The gene therapy that fixes a child's liver and leaves the family history alone

The gene goes into the liver, and stops there

On Wednesday the US drug regulator cleared the first treatment for glycogen storage disease type Ia [1]. The condition, also called Von Gierke disease, affects roughly 1 in 100,000 American infants [2].

The fault is in a single gene. It leaves the body short of one enzyme, glucose-6-phosphatase [1]. That enzyme is what releases stored sugar from the liver and kidneys into the blood between meals [1]. Without it, blood sugar drops dangerously whenever a person goes too long without eating [1].

The management is unending. Frequent meals, no simple sugars, and doses of raw or specially formulated cornstarch around the clock, day and night [1]. Cornstarch digests slowly, so it trickles sugar into the blood the way a working liver would.

The new therapy is Genglycos, given once. It uses AAV8 - a stripped-down virus used as a delivery carrier - to carry a working copy of the G6PC gene into liver cells [1]. It is approved for adults and children aged eight and over [1].

Read the trial closely. It ran 48 weeks, randomised, double-blind, against a dummy treatment [1]. The main goal was not blood sugar. It was cornstarch. Treated patients cut their daily intake by an average of 31%, about one dose a day, against placebo [1].

On the measure a patient cares about most, the result is honest and awkward. Treated patients spent a slightly larger share of time in the low-blood-sugar range than placebo patients did, by a mean of 3 percentage points [1]. This is an accelerated approval, granted on a stand-in measure, and Ultragenyx must run further trials to confirm the drug works [1].

The safety list is not short. Anaphylaxis, adrenal insufficiency, high lactate, low blood sugar [1]. The label carries warnings for liver toxicity and for the risk of tumours [1]. Nearly a third of treated patients developed high blood triglycerides, against 8% on placebo [1]. And one line points at everything that follows: the label says Genglycos should not be used during pregnancy [1].

Moderna and Beam Therapeutics have candidates of their own for the same disease, both still early [2].

Two children, two viral carriers, and the doses that killed them

The same week, the other end of the same technology surfaced. Two deaths in Chinese gene-therapy studies came to light, both from last year, neither disclosed at the time [3].

A six-year-old girl with a rare but non-fatal neurodevelopmental disorder died from a severe immune reaction. It came days after an experimental base-editing treatment her parents had helped pay for [3]. A boy in a HuidaGene trial for Duchenne muscular dystrophy died of acute respiratory distress after a CRISPR-based therapy [3].

Both cases share a shape. High doses, delivered by viral carriers, into children [3]. A viral carrier is not a live infection - the virus is emptied of its own genes and loaded with the therapeutic one. But the immune system does not always read it that way, and a big enough dose can provoke a reaction that kills.

Both trials ran through China’s investigator-initiated route, which lets academics and hospitals start early human studies faster and with fewer requirements than the standard drug-review path [3]. China issued a new regulation, Order 818, in October 2025; it came into force in May 2026, after both deaths but before either was public [3]. It sets tighter requirements on trial design, preclinical safety work and patient screening [3].

Gene-editing experts asked to review HuidaGene’s earlier data split sharply [4]. “I’m not impressed,” said Dongsheng Duan of the University of Missouri, who has spent a decade on gene editing for Duchenne [4].

The lines the labels draw

The regulator’s other approval this week runs on the same boundary.

Regeneron’s Pasatru was cleared on Wednesday for fibrodysplasia ossificans progressiva, a rare genetic disease in which muscle, tendon and ligament slowly turn into bone [5]. It is only the second drug ever approved for the condition, and it was roughly three decades in the making [5][6]. Regeneron shares rose 4% [7].

The trial numbers are worth seeing raw rather than as percentages. Across 63 adults over 56 weeks, the placebo group formed 19 new bone lesions among 21 patients [5]. The 10 mg/kg group formed two among 23; the 3 mg/kg group formed one among 19 [5].

The two doses did not rank the same way on everything. On clinician-assessed flare-ups over the same 56 weeks, the higher dose recorded 9, the lower dose 53, and placebo 66 [5]. Better on new bone at the low dose, far better on flare-ups at the high one.

Pasatru carries a warning for harm to a fetus. Patients who can become pregnant are told to use effective contraception during treatment and for six months after the last dose [5]. Genglycos is not to be used in pregnancy [1]. Two different medicines, two different mechanisms, and both labels stop at the same border. Pasatru was the 33rd novel drug the agency’s drug centre cleared this year [8].

The mark that is not written in the DNA

The freshest result of the week complicates the border rather than confirming it.

Researchers at the Marine Biological Laboratory reported evidence that a mother’s age shapes her offspring through changes in how genes are used, not through damage to the DNA itself [9]. Offspring of older mothers commonly fare worse, across species from invertebrates to elephants and primates [9].

“Nearly all forms of life show some level of maternal age effect,” said Kristin Gribble, who led the work [9]. Her lab studies rotifers, tiny aquatic animals that breed fast enough to track across generations [9].

The telling detail is reversibility. The effects did not compound generation on generation. They could be undone within a single generation [9]. That argues against slow damage or mutation piling up, and points instead at histone modifications - chemical tags that decide whether a gene is switched on or off [9]. The team is now testing that directly, and also whether mitochondrial DNA plays a part, since it comes only from the mother [9].

Hold the caveat firmly. This is rotifers, and the mechanism is a hypothesis under test, not a settled finding in people [9]. One strain even ran the other way, with offspring of older mothers living longer, which suggests genetic differences change how strongly any of it lands [9].

“It’s not just about what’s in your genome as an individual,” Gribble said [9].

Twenty years of a reprogramming trick

Nature marked an anniversary this week that sits underneath all of the above [10].

Twenty years ago Shinya Yamanaka and Kazutoshi Takahashi showed that switching on four genes could erase an adult mouse cell’s identity and return it to an embryo-like state [10]. Those induced pluripotent stem cells won a share of the 2012 Nobel prize [10].

The path from there to patients has been slow, and one reason is instructive. A cell that has not committed to a new identity can seed a tumour [10]. So every batch has to be checked for stragglers still in the embryo-like state before it can go into anyone [10].

Early on, Deepak Srivastava’s lab at the Gladstone Institutes could turn about 1% of these cells into heart cells [10]. Years of work took that past 90%, enough for the dish to beat in unison [10]. Japan granted conditional approval this year to two such therapies, one for Parkinson’s disease and one for heart failure [10]. Many more trials report over the next five years [10].

Nature’s editorial makes the argument for restraint plainly: one trial that harms its participants could freeze the whole field [10].

Around the edges

BioMarin bought Alesta and its experimental drug for a rare bone disorder for $275 million, its third acquisition in two years [11]. Argenx said an under-the-skin version of its blockbuster Vyvgart met its main goal in a late-stage myositis trial, and its shares rose more than 13% [12].

In Washington, Trump picked Heidi Overton, a doctor and deputy director of the White House domestic policy council, to run the FDA [13][14]. The agency has lost more than 3,000 staff over the past year [14]. One analyst noted her lack of drug-regulatory experience and called her likely effect on drug review uncertain [13].

Two smaller results worth noting. University of Chicago researchers engineered a gut bacterium, Bifidobacterium longum, to carry an immune-stimulating treatment into pancreatic tumours [15]. It slowed tumour growth, and worked better alongside chemotherapy or radiotherapy [15]. In animals only, so far [15]. And nutrition researchers pushed back on a review paper suggesting that restricting dietary protein promotes healthspan and lifespan, warning it would be a poor message for older adults in particular [16].

Separately, a Harvard-led team kept lab-grown brain organoids alive for years rather than months, long enough to watch them mature the way a developing brain does [17][18].