A child with a rare neurodevelopmental disorder died in a gene therapy trial. Diagnosed with developmental and epileptic encephalopathy (DEE) due to syntaxin-binding protein 1 (STXBP1) mutations, the patient was the first to be treated with one of Capsida’s adeno-associated viruses (AAV) engineered for delivering genetic medicines via IV across the blood-brain barrier (BBB).
The death of any individual from a promising experimental drug is a tragedy and an unimaginably devastating situation for the family and the DEE community, watching with uncertainty as to whether there will ever be a breakthrough therapy for the neurodevelopmental disorder. STXBP1-related DEE has no specific disease-modifying treatment; the best options available are symptomatic management approaches, such as anti-seizure medications.
This is not a one-off situation. It is not the first death related to an experimental gene therapy in the past few years, let alone months. The standoff between Sarepta Therapeutics and the FDA after three patients died after receiving Elevidys, a gene therapy for Duchenne muscular dystrophy (DMD) and other dystrophies, is still in the rearview mirror. The news from Capsida also wasn’t the first time a patient died after receiving a systemically delivered BBB-crossing AAV gene therapy, as there have been two reported deaths following the approval of Zolgensma, an IV-administered AAV gene therapy for spinal muscular atrophy (SMA) that targets the nervous system.
With at least four patients treated with gene therapies dying this year, and several more dying from AAV-delivered gene therapies in recent years, now is a time for deep reflection. What went wrong? How common are patient deaths from genetic medicines? Can anything be done to steer the ship in the right direction?
Engineered AAV: The more “experimental” component
A comprehensive evaluation of the cause of death in this situation has not yet been conducted, making it currently unclear which component of Capsida’s gene therapy, CAP-002, is the culprit. The cause of death could even have nothing to do with the “experimental gene therapy” at all—STXBP1-related epileptic encephalopathy disorder can be fatal, with intense seizures ending patients’ lives. The cause could be the “gene therapy” part of CAP-002. In other words, the uptake of DNA into the patient’s cells might have posed a problem due to the presence of foreign DNA or issues related to the gene expression of the delivered DNA. The evidence against this would be Capsida’s preclinical data demonstrating that in human neurons cultured in the lab, CAP-002 enabled the restoration of STXBP1 protein levels and the correction of neuronal network activity.
Syntaxin-binding protein 1 (STXBP1) is crucial for normal brain function by regulating the release of neurotransmitters between brain cells. It does this by interacting with proteins in the SNARE complex, enabling the fusion of synaptic vesicles with the cell membrane to release chemical messengers that relay signals. Pathogenic variants in the STXBP1 gene lead to STXBP1-related disorders, often causing insufficient functional protein, which impairs neurotransmitter release, leading to neurological issues like developmental delays, intellectual disability, and epilepsy. I can’t imagine that a boost in expression for a pathogenic STXBP1 in neurons would play out too well.
The other major component of the CAP-002 is the delivery method, which arguably was the more “experimental” part—an IV-infused BBB-crossing AAV therapy. Nitin Joshi, PhD, an assistant professor at Harvard Medical School who studies how to use nanoparticles and biomaterials for targeted drug delivery, told Inside Precision Medicine, “Capsida’s AAV capsids were designed to pass through the BBB and spread widely in the brain, while reducing delivery to other organs like the liver. This allows a single IV infusion to treat the widespread neuronal involvement seen in STXBP1 encephalopathy.”
If CAP-002 made it to the BBB but did not cross over into the brain, one of several mechanisms could cause damage to the BBB, such as a dangerous immune response. Plus, STXBP1 is not exclusively expressed in the brain—it can be found in pancreatic islets and neuroendocrine cells for the regulation of GLP-1 and hormone release, respectively. So, if CAP-002 were not to cross into the brain and instead target these organs, complications could arise.
Route of administration for brain targeting
There have been many successes with local injections of genetic medicines to the brain, which bypass the BBB and precisely target areas like the striatum or motor cortex, achieving high concentrations and long-term neuronal expression with lower vector doses and reduced off-target exposure.
David Schaffer, PhD, Hubbard Howe Professor of Chemical and Biomolecular Engineering, Bioengineering, and Molecular and Cell Biology at the University of California, Berkeley, said to Inside Precision Medicine: “In my opinion, and with few exceptions, the field should strongly favor local rather than systemic administration. Systemic maximally exposes the virus to the patient’s immune system, and we have seen the potential consequences. In contrast, local administration to organs and tissues has been predominantly safe.”
