How Water Decides What Your Brain Learns and What It Forgets

Somewhere in your brain right now, a tiny molecular gate is deciding whether an experience becomes a memory. The gate is a protein called an NMDA receptor, and the decision comes down to two ions, calcium and magnesium, that are chemically so similar they should, by most logic, be interchangeable. They carry the same electrical charge. They sit two squares apart on the periodic table. And yet the receptor treats them entirely differently, waving calcium through while using magnesium as a physical plug. For forty years, neuroscientists understood this had to be true. They couldn’t actually see why.

Now they can. A team led by Hiro Furukawa at Cold Spring Harbor Laboratory has captured, for the first time, the precise molecular mechanism by which calcium slips through the NMDA receptor channel while magnesium gets stuck, producing images clear enough to show individual water molecules clustered around each ion like a retinue. The work, published in Nature Neuroscience, resolves a question that has sat at the heart of learning and memory research since the 1980s.

The key, it turns out, is dehydration. Both ions travel through the brain wrapped in a shell of water molecules, but the tightness of that grip differs in ways that matter enormously. “Magnesium attracts water more strongly than calcium,” says Furukawa. “It’s more difficult to take out water molecules surrounding magnesium than calcium.” This has consequences when either ion approaches the business end of the receptor, a narrow bottleneck in the channel known as the asparagine cage, or Asn cage, after the amino acid residues lining its walls. To pass through, an ion must shed some of its water. Calcium can do this; the cage coordinates directly with a partially stripped calcium ion at five distinct positions as it transits the channel. Magnesium, gripping its water shell too tightly to let go, can’t. It sits outside the cage, fully hydrated, blocking the entrance. “It’s a sieve,” Furukawa says.

Elegant in retrospect, certainly. But the theory was proposed decades before anyone could actually confirm it, and confirming it required technology that simply didn’t exist then.

Fifty Thousand Movies of a Single Gate

The breakthrough came through single-particle cryo-electron microscopy, a technique that has transformed structural biology over the past decade by allowing researchers to image proteins at near-atomic resolution without the need for crystals. Furukawa’s postdoc Ruben Steigerwald prepared samples of NMDA receptors bound to calcium or magnesium, froze them in a glass-like layer of ice, and shot electrons through them from millions of different angles, generating roughly 50,000 movies of the receptor in action. The resolution achieved in the transmembrane region ranged from 2.3 to 3.0 angstroms, fine enough to place individual water molecules with confidence. “It’s all about resolution,” Furukawa says. Getting there required not just cutting-edge hardware but the lab’s high-performance computing cores churning through the raw data for weeks.

What emerged from that analysis was the first direct picture of how the Asn cage sorts ions by how easily they give up their water. Calcium appears at five positions inside the cage, each one showing a different degree of dehydration as the ion negotiates the narrowest constriction, formed by two asparagine residues contributed by the receptor’s GluN2B subunit. Magnesium shows up at only two positions, both outside the cage, embedded in a network of water molecules that connect it to the cage walls without ever letting it through.

There is a further twist. The team also found that two lipid molecules, tucked into pockets near the back of the Asn cage, appear to influence how voltage-sensitive the magnesium block is. Molecular dynamics simulations showed that when these lipids are present, magnesium moves closer to the cage entrance at resting membrane potential (around negative 70 millivolts) than it does at more depolarized voltages; remove the lipids, and this voltage dependence largely vanishes. Mutations in the amino acid residues lining these lipid pockets altered magnesium’s blocking potency in electrophysiology experiments on frog oocytes, which provided complementary evidence that the lipid environment around the channel is not merely structural scaffolding but an active participant in tuning receptor behavior.

A Molecular Filter with Clinical Stakes

Why does any of this matter beyond the satisfying resolution of a longstanding puzzle? The NMDA receptor is, in a real sense, the mechanism of learning. It functions as a molecular coincidence detector: it only opens, and only lets calcium in, when two things happen at once, a neurotransmitter arrives and the postsynaptic neuron is already electrically active. That simultaneous activation is what Hebbian plasticity requires, the “neurons that fire together, wire together” principle that underlies how synapses strengthen and new memories form. Magnesium’s role as the channel plug is what enforces this logic; at rest, it sits in the pore and blocks calcium entry even when glutamate binds. Depolarization pushes it out, the gate opens, and calcium floods in, triggering the downstream signaling cascades that strengthen the synapse.

The Asn cage is also a site of clinical concern. Spontaneous mutations in the asparagine residues that form it are associated with a class of severe developmental disorders called GRIN conditions, which can leave children non-verbal, unable to walk, and experiencing frequent seizures. Understanding the normal architecture of the cage, at the resolution this study achieved, is a prerequisite for understanding how individual mutations distort it.

Some questions remain open. How does extreme hyperpolarization eventually force magnesium back out through the channel, a phenomenon observed electrically but not yet explained structurally? What happens when high concentrations of calcium compete with magnesium for the same sites? The team suspects these answers will require datasets even larger than the ones that produced this study, and structures captured at a range of voltages rather than the fixed conditions possible with cryo-EM.

What the work does establish, with a clarity that has eluded the field for four decades, is that memory begins with water. Not in any mystical sense, but in a straightforwardly chemical one: the difference between learning something and forgetting it, at the molecular scale, comes down to which ions can shed their hydration shells tightly enough to squeeze through a protein bottleneck a few angstroms wide. It is, all things considered, a rather small gap on which to hang quite so much of what it means to be human.

https://doi.org/10.1038/s41593-026-02283-3

Frequently Asked Questions

Why does magnesium block the NMDA receptor channel instead of just passing through like calcium?

It comes down to how tightly each ion holds onto its surrounding water molecules. Calcium can shed part of its water shell relatively easily, allowing it to squeeze through the receptor’s narrow molecular filter. Magnesium grips its water molecules so strongly that stripping them away would cost far more energy than is available, so it sits outside the filter, fully hydrated, plugging the entrance instead. The new cryo-EM structures from Cold Spring Harbor Laboratory show this difference directly for the first time.

Is the magnesium block in NMDA receptors actually necessary for learning?

Yes, and the logic is elegant: magnesium’s plug is what makes NMDA receptors function as coincidence detectors. At rest, the plug sits in the channel and prevents calcium from entering even when the neurotransmitter glutamate binds. Only when the postsynaptic neuron is already electrically active does the plug get pushed out, allowing calcium to flow and trigger the synaptic strengthening that underlies memory formation. Without magnesium’s gating role, the receptor would lose its ability to detect whether two neurons are firing simultaneously, which is the core requirement for Hebbian plasticity.

What are GRIN disorders, and how does this research connect to them?

GRIN disorders are caused by mutations in the genes encoding NMDA receptor subunits, and they can cause severe developmental disabilities including epilepsy, intellectual disability, autism spectrum features, and loss of the ability to walk or speak. Many of these mutations fall directly in the asparagine cage region whose structure this study has now resolved at near-atomic resolution. Knowing exactly how the healthy cage coordinates calcium and blocks magnesium is a necessary foundation for understanding what goes wrong when individual residues are mutated.

How do the lipids found near the channel affect memory and learning?

Two lipid molecules sitting in pockets near the back of the channel’s selectivity filter appear to help tune how sensitively the magnesium block responds to changes in voltage. Molecular simulations in the study showed that removing these lipids blunts the voltage dependence of magnesium binding; mutations in the residues lining the lipid pockets altered magnesium’s blocking strength in live-cell electrical recordings. Whether lipid composition in neuronal membranes might influence learning capacity in intact animals is an open question this finding now makes worth asking.