Xinhua Hospital panel linked the death of a six-year-old participant to a gene editor injected into the cerebrospinal fluid
Xinhua Hospital panel linked the death of a six-year-old participant to a gene editor injected into the cerebrospinal fluid
On 23 July, Science and Retraction Watch reported that a girl had died after an early trial of a therapy targeting a CHD3 gene variant. An emergency panel at Xinhua Hospital concluded that the death was “definitely related” to the treatment and identified thrombotic microangiopathy as the cause. This condition damages small blood vessels through the formation of microscopic blood clots.
In March 2025, doctors injected two AAV9 viral vectors into the cerebrospinal fluid of a six-year-old participant whom the investigation calls Mei. AAV is a viral shell that can deliver genetic instructions into a cell. The girl died seven days later. The panel linked her death to the treatment, and the investigation describes a severe immune reaction.
The treatment was intended to correct the R1025W substitution in CHD3, a gene involved in controlling the activity of other genes during brain development. Zilong Qiu’s team chose a base editor for this purpose. This type of CRISPR changes a single DNA “letter” without cutting both DNA strands. The complete editor could not fit inside one AAV shell, so the team divided it between two. Both parts had to enter the same nerve cell, assemble into a functional protein, and modify the intended DNA sequence.
The Nature article, published on 18 February 2026, described this approach in mice carrying the human CHD3 variant. After editing, the animals had higher CHD3 protein levels and changes in behavioral measures. In an experiment involving two macaques, the researchers detected parts of the editor in neurons and measured their assembly. Reaching enough brain cells requires hundreds of trillions of viral particles. The investigation links delivery at this scale to immune risk, while the panel identified ТМА as the cause of death.
“A death should always be reported in a trial in which a treatment is administered to a human for the first time,” said bioethicist Hank Greely.
The trial record still lists the study as recruiting. The trial is designed to enroll one participant, and its primary safety outcome is treatment-related serious adverse events over 26 weeks. After the severe outcome following the first injection, that information must become part of the decision about the next administration for the family and the clinical team.
The personalized base editor for an infant with CPS1 deficiency was delivered to the liver using lipid nanoparticles. One year later, the child had improved clinically, and no serious adverse effects had been reported. In the CHD3 trial, the large genetic construct required two viral vectors, while delivery to the brain required hundreds of trillions of viral particles. The risk of the next injection depends on the target tissue, delivery vehicle, dose, immune protection, and disclosure of the severe outcome. Preclinical results can support a clinical decision only when considered together with these data.
Sources:
- Science and Retraction Watch investigation
- full text of the Nature article