A short clip is going around showing a neuron, an injection, and the words "blocking one gene restored key brain signalling linked to autism". It is not exaggerating the science by much, which is unusual. The study is real, it is from a serious lab, and it was published in Nature Communications in May 2026. It is also, importantly, a mouse study, not a human treatment. Here is what it actually shows.
The Broken Part: NMDA Receptors
Two of the best-established autism-risk genes are SHANK2 and SHANK3. SHANK3 in particular causes Phelan-McDermid syndrome, a genetic condition strongly linked to autism that we covered in 1 in 7,300: the autism-linked syndrome most people have never been tested for. These genes build scaffolding proteins inside neurons, the structural glue that holds signalling machinery in place at the synapse, the junction where one neuron talks to the next.
One of the things that scaffolding supports is the NMDA receptor, a key gatekeeper for learning and brain plasticity. Unlike most receptors, an NMDA receptor needs two different signals to open at once: glutamate, the brain's main "go" signal, and glycine, acting as a co-agonist. Without enough glycine nearby, the receptor stays shut even when glutamate is present. In mice missing working copies of SHANK2 or SHANK3, NMDA receptor activity is suppressed, and this has been linked to the social and repetitive-behaviour differences seen in autism.
The Fix: Block the Glycine Vacuum
Researchers at South Korea's Institute for Basic Science, led by Eunjoon Kim, targeted a different gene: SLC6A20, which builds a transporter that pulls glycine away from the synapse, like a vacuum clearing it from the area. Block that vacuum with an antisense oligonucleotide, a short synthetic strand that silences the gene's instructions, and glycine builds up locally. With more glycine around, NMDA receptors that were starved of their co-agonist can open again, even without touching SHANK2 or SHANK3 directly. As Kim put it: "Unlike gene re-expression strategies, SLC6A20 inhibition works by modulating endogenous signalling."
The team injected the treatment directly into the brains of adult mice, aged 3 to 7 months, not newborns, and tested them four weeks later. A single injection stayed effective for at least eight weeks.
What Improved, and What Did Not
The results are genuinely mixed, and the paper is upfront about it. In mice missing SHANK2, three-chamber social interaction tests improved, home-cage social approach improved, and courtship vocalisations were partially rescued. But hyperactivity and marble-burying, a repetitive-behaviour test, were not rescued. In mice with SHANK3 mutations, excessive self-grooming was strongly rescued, along with social interaction and vocalisations, but marble-burying again did not respond. The treatment also restored the underlying NMDA receptor signal itself, directly measured, in a third mouse line.
The researchers went a step further and tested human relevance directly: cortical organoids, small lab-grown clusters of human brain cells, engineered with the same SHANK2 or SHANK3 mutations using CRISPR. The same treatment restored NMDA receptor signals in these human cells back to typical levels.
Why It Worked in Adults
The detail most worth noting is that this worked in fully grown mice. A lot of neurodevelopmental research assumes a narrow prenatal or early-childhood window after which the wiring is fixed. Here, adult brains regained function from a treatment given well after any such window had closed. That does not mean the same holds in an adult human, brains are vastly more complex, but it is the finding the authors highlight most.
What This Is Not
This is not a treatment you or your child can access. There is no human trial. The authors themselves list real obstacles: NMDA receptors vary in composition and in how much they rely on glycine versus a related co-agonist, D-serine, across brain regions and developmental stages, which makes dosing genuinely hard to get right. Autism itself is clinically diverse, different genetic causes, different cognitive profiles, different co-occurring conditions, so a treatment built around one pathway will not help everyone. And a meaningful share of the mouse behaviours, hyperactivity and repetitive behaviour among them, did not improve at all, which tells us NMDA restoration is not the whole story even in this narrow, single-gene model.
What it does show is a plausible, testable mechanism, evidence it works in adult brains and not only during development, and a first bridge into human tissue. That is a genuinely good place for a 2026 paper to be. It is a long way from a prescription.
If This Is Personal for You
Most autism is not caused by a single SHANK2 or SHANK3 mutation; those are rare, high-impact causes among many contributing genes, as we explain in ADHD genes beyond dopamine for the ADHD side of the picture. If a diagnosis, your own or a family member's, is what brought you here, a validated screener is still the sensible next step, not a wait for future pharmacology. The AQ-10 takes about five minutes, is free, and gives you a score to bring to a professional.
Source: Roh JD, Bae M, Oh Y, et al. Glycine-modulating Slc6a20a-ASO restores NMDA receptor function in SHANK2 and SHANK3-mutant mice and cortical organoids. Nature Communications, 29 May 2026. DOI 10.1038/s41467-026-73881-9. Institute for Basic Science (South Korea). Figures and quotations are from the article and the institute's release; the explanations in plain language are ours. This article is educational and describes preclinical animal and cell research, not an available human treatment.