D-Serine and the NMDA Receptor: The Brain's Co-Agonist

The NMDA receptor is the brain's coincidence detector — a protein channel that opens only when two conditions are met at once, and whose opening is the physical event that turns a fleeting experience into a lasting memory. For decades the second condition was assumed to be glycine. We now know that at most of the synapses that do learning, the real second key is D-Serine, a molecule the brain makes on purpose and releases mainly from astrocytes. This page explains, in plain language but without cutting corners, how that co-agonist mechanism works, how D-Serine is made and destroyed, and why it sits at the center of synaptic plasticity. It is a mechanism explainer, not health advice: D-Serine itself remains an investigational research compound.

Table of Contents

  1. The NMDA Receptor Needs Two Keys
  2. The "Glycine" Co-Agonist Site — Occupied by D-Serine
  3. D-Serine Is a Gliotransmitter: The Astrocyte Story
  4. Making and Breaking D-Serine: Serine Racemase and DAAO
  5. Long-Term Potentiation: The Molecular Basis of Memory
  6. Synaptic vs Extrasynaptic Receptors: A Division of Labor
  7. The Magnesium Block and the Calcium Signal
  8. The Double Edge: Plasticity vs Excitotoxicity
  9. What This Does and Does Not Mean for You
  10. Key Research Papers
  11. External Resources
  12. Connections
  13. Featured Videos

The NMDA Receptor Needs Two Keys

Glutamate is the brain's main excitatory neurotransmitter, and it acts on several kinds of receptors. Most of them — the AMPA and kainate receptors — behave simply: glutamate binds, the channel opens, the neuron is nudged toward firing. The NMDA (N-methyl-D-aspartate) receptor is deliberately fussier. It is built to open only when several things are true at the same moment, which is exactly what makes it useful as a detector of meaningful activity rather than background noise.

The NMDA receptor requires, simultaneously: (1) glutamate bound at its main agonist site on the GluN2 subunit; (2) a co-agonist — glycine or D-Serine — bound at a separate site on the GluN1 subunit; and (3) the surrounding neuron to be already partly depolarized, which physically pops a magnesium ion out of the channel mouth. Think of it as a bank vault that needs two different keys turned together and the room lights already on. Only then does the channel open and let calcium in. This triple requirement is why the NMDA receptor is described as a "coincidence detector": it fires when a presynaptic signal (glutamate) and a postsynaptic state (depolarization) coincide, and only if the co-agonist supply is adequate.

That third element — the co-agonist requirement — is where D-Serine lives, and it is not a minor footnote. If the co-agonist site is empty, no amount of glutamate will open the channel. The brain therefore has a tunable dial on its own learning machinery: by controlling how much D-Serine is available at a given synapse, it controls how easily that synapse can undergo plastic change.

Back to Table of Contents

The "Glycine" Co-Agonist Site — Occupied by D-Serine

The co-agonist site on the GluN1 subunit was discovered in the 1980s and named the glycine site (or glycine-B site, to distinguish it from the inhibitory glycine receptor elsewhere in the nervous system). The name stuck, and it is somewhat misleading, because glycine is not always the molecule that actually occupies it.

In a landmark 2000 paper, Mothet and colleagues showed that D-Serine is an endogenous ligand for this site — that the brain's own D-Serine is a genuine physiological co-agonist, not a laboratory curiosity (Mothet et al., PNAS 2000; PMID 10781100). Enzymatically degrading D-Serine in brain tissue sharply reduced NMDA-receptor responses, proving that D-Serine was doing real work there in the living system.

The picture that emerged over the following decade is a division of labor by location. At the synaptic NMDA receptors that mediate learning — especially in the hippocampus and cortex — D-Serine is the dominant co-agonist. At many extrasynaptic receptors, glycine takes over. This is not a tidy either/or in every brain region, but the general rule is robust and was demonstrated directly by Papouin and colleagues (Cell 2012; PMID 22863013). The practical upshot: for the receptors that matter most to memory, D-Serine — not glycine — is the co-agonist that counts.

Back to Table of Contents

D-Serine Is a Gliotransmitter: The Astrocyte Story

One of the most surprising features of D-Serine is where it comes from. Classical neurotransmitters are released by neurons. D-Serine is released substantially by astrocytes — the star-shaped glial cells that were long dismissed as mere structural "glue" holding neurons in place. Molecules released by glia to influence neurons are called gliotransmitters, and D-Serine is one of the clearest examples.

