Half a girl’s brain cells carry the working copy of MeCP2. Half carry the broken one. For three decades the working bet has been that the broken cells are the problem and the healthy ones are the rescue, and that the gap between them is what eventually takes a Rett kid’s words, her hands, her gaze. A new single-cell map of the Rett brain from Huda Zoghbi’s lab at Baylor College of Medicine, published this week in Science Advances, reorders that picture. The working cells were never really fine. They were being pulled off course by the broken ones living next to them, weeks before any symptom shows.

Which means Rett is already corrupting the brain during what the clinic still calls presymptomatic. That moves the goalposts on what “early intervention” gets to mean.

Rett is the kind of disease pediatric textbooks present as a clean story. Girls develop normally through infancy. Then between six and eighteen months something flips: they lose speech, lose intentional hand movement, lose the social spark. The genetic culprit, identified by Zoghbi’s own lab back in 1999, is MECP2 on the X chromosome. Because girls have two X’s and silence one in each cell at random, roughly half of every Rett girl’s neurons run on a working copy and half run on the mutant. The disease is mosaic at the cellular level. That fact has always made the biology more interesting and the treatment problem harder.

What the Baylor team did was deceptively simple. They took mouse models of Rett, physically sorted the MeCP2-positive cells from the MeCP2-negative ones inside the same female brain before the mice started showing any symptoms, and read out what every cell type was doing with bulk RNA sequencing and single-nucleus RNA sequencing in parallel, more than 120,000 nuclei from the hippocampus, the region that handles memory and context. The cross-comparison is the trick. Bulk sequencing averages every cell type into one signal, the way you’d average a city’s noise level into a single number. Single-nucleus sequencing tells you which neighborhood is actually shouting.

The result was a short list of twelve core genes whose activity had already drifted by the time you ran the assay, almost all of them involved in how neurons talk to each other across synapses. Twelve is small enough to be a fingerprint, and that is the point. The team is offering this as a molecular signature you can watch, in principle, to see whether an intervention is reaching the cells that matter, weeks before the behavior changes that drugs are currently graded on. “Understanding these early and cell-specific changes,” Zoghbi told reporters, “provides markers to monitor efficacy of interventions and also entry points to understand the brain circuits driving Rett features.”

The character no one in the field was watching turned out to be a kind of neuron called the trilaminar interneuron, named for the way it stretches across three layers of the hippocampus to coordinate signaling between them. It was the most disrupted cell type the team found, and it had never been flagged as vulnerable in Rett before. The reason it stayed invisible, the paper argues, is that bulk-tissue methods averaged the interneuron’s distress into the chorus of much more numerous excitatory neurons, where it disappeared. Single-cell methods picked it back out.

The other piece reorders the whole therapeutic problem. In a mosaic brain, the MeCP2-positive neurons, the ones with the working gene, were not transcriptionally normal either. The presence of their broken neighbors was changing them. That is a different kind of disease than the one I would have described a week ago. The target is not just “fix the broken cells.” It is the circuit in which broken and working cells are talking past each other.

This matters because Rett girls right now have exactly one FDA-approved drug, trofinetide, sold by Acadia as DAYBUE, approved in March 2023. It is a synthetic version of a tripeptide called glypromate (glycine-proline-glutamate), a fragment derived from IGF-1. In animals it boosts dendritic branching and synaptic signaling. In humans, Acadia’s own approval release concedes that the mechanism by which trofinetide exerts therapeutic effects in patients with Rett syndrome “is unknown”, and the drug is graded on parent-rated behavioral scales after weeks of dosing. The real-world data the International Rett Syndrome Foundation tracks is sobering: in the 12-week LAVENDER trial that won approval, 82% of patients on DAYBUE developed diarrhea versus 20% on placebo. In the 40-week LILAC-1 extension, 74.7% still had it. By the 32-month LILAC-2 extension, 53.2%. Roughly a third of families who start the drug have stopped it by the six-month mark. It is the best we have, and on the most generous reading it is a blunt instrument aimed at a mechanism nobody can name, with significant collateral. There is a real distance between that and “protect this specific interneuron from this specific transcriptional drift,” and the Baylor work is offering the first set of coordinates you would need to start closing the gap.

I came into this paper braced for another atlas. It is more than that. It reframes when Rett actually starts (earlier, biochemically, than the clinical window suggests), what to watch (a specific signature in a specific cell type, not a behavior scale), and what the working cells are doing during the so-called healthy months (not what we thought). The honest caveats are real. This is mouse work, in one brain region, in a female model of the disease, with no human biomarker on the back end yet. Translating that to a girl in a clinic is a stack of unsolved problems, and Zoghbi’s group does not pretend otherwise.

If I had a daughter heading into that six-to-eighteen-month window, I would not be telling you a treatment is coming. I would be telling you that the field finally knows what to measure, and that has to be true before anything else gets a chance to work.

Sources

  1. Science Advances – Anderson, Li, et al., “Single-nucleus profiling reveals a core disease signature and cell type–specific vulnerabilities in early Rett syndrome” (2026)
  2. Texas Children’s Hospital – “Understanding Rett syndrome before symptoms appear”
  3. News-Medical – “New study maps early genetic disruptions in Rett syndrome” (10 June 2026)
  4. Medical Xpress – “Early Rett syndrome clues emerge as 12 genes shift before symptoms appear” (June 2026)
  5. Acadia Pharmaceuticals – FDA approval of DAYBUE (trofinetide) press release (March 2023)
  6. International Rett Syndrome Foundation – DAYBUE in the Real World