Rett syndrome does something almost theatrically cruel to development. A girl develops normally for her first six to eighteen months. She babbles, reaches for toys, starts crawling. Then the disease pulls every milestone backward. Hands that once grasped objects begin wringing compulsively. Words vanish. Walking becomes unsteady or impossible. Regression can take weeks or stretch across years, but the direction never reverses on its own.
A natural history study published this year in the Journal of Neurodevelopmental Disorders tracked 1,228 females with classic Rett syndrome and documented that trajectory in pitiless detail: skill acquisition peaks before age two and essentially stops by age six, with the one-year incidence of gaining or regaining any of 51 developmental skills dropping to near zero after that threshold. Once the regression window closes, the disease locks in whatever function remains, and families spend decades managing a condition that took everything in its first few years.
On June 29, Neurogene reported 30-month follow-up data from its Phase 1/2 trial of NGN-401, a gene therapy that delivers a full-length copy of the MECP2 gene directly into the cerebrospinal fluid. Ten participants, spanning pediatric through adolescent and adult age groups, gained 47 developmental milestones total. Every single one of them improved. Not one lost a milestone she had gained.
Both pediatric and adolescent/adult participants responded. What makes the adult response extraordinary is how directly it collides with the 1,228-patient natural history: after age six, the disease simply does not produce new milestones on its own.
The Acceleration Nobody Expected
Raw response rates in small gene therapy trials are common enough to dismiss. Ten for ten sounds good until you remember that trial participants are selected, endpoints can be generous, and Rett syndrome's pseudo-stationary phase can produce minor fluctuations that look like improvement if you squint. But the trajectory of the NGN-401 data is harder to explain away.
Milestone gains increased by 95% between the six-month and twelve-month marks. Over the full follow-up period extending to 30 months, they increased by 147% relative to the six-month count. In a linear model of improvement, where the treatment effect plateaus after initial gains, you would expect roughly a 100% increase when you double the observation window from six to twelve months and then a gradual flattening. Instead, the curve bent upward, with each additional month producing more milestones than the month before, not fewer.
Neurogene's CEO Rachel McMinn framed the pattern as a "restart of developmental progression," and the data bear that interpretation with unusual directness. Participants gained milestones in a stepwise, developmentally ordered sequence: hand function building on trunk stability, communication building on hand function, each skill scaffolding the next one in the same order that typically developing children follow. Seven of the ten participants gained milestones across at least two of the three core Rett domains, meaning hand function, gross motor, and communication, rather than clustering their gains in a single area. Median time to first clinical improvement was two months.
The MECP2 Dosage Problem
Rett syndrome should be a straightforward gene therapy target. A single gene, MECP2, sits on the X chromosome. Mutations in that gene cause the disease. Deliver a working copy and the problem should resolve itself.
Except that MECP2 is one of the most dosage-sensitive genes in the human genome, and the consequences of getting the dose wrong are symmetric. Too little MECP2 protein produces Rett syndrome. Too much produces MECP2 duplication syndrome, a condition with its own devastating constellation of intellectual disability, seizures, and early death. No comfortable range exists on either side of the therapeutic window. It is a knife edge where both directions lead to neurological catastrophe.
This dosage sensitivity killed earlier gene therapy attempts in animal models. Conventional AAV9 vectors carrying a full-length MECP2 gene reliably produced toxic overexpression in wild-type animals. If a cell that already has one functional copy of MECP2 receives a second copy from the gene therapy vector, it can tip into the overexpression range. In Rett syndrome, where approximately half of a patient's cells have the mutant MECP2 silenced by X-inactivation and the other half express normal MECP2, the gene therapy must selectively rescue the deficient cells without poisoning the healthy ones. Nobody figured out how to do this reliably for more than a decade.
Neurogene's solution, called EXACT (Expression Attenuation via Construct Tuning), embeds a microRNA-based self-regulating circuit directly into the gene therapy construct. Every time a cell transcribes the therapeutic MECP2 gene, it simultaneously transcribes a microRNA that inhibits further expression of that same transgene. EXACT creates a ceiling on MECP2 protein production that the cell cannot breach, regardless of how many copies of the vector it receives. In preclinical studies published in Science Translational Medicine, the EXACT circuit prevented toxic overexpression in wild-type female mice while extending survival in Mecp2-null males from roughly 10 weeks to 37 weeks, a near-quadrupling of lifespan.
The Competitor and the Design Divergence
Taysha Gene Therapies is pursuing the same target with a different architecture. TSHA-102 uses a miniaturized version of the MECP2 gene paired with a regulatory element called miRARE, and delivers the payload intrathecally, into the spinal fluid, rather than intracerebroventricularly, directly into the brain's ventricular system. Taysha's truncated gene construct freed space for the regulatory element, and in preclinical mouse studies, TSHA-102 extended knockout survival by 56% while avoiding the toxicity that unregulated constructs produced. Phase 1/2 trials (REVEAL) are ongoing in both adult and pediatric cohorts.
Those design differences matter beyond academic taxonomy. Neurogene's decision to deliver the full-length gene rather than a mini-gene preserves all of MECP2's known regulatory domains, which control gene expression across thousands of downstream targets. ICV delivery aims for broader brain distribution than intrathecal injection, which must diffuse upward from the spinal canal. But the full-length gene is harder to package into the AAV9 capsid, leaving less room for regulatory elements, and ICV delivery requires a more invasive surgical procedure.
Neither program has released head-to-head data, and the trials differ in patient selection, endpoints, and follow-up duration, so direct comparison is premature. What the Neurogene data establish is that at least one MECP2 gene therapy architecture can produce durable, multidomain milestone gains across a wide age range with an acceptable safety profile through 30 months, in a disease that a 1,228-patient natural history study says should not allow those gains to happen.