Yet neurosurgical delivery is invasive, carrying risks such as bleeding, infection, or tissue damage. This is a major point for Joshi, who said, “Compared with intrathecal or intracerebroventricular injections, systemic delivery is minimally invasive, improving safety, convenience, and patient compliance, which is particularly important for young children, who were the target patient population in this trial.”
Direct injection also has a reach limited to a few millimeters from the injection site, making it unsuitable for diffuse neurodegenerative diseases, and the procedures are far more costly and resource-intensive than routine intravenous infusion. Mark Kay, MD, PhD, the Dennis Farrey Family Professor in the departments of pediatrics and genetics, as well as head of the division of human gene therapy in pediatrics at the Stanford University School of Medicine, explained to Inside Precision Medicine, “Direct injection at this point does not result in a large spread of the vector throughout the brain. So until this is possible for diseases where one needs wide scale gene delivery into a large number of regions in the brain, systemic delivery is required.”
Kay said that systemic administration, particularly with engineered AAV capsids, offers the possibility of broad CNS coverage through the circulation and greater scalability for widespread conditions, though it must overcome challenges like BBB penetration and peripheral toxicity. This trade-off—procedural risk versus systemic exposure—means the optimal approach depends on disease, safety, and logistics, driving companies like Capsida to focus as much on delivery innovation as on genetic payloads.
According to Kay, some groups have had success in making AAV variants that are more specific and, in theory, should require a lower dose. “Like any drug, there is a great deal of variation between individuals who have side effects,” said Kay. “Look at the COVID vaccine—some people have no side effects, and others can really have more severe side effects with high fevers and myalgia. From this one patient at this point, it is unclear if this particular new capsid and DNA expression cassette have a greater likelihood of severe side effects compared with others at this time.”
Since the brain is composed of many cell types, the mechanism of disease will dictate which delivery type is more likely to be successful. For some indications, like Parkinson’s disease and AADC deficiency, Kay thinks that local delivery is likely the best approach, and there are diseases for which there has been some success.
Gene therapy delivery vehicles
Capsida is not alone in pursuing systemic injection of BBB-crossing vehicles. Several major players are pushing innovation in this space: Biogen and Pfizer are advancing novel vectors and delivery technologies in preclinical programs, while smaller biotechs and academic labs are refining receptor-mediated transport systems and tailored lipid nanoparticles to boost systemic delivery efficiency. Voyager Therapeutics is developing next-generation AAV capsids designed for efficient BBB penetration and widespread central nervous system (CNS) distribution, with programs for Huntington’s disease and other neurodegenerative disorders. Denali Therapeutics, Ionis Pharmaceuticals, and Passage Bio are likewise building BBB-crossing platforms or gene therapies for Parkinson’s disease, lysosomal storage disorders, and frontotemporal dementia.
Meanwhile, non-viral methods are expanding the BBB-crossing toolkit. Denali’s Transport Vehicle platform (DNL310), Roche/Genentech’s transferrin-receptor “brain-shuttle” antibodies such as trontinemab, and Angiopep-2 conjugates like ANG1005 use receptor-mediated transcytosis to ferry drugs into the brain, with some already in human trials. Targeted nanoparticles and immunoliposomes—often Angiopep-2– or transferrin-receptor–decorated—show preclinical promise for RNA or small molecules, while intranasal routes, validated by clinical studies of insulin and oxytocin, are being refined with nanoparticle formulations. Together, these strategies complement viral vectors, offering safer, repeatable, and potentially more versatile options for CNS therapy.
Both Kay and Schaffer agreed that non-viral approaches for DNA delivery have not worked well to date. Schaffer, founder of seven companies from his lab, including 4D Molecular Therapeutics (4DMT), said, “Viral vehicles remain substantially more efficient than non-viral, and AAV remains the vehicle of choice for delivery and long-term expression of <5 kb of therapeutic DNA. Non-viral delivery, while promising and well-suited for applications like vaccines and genome editing, cannot efficiently deliver DNA at this time.” The genetic medicine company 4DMT has several clinical programs, the most advanced being 4D-150, which is currently being evaluated in the 4FRONT-1 Phase III clinical trial and the ongoing Phase I/II PRISM clinical trial in adults with wet AMD. With 4D-150, 4DMT is using a directly injected viral approach.
