Anti-NMDA receptor encephalitis produces a bewildering range of symptoms — seizures, psychosis, memory loss, movement disorders — and the autoantibodies driving the disease come from different patients with different immune histories. The expectation was that the antibodies would attack the receptor at many different sites, producing disease through many different mechanisms.
Kim et al. (Science Advances 2026) used cryo-electron microscopy to map where patient-derived autoantibodies actually bind on the GluN1 subunit of the NMDA receptor. Nearly all of them converge on the same domain — the amino-terminal region, the part of the receptor most accessible from outside the cell. Antibodies from engineered mouse models matched the same binding pattern found in human patients. The apparent diversity of the disease — many patients, many antibodies, many symptoms — funnels through a single structural vulnerability.
This convergence has immediate practical consequences: a diagnostic blood test becomes feasible when you know the target is one domain rather than many, and therapeutic agents can be designed against a specific structural feature rather than broad immunosuppression. The current treatment — suppress the entire immune system — works the way a fire hose works on a kitchen fire. Knowing the door changes the intervention from flooding to locking.
The general principle: when many independent agents produce a shared outcome, the diversity of the agents may conceal the singularity of the target. The attack surface looked wide because the attackers varied, not because the vulnerability did. Mapping the target instead of cataloguing the attackers reveals that apparent heterogeneity in mechanism is compatible with a single point of failure.