Gene therapy has been a goal of medicine since the first “inborn errors of metabolism” were identified by Sir Archibald Garrod in the early twentieth century. This was before anyone had a good idea of what a gene was, but the principles of Mendelian genetics were used by Garrod, with the assistance of William Bateson, to show that alkaptonuria (original paper from 1902, no paywall) is a heritable condition. A broad view of gene therapy and its connection to society at large can be found in The Science of Human Perfection: How Genes Became the Heart of American Medicine by Nathaniel Comfort. The primacy of genes and modern eugenics is relevant to current beliefs at the apex of our societal pyramid but is for another time.
The first successful gene therapy was used to cure Severe Combined Immune Deficiency (SCID) caused by adenosine deaminase deficiency (ADA) in 1990 when William French Anderson and colleagues at NIH used a retroviral vector to transfer the ADA gene to the blood cells of a four-year-old patient. [1] The patient required regular subsequent treatment with modified immune cells, but she was able to live a normal life. Since then, improved vectors have been used to cure this form of SCID.
This remarkable success led to the naïve concept that to cure an inherited disease, one had only to:
- Identify and clone the defective gene using standard techniques
- Prepare a vector to deliver the gene to the patient
- Proceed with gene therapy in patients with the disease
The completion of the human genome sequence about twenty-five years ago only reinforced these notions. Gene therapy, of course, is much easier said than done, as recent experience has reminded us with gene therapy for Duchenne Muscular Dystrophy (DMD) and STXPBP1-related disorders. [2] DMD is well known and unfortunately all too common, with an incidence of 1-in-5000 males (DMD is a sex-linked disease). DMD is caused by mutation of the gene for dystrophin and leads inexorably to skeletal muscle breakdown, with the muscle tissue invaded by fat as it gets larger but weaker. Mean survival is 24 years.
STXBP1 encodes syntaxin-binding protein required for the release of neurotransmitters in the nervous system. Mutations in STXBP1 (incidence of 1-in-30,000) lead to:
A spectrum of neurodevelopmental disorders that can include early-onset epilepsy and developmental delay, sometimes accompanied by autism spectrum disorder, increased or decreased muscle tone, or movement disorders. The symptoms a child experiences and the severity of the disorder can vary widely.
Problems with recent approaches to DMD and STXBP1 disorders have illustrated some of the barriers to effective treatment of these condition using gene therapy. These obstacles are likely to be more common than not.
Dystrophin is a very large protein that is an integral part of a complex that stabilizes muscle cells during cycles of contraction and relaxation. The size of dystrophin presents problems for the delivery of the normal gene during gene therapy, primarily because the DMD-coding region is simply too long to fit into the viral vectors currently used. Sarepta Therapeutics has attempted to solve this problem with Elevidys, which is a micro-dystrophin that according to the manufacturer “is similar to the body’s natural dystrophin. While not exactly the same as dystrophin, it’s designed to help protect muscles from damage.” The gene for micro-dystrophin will fit into the adeno-associated virus (AAV) vectors commonly used in gene therapy. Given the constraints of current gene delivery technology, this is the only imaginable approach to DMD gene therapy.
Micro-dystrophin may work, but unfortunately treatment with Elevidys led to the deaths of two patients earlier this year. Both were caused by liver failure unrelated to the underlying condition. It seems likely this was due to a severe immune reaction to the AAV variant used in the therapy. This may take us back to the sad case of Jesse Gelsinger, who volunteered to be a subject of a gene therapy vector test in 1999. Jesse had an inherited enzyme deficiency that could eventually be treated with gene therapy. [3] He died from immune-mediated multiple organ failure within a few days of the infusion of the viral vector. This was unexpected, largely because adenoviruses were thought to be innocuous, and caused a major interruption in the progress of gene therapy research, as it should have.
The future of Elevidys is currently uncertain. Sarepta has also developed a gene therapeutic approach to limb-girdle muscular dystrophy (LGMD), which is a syndrome more complicated than DMD (32 genes can be involved instead of one). A patient under a Sarepta treatment for LGMD also died of liver failure in 2025.