Even Joshi, whose lab focuses on non-viral delivery approaches, said, “Today, viral vectors remain the only clinically validated way to deliver genes broadly to the brain systemically. Non-viral approaches, such as nanoparticles or other delivery platforms, are actively being developed and hold great promise, but none have advanced to clinical trials for brain-wide gene delivery via systemic administration. Also, non-specific biodistribution, particularly to the liver, still remains a major bottleneck with non-viral platforms.”
Joshi continued, “Non-viral nanoparticles are exciting because they are scalable and less immunogenic, but they still face major challenges, including crossing the blood–brain barrier efficiently, avoiding liver accumulation, and achieving durable expression. Because non-viral transduction is typically short-lived, repeated dosing may be needed, which is another barrier to clinical adoption for their local administration via intrathecal or other routes and an area where significant advances are required. If these challenges are overcome, the field could ultimately shift toward non-viral approaches as the preferred strategy for many CNS applications.”
Frequency of gene therapy deaths
There have been issues with experimental gene therapies dating back to the 1990s—this is not new territory. Since the recent resurgence in experimental gene therapies, we continue to see serious issues. If tied to treatment, this would be the fifth death linked to gene therapy in the past year and possibly the 13th since 2019. I’ve written about the three deaths from Sarepta’s Elevidys; I’ve discussed the events on the Behind The Breakthroughs podcast. So, unfortunately, this isn’t a one-off or negligible. But is it a frequent failure? A cursory search shows that since 2019, regulatory agencies worldwide have approved roughly a dozen to over 30 gene therapies, depending on whether you count only in vivo gene transfers or also gene-modified cell treatments. Approvals have accelerated sharply, with the FDA alone clearing one therapy in 2020, two in 2021, four in 2022, seven in 2023, and six in 2024.
Despite this progress, commercial uptake remains modest because most indications are rare and rollout is slow. Only a few hundred to a few thousand patients have been treated so far. Zolgensma for spinal muscular atrophy has reached thousands globally, and Elevidys for DMD has surpassed 800 patients. So, in the past six years, there has been rapid growth of gene therapy approvals and a limited rollout that has still likely led to thousands of patients benefiting from the experimental medicines in clinical trials or commercially.
Additionally, there have been recent examples of gene editing therapeutics—an even more “experimental” genetic medicine than gene therapy—that have stopped or even reversed the course of deadly conditions. A few months earlier this year, the story of baby KJ Muldoon, who received an on-demand personalized genetic medicine unlike anything seen before, marked a watershed moment for genetic medicines. KJ went from likely never leaving the hospital, succumbing to the disease before turning one, to being discharged with a parade-like police escort.
Predicting patient responses
Time and time again, we’ve seen drugs cure mice that fail and even kill humans. NHP studies are not enough, are extremely expensive, and are the subject of intense ethical debate. A proxy for testing the safety and efficacy of the drug in a specific patient is urgent. In some cases, this can be done with relative ease, testing primary cells from a patient biopsy. In other cases, such as when dealing with brain-related disorders, this is not possible. The closest would require approaches like reprogramming patient cells into induced pluripotent stem cells (iPSCs) and then differentiating them into cortical neurons or some sort of organoid or assemblies—a process that could take weeks, even months. This may work in some situations where the disease progresses little or not at all over a significant time, but for anything that’s deteriorating fast, it would be as useful as a million-dollar chocolate teapot.
We are at the very beginning of the precision medicine era, laying the road for treatments that provide hope where there is none—bespoke genetic medicines and computer-designed drugs. But the success of personalized approaches is highly dependent on understanding the underlying biology and predicting patient prognoses and responses to drugs to tailor the right approach.
When it comes to the brain, Joshi doesn’t think that a single delivery approach will win outright, as each disease will call for different delivery approaches. “For conditions that need broad brain coverage, systemic AAV gene therapy remains the most proven option, enabling whole-brain delivery with a single IV dose,” said Joshi. “If deep-brain targeting isn’t needed, intrathecal delivery can work well, while intraparenchymal injections are best reserved for very focal diseases.”
In the future, precision medicine may be able to offer finding the right modality for treating that disease and how to tailor the therapy for each and every patient, such as a child with DEE, providing the first steps to a whole new, healthy life.
