Engram Cells: Part One: The Memory That Was Still There

Abstract

In 2015, a group at MIT gave mice a drug that blocks memory consolidation, confirmed the mice had no memory of what happened to them, and then switched the memory back on with a pulse of blue light.

The animals froze in fear at a context they could not otherwise recall. The memory had not been erased. It had become unreachable.

This is the first of two parts on engram cells — the population of neurons that physically holds a specific memory. It is the most spectacular result in this library, and it is entirely in mice, by a method that cannot be used in a person.

1.How You Label a Memory

The technique rests on a trick of timing. When a neuron is strongly active, it switches on immediate-early genes, of which c-fos is the best known. If you couple the c-fos promoter to a genetic switch, then any cell active during a particular window gets permanently tagged.

In Tomás Ryan's experiments, the dentate gyrus engram was tagged with a fluorescent marker, mCherry, using learning-dependent c-fos-driven labelling. The cells active while the animal learned became visible, and stayed visible afterwards.

Then the second half: presynaptic neurons in the entorhinal cortex were made to express channelrhodopsin-2, a light-sensitive ion channel, delivered by virus. Shine blue light and those cells fire.

Together these give you a way to find the cells that were active during a specific experience, and to switch them on again later, at will.

2.The Experiment

Mice underwent contextual fear conditioning in one chamber — they learned that a particular place was dangerous. Immediately after training, half received anisomycin, a protein synthesis inhibitor that prevents memory consolidation. The others received saline.

The authors verified the drug worked, confirming that the dose used did inhibit protein synthesis in the dentate gyrus. A second inhibitor, cycloheximide, produced comparable results.

Tested normally, the anisomycin mice were amnesic. Returned to the dangerous chamber, they did not freeze. By every standard behavioural measure the memory was gone.

Then the animals were placed in a different chamber — a neutral one they had no reason to fear — and the tagged cells were stimulated with light in alternating three-minute epochs across a twelve-minute session.

Both groups froze significantly during the light-on epochs, at equivalent levels. The amnesic mice were expressing a fear memory they could not otherwise retrieve, in a place where nothing had ever happened to them.

3.What Was Missing, and What Was Not

The physical differences between consolidated and amnesic engrams were then measured, and this is where the result becomes interpretable rather than merely startling.

Consolidated engram cells in the saline group showed significantly greater AMPA to NMDA current ratios than both non-engram cells and engram cells from the anisomycin group. They also showed increased dendritic spine density; in the anisomycin group, spine densities of tagged and untagged cells were similar.

So the amnesic engram lacked both hallmarks of consolidation: it had not been strengthened, and it had not grown new connections.

What it did retain was its wiring. Amnesic engram cells kept their preferential connectivity with downstream engram cells in CA3 — and this engram pathway-specific connectivity was unaffected by post-training anisomycin.

The pattern survived. The amplification did not.

4.Storage Against Retrieval

The authors' interpretation reverses an assumption a century old. Synaptic strengthening had been treated as the memory itself — the physical form the trace takes.

Their proposal is that the rapid increase in synaptic strength and the growth of dendritic spines during consolidation are not a crucial component of the stored memory. What may represent a fundamental mechanism of memory information storage is the persistent, specific connectivity between engram cells.

And the role of the strengthening is then something else: to provide natural recall cues with efficient access to the soma of engram cells, so that they can be reactivated.

On that reading, consolidation does not write the memory. It builds the road to it. Amnesia becomes a retrieval failure — the trace intact, the access route missing.

The finding held across several variations: optogenetic place avoidance, engram activation in CA1 and in the lateral amygdala, and a reconsolidation amnesia paradigm.

5.What This Cannot Establish

Everything above depends on optogenetics, which requires introducing a virus into the brain, expressing a foreign light-sensitive protein in specific neurons, and implanting a fibre to deliver light. None of this is possible in a human being, and no version of it is close.

The memory in question is also a specific kind: a fear association formed in minutes, in a rodent, about a place. Human episodic memory is richer, more reconstructive, and less well modelled by freezing behaviour than the elegance of this experiment can suggest.

And a stimulated engram is not a recalled memory. Driving a labelled population with light produces a behaviour that resembles recall. Whether the animal experiences anything corresponding to remembering is not a question this method can reach.

Those limits are worth holding because of how this work travels. It is routinely reported as showing that lost memories can be recovered — which, in mice, under these conditions, is exactly what it shows, and which says nothing about anyone's lost memories.

Editorial Comment

MindHeaven® makes no claim relating to memory engrams, consolidation, retrieval or recovery of lost memories, and nothing in our formulations is proposed to influence any of them. No supplement has been shown to act on any mechanism described here.

We publish this because it is the clearest demonstration in modern neuroscience that a memory is a physical object in a specific set of cells — and because the same result draws a hard line under what physical means. The trace survived a drug that abolished the behaviour.

Part Two follows the same logic into a mouse model of early Alzheimer's disease, where the distinction between a memory that is gone and a memory that cannot be reached stops being an abstraction.