Delivery of a therapeutic gene to the proper target cells, tissues, and organs in the body is the difficult “engineering” task, often glossed over in discussions of gene therapy. This is especially true if the target is in the brain, which is where STXBP1 must be delivered, through the blood-brain barrier, if it is to cover a mutant. Jason Mast writes about this in Child’s sudden death unnerves a promising area of gene therapy research (Trial was a first test of viruses engineered to get past the blood-brain barrier):
Gene therapy researchers were converging on a holy grail. A few years ago, researchers at labs and companies reported they had engineered viruses that could ferry corrective genes deep into the brain, giving potential entry to a new world of treatments for Alzheimer’s, Parkinson’s, and a slew of rare genetic diseases.
This summer, after years of careful study, the first person underwent gene therapy using one of the new viruses. The patient, a young child, died two and a half days later.
The death has sent concern and uncertainty rippling through labs and companies developing gene therapies for the brain, along with rare disease groups who hoped these tools could deliver long-sought cures. They worry that Capsida Biotherapeutics unearthed a broader risk for other viruses designed to travel like a messenger pigeon to our brains, one that could derail years of progress.
Yes. The Capsida Biotherapeutics technology (Capsida is apparently a neologism based on “capsid,” which is the “protein shell of a virus”) is based on the very impressive research from the laboratory of Viviana Gradinaru at Caltech [reviewed in Adeno-Associated Virus Toolkit to Target Diverse Brain Cells (very technical but no paywall)]:
Abstract: Recombinant adeno-associated viruses (AAVs) are commonly used gene delivery vehicles for neuroscience research. They have two engineerable features: the capsid (outer protein shell) and cargo (encapsulated genome). These features can be modified to enhance cell type or tissue tropism and control transgene expression, respectively. Several engineered AAV capsids with unique tropisms have been identified, including variants with enhanced central nervous system transduction, cell type specificity, and retrograde transport in neurons. Pairing these AAVs with modern gene regulatory elements and state-of-the-art reporter, sensor, and effector cargo enables highly specific transgene expression for anatomical and functional analyses of brain cells and circuits. Here, we discuss recent advances that provide a comprehensive (capsid and cargo) AAV toolkit for genetic access to molecularly defined brain cell types.
The transgene used to complement the stxbp1 mutant in the patient is normal STXBP1. Expression of STXBP1 in the target brain cells would correct the deficiency and “cure” the condition. The basic research and “engineering” involved in preparing the transgene delivery were carefully done. All indications were that this would work. Nevertheless:
Capsida has declined to answer questions about the death beyond a brief statement. Its CEO has departed. The information that has leaked out is troubling. The child died of cerebral edema — brain swelling — a clinical course distinct from other deaths tied to gene therapy over the last decade, according to a person familiar with the matter.
Most disturbingly, none of the animal and lab studies Capsida presented indicated such a calamity was possible, making it unclear how other researchers and companies would test for such a risk.
“This is an outlier that to me is the most material event that I’ve seen in the field of genetic medicine for 10 years, where there was no suggestion that this was going to happen,” Jim Wilson, a prominent gene therapy researcher, told a rare disease conference in late October. “And this is scary, I’m sure, for all those involved.”
The field, Wilson cautioned in an interview, shouldn’t abandon these viruses. The potential is too vast. Since 2020, over $800 million has been invested into companies working on at least a dozen different brain disorders, with far more undisclosed and in academia. Much remains unknown about Capsida’s trial — it’s still yet possible the death is unrelated to gene therapy. And it’s unlikely every virus carries this risk. But executives, researchers, and regulators need to understand what happened and ideally develop an animal model for evaluating future programs.
Still, the Capsida approach is likely to work, eventually, and the research should be continued. However, without a good animal model progress will be slow. The proper usage of AAV gene therapy vectors was thought to be well understood. Perhaps the brain is just different from liver, muscle, and blood-forming and immune cells when it comes to gene therapy. This would not be a surprise.