Astrocytes send fine processes that wrap around synapses, forming what neuroscientists call the "tripartite synapse": the presynaptic terminal, the postsynaptic spine, and the astrocytic process, all in intimate contact. Panatier and colleagues showed that glia-derived D-Serine controls NMDA-receptor activity and is required for normal synaptic memory in the hippocampus (Cell 2006; PMID 16713567). Henneberger and colleagues then demonstrated that long-term potentiation itself depends on the release of D-Serine from astrocytes — blocking that release blocked the plasticity, and adding D-Serine back restored it (Nature 2010; PMID 20075918).

Why does the brain outsource this job to astrocytes? Partly because a single astrocyte contacts thousands of synapses, so it can act as a local supervisor: it "listens" to how active a neighborhood of synapses is and releases D-Serine to license plasticity precisely where and when activity warrants it. The result is that whether a synapse can learn is not set globally by some brain-wide average, but locally, in real time, by the glia sitting right on top of it. (Some neuronal release and neuron-based synthesis also contribute; the field's exact accounting of neuronal versus astrocytic sources continues to be refined, but the astrocyte's role in supplying the co-agonist pool for plasticity is well established.)

Back to Table of Contents

Making and Breaking D-Serine: Serine Racemase and DAAO

Because D-Serine is a signaling molecule, the brain needs precise control over how much of it is around. Two enzymes set that balance.

Serine racemase is the enzyme that makes D-Serine. It takes ordinary L-Serine — the everyday dietary form of the amino acid — and flips its three-dimensional shape into the D-form. Wolosker and colleagues identified and characterized this enzyme, originally describing it as a glial enzyme synthesizing D-Serine to regulate NMDA transmission (PNAS 1999; PMID 10557334). Serine racemase requires vitamin B6 (as pyridoxal 5′-phosphate) as a cofactor, which is one of several links between everyday nutrition and this exotic-seeming brain chemistry. The supply of its substrate, L-Serine, comes largely from the astrocytic "phosphorylated pathway" of serine synthesis.

D-amino acid oxidase (DAAO) is the enzyme that destroys D-Serine, oxidizing it and clearing it from tissue. DAAO is especially active in the cerebellum, brainstem, and — importantly for the safety story — the kidney. The balance between serine racemase (synthesis) and DAAO (degradation) determines local D-Serine concentration. This is not just theoretical bookkeeping: overactive DAAO lowers D-Serine and has been genetically linked to schizophrenia risk, while inhibiting DAAO is an actively pursued drug strategy to raise endogenous D-Serine without administering it directly (see the Schizophrenia Research page). Wolosker and Mothet's review lays out this synthesis-degradation logic in detail (FEBS J 2008; PMID 18564180).

Back to Table of Contents

Long-Term Potentiation: The Molecular Basis of Memory

Long-term potentiation (LTP) is the best-understood cellular model of learning. When two neurons fire together repeatedly, the synapse between them strengthens — the classic "cells that fire together, wire together." At the molecular level, LTP begins when NMDA receptors open and admit a pulse of calcium into the postsynaptic spine. That calcium acts as a trigger: it activates enzymes (notably CaMKII) that insert more AMPA receptors into the synapse and reorganize the scaffolding proteins, physically making the synapse bigger and more responsive.

Because the NMDA receptor is the gatekeeper for that calcium pulse, and because D-Serine is required to open the NMDA receptor, D-Serine availability directly sets the ease of LTP. Where D-Serine is plentiful, synapses potentiate readily; where it is depleted, LTP is weak or absent even if glutamate signaling is intact. This is the through-line that connects a single mirror-image amino acid to something as grand as memory: no co-agonist, no calcium; no calcium, no potentiation; no potentiation, no durable strengthening of the synapse.

This mechanistic chain also explains why researchers became interested in D-Serine as a potential intervention for conditions of impaired plasticity — from schizophrenia to age-related memory loss. But it is worth stressing that "this molecule is necessary for the mechanism" is a very different claim from "adding more of this molecule improves the outcome in people." The first is well established; the second is exactly what remains investigational.

Back to Table of Contents

Synaptic vs Extrasynaptic Receptors: A Division of Labor

NMDA receptors are not all in the same place, and location changes their job. Receptors sitting within the synapse tend to drive plasticity and cell-survival signaling — the "good" learning-related outcomes. Receptors sitting outside the synapse (extrasynaptic), activated by glutamate that spills over or lingers, are more associated with pro-death signaling when over-activated.