The Safety Dataset and What It Can (and Cannot) Tell You
Across 35 patients dosed at the therapeutic level (1E15 viral genomes), Neurogene reported zero treatment-related serious adverse events and zero dose-limiting toxicities as of June 16, 2026. All treatment-related adverse events were Grade 1 (mild) or Grade 2 (moderate), and the majority were known AAV-related effects that resolved or were resolving. No new safety signals have emerged since October 2025.
For a gene therapy targeting one of the most dosage-sensitive genes in neurology, this is a dataset worth reading carefully rather than casually. Thirty-five patients with zero SAEs gives an upper bound on the true SAE rate of approximately 8.2% at the 95% confidence level, calculated as one minus the 35th root of 0.05. If the real rate of serious events were higher than one in twelve, the odds of seeing none in 35 patients would fall below 5%. That is not proof of safety. It is a lower bound on confidence that the EXACT regulatory circuit is doing what it was designed to do, specifically preventing the MECP2 overexpression that destroyed earlier unregulated constructs in animals.
Embolden, the registrational trial, has completed dosing, with topline data expected in the second half of 2027. That dataset will carry the statistical power to establish whether the Phase 1/2 results replicate at scale.
The After-Six Wall
The deepest implication of the Neurogene data is not the response rate or the safety profile or even the milestone count. It is the comparison to natural history.
In the 1,228-patient natural history study, the researchers tracked 51 developmental skills from birth through age 20. They found that skill acquisition was greatest for lower-level skills, peaked around age 6, and then functionally ceased. Regain of lost skills, where a patient recovers something she had lost during regression, was infrequent for most domains, affected fewer than 30% of patients, and occurred mainly before age 6. After age 6, the disease enters what clinicians call the pseudo-stationary phase: a long plateau where function neither improves nor dramatically worsens, except for continued gross motor deterioration in the later stages.
The NGN-401 trial included both pediatric and adolescent/adult participants, and both groups gained milestones. Bernhard Suter, Medical Director of the Blue Bird Circle Rett Center at Texas Children's Hospital and principal investigator in the trial, described the pattern as participants "gaining developmental milestones in a manner that suggests restarting developmental progression, something not observed in the natural history of the disease." Clinical significance rests entirely on the 1,228-patient dataset: when you have natural history data showing that milestone acquisition after age six has an annual incidence approaching zero, and you then observe treated participants in that age range gaining milestones on an accelerating trajectory through 30 months, the treatment is producing an outcome the disease does not generate on its own.
Limitations
The Phase 1/2 dataset is ten patients, open-label, with no placebo control. Every efficacy signal in an open-label Rett trial must contend with the known variability of the pseudo-stationary phase, the powerful placebo-adjacent effects of intensive clinical attention and hope, and the possibility that selected trial participants represent the most responsive end of the disease spectrum. An accelerating milestone trajectory is encouraging precisely because it is hard to attribute to attention effects, which typically produce their largest gains early and then flatten, but the definitive test is the blinded Embolden registrational trial.
As a retrospective, observational dataset, the 1,228-patient natural history comparison is not a controlled parallel arm. Treated participants were not matched to untreated controls from the natural history cohort on mutation type, baseline severity, or age at treatment. The comparison provides biological plausibility for a treatment effect, not a controlled estimate of effect size.
Gene therapy durability beyond 30 months remains unknown. AAV-mediated gene expression is generally considered durable in non-dividing neurons, but the developing brain introduces cellular turnover that could dilute the therapeutic transgene over years. No MECP2 gene therapy has follow-up beyond three years in humans.
MECP2 is expressed in virtually every cell in the body, not just neurons. Long-term systemic effects of the gene therapy, even with the EXACT regulatory circuit, cannot be fully characterized from a 30-month safety dataset in 35 patients.
What You Can Do
If you are a parent or caregiver of someone with Rett syndrome: the Embolden registrational trial has completed dosing, but Neurogene maintains a patient registry at neurogene.com. Topline data in the second half of 2027 will determine whether the Phase 1/2 results hold in a larger, controlled setting, and a positive outcome could support an FDA application under the Breakthrough Therapy, RMAT, and Fast Track designations that NGN-401 already holds.
If you work in gene therapy development: the EXACT self-regulatory circuit is a proof of concept for dosage-sensitive targets beyond MECP2. Any gene where overexpression produces pathology, a list that includes PMP22 (Charcot-Marie-Tooth 1A), APP (amyloid precursor protein), and SNCA (alpha-synuclein, Parkinson's), could potentially benefit from a similar auto-attenuating circuit architecture. Neurogene's 35-patient, zero-SAE safety dataset is the strongest clinical evidence to date that self-regulating gene therapy constructs can solve the overexpression problem in humans.
If you are an investor: Neurogene trades on NASDAQ under NGNE. The company's valuation will pivot on the Embolden topline data in 2H 2027, which will determine whether the 100% response rate and accelerating milestone trajectory replicate in a controlled setting with adequate statistical power.
The Bottom Line
Rett syndrome's cruelty has a signature: it gives you a typically developing child and then takes her back, one milestone at a time, and the natural history of 1,228 patients says that once those milestones are gone, they stay gone. Neurogene's NGN-401 trial, in ten participants followed for up to 30 months, produced an outcome that 1,228-patient natural history says should not exist: multidomain milestone acquisition on an accelerating trajectory, in patients both young and old enough to have hit the after-six wall where the disease stops yielding anything at all. Ten patients is not proof. But ten for ten, with no milestone losses, no serious adverse events in 35 dosed, and milestones arriving faster in the second year than the first, is the kind of early signal that makes the registrational trial the most consequential dataset in rare disease gene therapy in 2027.