Gene therapy for GM1 gangliosidosis moved the biochemistry. The clinical picture is mixed

GM1 gangliosidosis is caused by biallelic variants in GLB1, which produce a deficiency of the lysosomal enzyme beta-galactosidase [s1]. Without that enzyme, GM1 ganglioside accumulates. The disease is neurodegenerative, fatal, and currently has no effective therapy [s1].

A phase 1–2 trial published in the New England Journal of Medicine on 6 February reports results from nine children with type II GM1 gangliosidosis — late-infantile or juvenile onset — who received immunosuppression and a single intravenous infusion of adeno-associated virus serotype 9 encoding beta-galactosidase [s1].

The primary endpoint was safety [s1]. Reading the paper for what it establishes rather than what it promises means starting there.

What happened on safety

Over a three-year period, 124 adverse events occurred across the nine participants [s1]. Thirty were judged by the investigator as possibly, probably or definitely related to the gene therapy: eight gastrointestinal events, 21 laboratory abnormalities associated with inflammation, and one tachycardia event [s1].

Five serious adverse events occurred, including vomiting that led to hospitalisation in one participant, which was attributed to the gene therapy [s1].

Serum aspartate and alanine aminotransferase levels increased in all participants and returned to baseline by 18 months [s1]. Transaminase elevation is a recognised feature of systemic AAV administration; that it occurred in every participant, and that it took up to 18 months to resolve, is part of what an open-label safety readout is for.

What happened on biochemistry

Here the results are uniform. In all participants, cerebrospinal fluid beta-galactosidase levels increased and CSF GM1 ganglioside levels decreased [s1].

That is the mechanistic claim demonstrated: an intravenous AAV9 vector produced measurable enzyme activity in the central nervous system and reduced the substrate that accumulates there. For a neurodegenerative lysosomal storage disease, crossing into the CNS compartment at all is not guaranteed, and this trial shows it happened.

Biochemical correction is not the same as clinical benefit. It is the precondition for it.

What happened clinically

This is where the reading requires care.

Expressive communication and gross motor scores appeared stable, but fine motor and receptive communication scores decreased [s1]. Two domains held; two did not.

The Clinical Global Impression–Improvement score is rated on a scale from 1, very much improved, to 7, very much worse [s1]. The median CGI-I score was 3 — minimal improvement — at two years, and 4 — no change — at three years [s1]. The authors note that in historical controls, scores have been shown to increase, indicating worsening, over time [s1].

Neuroimaging showed patterns consistent with reduced rates of cerebral atrophy and favourable changes in myelination compared with baseline [s1].

The authors' own conclusion is stated with matching restraint: secondary endpoint results suggested improvements in biochemical markers and neuroimaging patterns and stable or reduced rates of developmental deterioration in some measures [s1].

The comparison problem

The trial was open-label, dose-escalating, and had nine participants [s1]. There was no randomised control group; the comparison is against historical controls [s1].

That design is standard for a first-in-human trial in an ultra-rare fatal disease, and it is also the design that makes clinical inference hardest. Historical controls differ from trial participants in ways that are difficult to adjust for — era of diagnosis, supportive care, ascertainment, and the selection that determines which families enrol in an experimental trial at all.

A median CGI-I of 4 at three years means the median child was rated as unchanged. Against the natural history of a progressive neurodegenerative disease, unchanged is a meaningful observation. Against no control arm, it is an observation whose meaning depends entirely on how confident one is in the historical comparison.

The divergence between domains — expressive communication and gross motor stable, fine motor and receptive communication declining — also resists a simple summary. It is compatible with partial benefit, with differential rates of natural progression across domains, or with measurement properties of the individual scales.

The trial was funded by the National Human Genome Research Institute and others and is registered as NCT03952637 [s1].

What it establishes

Three things, at this stage.

That a single intravenous AAV9 infusion encoding beta-galactosidase produced CNS enzyme activity and reduced CNS substrate in every one of nine children with type II GM1 gangliosidosis [s1].

That the approach carried a defined safety burden — universal transaminase elevation resolving by 18 months, one hospitalisation for vomiting, five serious adverse events over three years across nine participants [s1].

And that the clinical signal, on secondary endpoints and against historical comparison, is suggestive rather than demonstrated.

What to watch

Whether a controlled trial follows, and what it uses as a comparator. For diseases where randomising to no treatment is difficult, the design of the next study is the field's real decision point.

Whether the biochemical response persists. Three years of follow-up in nine children tells you the enzyme reached the CNS; durability over the timescale of childhood is a separate question.

And whether treating earlier changes the picture. In neurodegenerative disease, the interval between onset and intervention is often the variable that dominates outcome, and a trial in late-infantile and juvenile onset patients cannot answer what would happen if the same vector were given sooner.

This article describes trial results and is not clinical guidance. Treatment decisions in rare genetic disease belong with the clinicians and families involved.

Sources

• [s1] Lewis CJ, D'Souza P, Johnston JM, et al., "AAV9 Gene Therapy in Type II GM1 Gangliosidosis — A Phase 1–2 Trial," The New England Journal of Medicine, published 6 February 2026. https://doi.org/10.1056/NEJMoa2510935