The Papouin 2012 work showed these two pools are gated by different co-agonists: synaptic NMDA receptors are controlled mainly by D-Serine, extrasynaptic ones mainly by glycine (Cell 2012; PMID 22863013). This is a beautiful piece of biological engineering: by assigning different co-agonists to different receptor populations, the brain can, in principle, tune plasticity-promoting signaling and injury-promoting signaling somewhat independently. It also complicates any naive "just add D-Serine" logic, because flooding the brain with a co-agonist could affect the balance between these pools in ways that are hard to predict — another reason careful, dose-controlled research is required rather than casual self-experimentation.

Back to Table of Contents

The Magnesium Block and the Calcium Signal

The NMDA receptor's third requirement — that the neuron already be partly depolarized — deserves its own note, because it is what makes the receptor a true coincidence detector rather than a simple gate. At normal resting voltage, a magnesium ion sits lodged in the channel pore, physically plugging it. Even with glutamate and D-Serine both bound, calcium cannot flow while the magnesium plug is in place.

Only when the postsynaptic membrane is depolarized — typically because other inputs (via AMPA receptors) have already fired and raised the local voltage — does the magnesium ion get electrostatically pushed out, unblocking the pore. Now, and only now, does the fully assembled combination (glutamate + D-Serine + depolarization) let calcium rush in. This is why the NMDA receptor detects coincidence: it opens when presynaptic activity (glutamate release) coincides with strong postsynaptic activity (depolarization), and it does so most reliably when the co-agonist supply of D-Serine is adequate. The calcium that enters is not just an ion current; it is a chemical message that instructs the synapse to remodel itself.

Back to Table of Contents

The Double Edge: Plasticity vs Excitotoxicity

Everything that makes the NMDA receptor powerful also makes it dangerous. The same calcium influx that encodes memory can, in excess, kill the neuron — a process called excitotoxicity. Massive or prolonged NMDA-receptor activation floods the cell with calcium, activating destructive enzymes and free-radical cascades. Excitotoxicity is implicated in the neuronal death of stroke, traumatic brain injury, and some neurodegenerative diseases.

This double edge is why the brain regulates the co-agonist so tightly, and why researchers are cautious about manipulating it. D-Serine sits at the fulcrum: too little and synapses cannot learn; too much and the door to excitotoxic injury opens wider. In fact, D-Serine has been studied as a possible contributor to excitotoxic injury in some disease models, not only as a helper for plasticity. This is not a molecule where "more is better" is a safe assumption, and it is one more reason the human research proceeds slowly and under close supervision.

Back to Table of Contents

What This Does and Does Not Mean for You

Here is the honest translation of the mechanism into everyday terms:

The elegance of the co-agonist mechanism is genuinely one of the highlights of modern neuroscience. Appreciating it is worthwhile in itself. Turning it into a bottle on a shelf is a leap the evidence does not currently support.

Back to Table of Contents

Key Research Papers

  1. Mothet JP, Parent AT, Wolosker H, et al. (2000). D-serine is an endogenous ligand for the glycine site of the N-methyl-D-aspartate receptor. Proc Natl Acad Sci USA. — PMID 10781100
  2. Wolosker H, Blackshaw S, Snyder SH (1999). Serine racemase: a glial enzyme synthesizing D-serine to regulate glutamate-N-methyl-D-aspartate neurotransmission. Proc Natl Acad Sci USA. — PMID 10557334
  3. Panatier A, Theodosis DT, Mothet JP, et al. (2006). Glia-derived D-serine controls NMDA receptor activity and synaptic memory. Cell. — PMID 16713567
  4. Henneberger C, Papouin T, Oliet SHR, Rusakov DA (2010). Long-term potentiation depends on release of D-serine from astrocytes. Nature. — PMID 20075918
  5. Papouin T, Ladepeche L, Ruel J, et al. (2012). Synaptic and extrasynaptic NMDA receptors are gated by different endogenous coagonists. Cell. — PMID 22863013
  6. Wolosker H, Mothet JP (2008). D-amino acids in the brain: D-serine in neurotransmission and neurodegeneration. FEBS J. — PMID 18564180
  7. Wolosker H (2018). The Neurobiology of D-Serine Signaling. Adv Pharmacol. — PMID 29413526
  8. Panizzutti R, Scoriels L, Avellar M (2014). The co-agonist site of NMDA-glutamate receptors: a novel therapeutic target for age-related cognitive decline. Curr Pharm Des. — PMID 24410562

PubMed Topic Searches

Back to Table of Contents

External Resources

Back to Table of Contents

Connections

Back to Table of Contents