Long-Term Potentiation: Part Two: The Memory Molecule

Abstract

If a synapse becomes stronger and stays stronger for days, something has to hold the change in place — through protein turnover that replaces the molecules involved many times over.

The leading candidate has been the same enzyme for four decades: calcium/calmodulin-dependent protein kinase II. Roger Nicoll and Howard Schulman reviewed the case in Physiological Reviews in 2023, and open with a line that captures the field's mood: CaMKII and long-term potentiation were discovered within a decade of each other and have been inextricably intertwined ever since — but like many marriages, it has had its ups and downs.

They set out a seven-step model, and then do the thing that makes the review valuable: they state which steps are firmly grounded and which are not.

1.Why a Memory Molecule Is Needed

The problem is persistence. Proteins in a synapse are continuously degraded and replaced, on timescales of hours to days. A memory lasting years cannot be a particular set of molecules sitting there.

What is required is a mechanism that reconstitutes itself — something that, once switched on, keeps switching on its own replacements.

CaMKII was proposed as a memory molecule on exactly these grounds, on the basis of its unusual biochemistry, before any physiological link to LTP had been made. It is a rare case of a molecule being nominated for a role on theoretical grounds and then, slowly, being shown to occupy it.

The authors note how long that took. The convincing linkage of CaMKII to synaptic physiology and behaviour took many decades, and depended on technologies that did not exist at the outset: brain slices, mouse genetics, single-cell molecular genetics, pharmacological reagents, protein structure and two-photon microscopy.

2.The Seven Steps

Their proposed model is unusually explicit, and worth setting out because most accounts of LTP stop at "calcium comes in and the synapse gets stronger".

Calcium enters through NMDA receptors and activates CaMKII. The enzyme then autophosphorylates — modifies itself — which makes it constitutively active, no longer requiring calcium, and exposes a binding site for the NMDA receptor subunit GluN2B.

Active CaMKII moves to the postsynaptic density and binds the cytoplasmic tail of GluN2B. That complex initiates a structural rearrangement of the postsynaptic density, which may involve liquid–liquid phase separation — the formation of a distinct molecular droplet within the cell.

The rearrangement involves the scaffolding protein PSD-95 and results in accumulation of AMPA receptors at the synapse. More receptors means a larger response to the same transmitter release, which is the potentiation.

And the persistence: stability of the modified postsynaptic density is maintained by stability of the CaMKII–GluN2B complex, which survives protein turnover through subunit exchange or phosphorylation between holoenzymes. The complex replaces its own parts while remaining the same complex.

3.Autophosphorylation as a Switch

The second step carries most of the conceptual weight and deserves separating out.

An enzyme that requires calcium to function is a detector: it reports what is happening now. An enzyme that modifies itself so it no longer requires calcium has become a record: it reports that something happened.

That transition — from calcium-dependent to constitutive activity — is the moment a transient event becomes a stored one, expressed in a single molecular step. It is why the molecule was nominated before the evidence existed.

4.What the Authors Do Not Claim

The review's most useful feature is its refusal to present the model as settled. The authors state that they critically discuss the data underlying each step, that some steps are more firmly grounded than others, and that they offer suggestions for how the weaker evidence could be strengthened or, on new data, replaced.

That sentence is unusual in a review of a forty-year-old hypothesis by two of its principal architects. It also means a reader should treat the seven steps as a scaffold with known soft joints, not as a mechanism established end to end.

Their closing assessment is optimistic and carefully bounded: the prospect of having a detailed cellular and molecular understanding of learning and memory is at hand. At hand, not achieved.

5.The Problem This Model Does Not Solve

Everything above explains how a synapse could remain strong. It does not establish that the strength is the memory.

The engram work covered elsewhere in this section bears on precisely this. Amnesic mice retained a memory that could be reactivated optogenetically while lacking the AMPA-to-NMDA ratio increase and the spine density growth that this model produces. The molecular machinery of potentiation was absent; the trace was not.

One reading is that CaMKII-dependent potentiation builds access to a memory held in the pattern of connections, rather than constituting the memory. On that account the seven steps above are correct about what they describe, and describe something other than storage.

Part Three takes up a different challenge to the synaptic account, from a mechanism that is not synaptic at all.

Editorial Comment

MindHeaven® makes no claim relating to CaMKII, AMPA receptors, synaptic potentiation or memory formation. Nothing in our formulations has been shown to act on any step of this model, and we are not aware of any supplement that has.

The reason to publish a molecular review in a library aimed at general readers is the shape of the argument rather than the biochemistry. Here is a hypothesis proposed on theoretical grounds in the 1980s, pursued for forty years, supported step by step as new methods arrived — and still described by its own proponents as having steps that need strengthening or replacing.

That is what a mature scientific claim looks like from the inside, and it is a useful calibration against the certainty with which mechanisms are invoked in marketing.