years and they're on control, by the time they exit that study and become unblinded, they're unlikely to be a candidate for a disease-modifying treatment because they've advanced. There are real practical issues that patient advocates have come out against. I do want to provide you time. I'm just really delighted to be in a leadership position and to help create a pathway for others to take advantage of because I truly believe in it. Yeah. Zandy? I was so engrossed in what you said. I forgot the question. We've only got 35 seconds. I'll leave the last word to you. OK. Someone else. Maybe we'll end on another question in terms of access. With your program focusing on local delivery, what would you say are the greatest limiting factors for broad access? Is that the number of patients, the number of treatment centers, the number of surgeons who are qualified to do these surgeries? What are the biggest challenges here? Let's start with you. Sure. All of those are normal launch challenges for any product that is administered in a center of excellence. I think the key thing is if you have great data, people will be willing to be trained on it. If you guys have great data, I'm sure people will be willing to train on it. Yes, the more complex the procedure is, the more difficult the uptake is going to be. It is going to be more of a hurdle with different stakeholders, whether it be hospitals or different doctors. If you have good data, people will come, especially in something like Huntington's or frontotemporal dementia with GRN mutations. Our Parkinson's treatment was designed by the Vice Chair of Surgery at Cornell. The surgery is identical to the delivery of the wires and DBS. It was designed so that already trained all over the world are surgeons who are ready to do this procedure. OR time is really short. Infusion is in the recovery room. There's no general anesthetic. With respect to access and training, it really is a therapy that was developed with hospitals and surgeons and what was existing in mind. We think the data will speak for itself, particularly any disease modification data. This is for patients who no longer respond to dopamine, so they don't have good therapies. It's a big unmet need. We got RMAP and fast track breakthrough, all of that. We think if the data is as good as the phase II, it will be taken up very rapidly. Got it. Thank you. Matt? We're out of time. I just would say it's a strange thing in the biopharma space to have a procedural element. Procedures exist all the time. People get orthopedic implants, and they get angioplasty, and they get drug-eluting stents. If you have data that demonstrates therapeutic benefit and you have a procedural element to it, there's some infrastructure that needs to be built out. These things can be very successful. That's what we all hope to plan. Got it. Excellent. Yes, as we mentioned, we're out of time. Again, thank you very much for joining the panel, and thank you very much for attending our conference. Have a great rest of your day. Thanks for having me.