“Genetic engineering” has been a term of art for nearly fifty years, but these recent outcomes also show that sometimes the use of the tools of modern molecular biology represents engineering only in a very loose sense. The “engineering ideal in biology” is more than a hundred years old. It led Jacques Loeb astray in the early twentieth century. It can do the same to much more “knowledgeable” biologists now. There are many unknowns between the DNA sequence and its function as a gene in the cell, tissue, organ, and organism. Apparently there are also unknowns between the function of a bespoke AAV capsid in the brain of a mouse or a monkey and in the brain of a human patient. At one level, “Who knew?” But at another level, “Why is this a surprise?” The model is only that, a model. Choosing the correct experimental model for the question asked is the hardest task for a biomedical scientist. And sometimes there is no good model. This could be one of those.
But as noted in the STAT article, there was a “false promise of safety” in this trial (as also in the preliminary experiment that killed Jesse Gelsinger):
Capsida’s death has already affected other programs. In addition to STXBP1, Capsida also registered a trial in Parkinson’s and signed partnerships with Eli Lilly and AbbVie. The Parkinson’s trial has been suspended. The status of the partnerships remains unclear.
Allyson Berent, chief science officer at the Foundation for Angelman Syndrome Therapeutics, said her group has “taken a step back” on a next-generation gene therapy for the neurodevelopmental disease.
“The really, really hard part is nothing predicted it,” she said. Adding, “If a monkey can’t predict that, then that gives a lot of pause.”
The cases of Sarepta/Elevidys and Capsida/STXBP1 have certainly reemphasized that the facile success gene therapy for SCID/ADA deficiency in 1990 was a false dawn. That target was easy to hit, just as the liver has been easy to hit in gene therapy for hemophilia A and B. Whether a micro-dystrophin can even work in the skeletal muscles of a DMD patient remains a question, as does whether all the target muscles can be hit. Gene-based therapeutics from CAR-T therapy for cancer treatment to pancreatic cancer “vaccines” to the use of siRNA (e.g., LEQVIO) have been very effective. But their development was long and incremental, with virtually all the research funded publicly by committed scientists whose livelihood did not depend on “making the experiment work.”
Thus, it may be important that Capsida Therapeutics seems to be a creature of venture capital (described very well in the current context in Gilded Rage by Jacob Silverman, 2025). As noted above, “Since 2020, over $800 million has been invested into companies working on at least a dozen different brain disorders…” Money has its prerogatives. However, one cannot help but wonder if this is really the best way to do science at the leading edge of biomedical therapeutics, from DMD to cancer to Alzheimer’s disease to Parkinson’s disease? Is the $3,000,000 rack rate for Elevidys reasonable? Or would it be more efficient and effective to have NIH and its cognate agencies around the world put their minds together in a non-proprietary attitude to get this research done and make it available to all?
No, that is not really a question, except in Silicon Valley.
Notes
[1] Retroviruses (e.g., lentivirus) insert their genome, mostly at random, into the chromosomes of infected cells. When elements that direct the expression of an exogenous gene (transgene) are included, infected cells will produce the protein encoded by that gene, usually at a low level. If the protein is an enzyme such as ADA, this will be enough because enzymes present at a few percent of wild-type level are often sufficient to produce a normal result.
[2] A note on terminology genetic terminology: STXPBP1 = normal gene, STXBP1 = normal protein, stxbp1 = mutant gene.
[3] From the linked NYT article by Sheryl Gay Stolberg: Jesse Gelsinger was not sick before died. He suffered from ornithine transcarbamylase (OTC) deficiency, a rare metabolic disorder, but it was controlled with a low-protein diet and drugs, 32 pills a day. He knew when he signed up for the experiment at the University of Pennsylvania that he would not benefit; the study was to test the safety of a treatment for babies with a fatal form of his disorder. Still, it offered hope, the promise that someday Jesse might be rid of the cumbersome medications and diet so restrictive that half a hot dog was a treat. ”What’s the worst that can happen to me?” he told a friend shortly before he left for the Penn hospital, in Philadelphia. ‘I die, and it’s for the babies.’” Jesse Gelsinger had a noble soul.
