Memory reconsolidation and extinction have distinct temporal and biochemical signatures

What was read

• Full published PDF from PubMed Central (PMC6729467): abstract, introduction, materials and methods, results, discussion, figure legends. Read in full.

Question

Retrieval is not passive: it can trigger reconsolidation (a protein-synthesis-dependent re-stabilisation of the original memory, during which the memory is vulnerable) or extinction (new learning of "cue predicts nothing" that competes with the original). Which happens, and what decides it?

Method

• Male C57BL/6 mice; contextual fear conditioning (one 0.75 mA shock, or three shocks for "strong" memories) and the Morris water maze. Systemic anisomycin (ANI, 150 mg/kg, blocks > 90% of brain protein synthesis for about 2 h) or saline 30 min before a re-exposure to the training context without shock; test 24 h later. Also CB1 antagonist (SR141716A), L-type calcium channel blocker (nimodipine), NMDA antagonist (CPP). Experimenters blind to treatment. n = 9–12 per group.
• Variables manipulated: re-exposure duration (0, 1, 3, 30 min; 3, 5, 10 min for strong memories), memory strength (one vs three shocks), memory age (1, 3, 8 weeks), interval from re-exposure to test (2 h, 24 h, 1 week), and for the water maze the number of probe trials (1 vs 10).

Results

1. Duration decides the process. With a one-day-old, one-shock memory: 0 or 1 min re-exposure plus ANI, no effect on later freezing; 3 min plus ANI, freezing lost at test (reconsolidation blocked); 30 min re-exposure, freezing declines within the session in both groups (extinction), and ANI blocks the extinction while leaving the original memory intact at test. Three phases: an initial window in which the trace is immune, a reconsolidation window, then extinction learning.
2. Timing of the deficit. The ANI effect is absent at 2 h after re-exposure and present at 24 h and 1 week: it is a long-term-memory deficit, short-term expression is spared, and protein synthesis is not required for recall itself.
3. Strength. Three-shock memories are not disrupted by ANI after 3 or 5 min re-exposure, but are after 10 min. Stronger memories need longer re-exposure to become labile.
4. Age. One- and three-week-old memories are disrupted after 3 min re-exposure; eight-week-old memories are not at 3 min, but are at 10 min. Older memories need longer re-exposure to become labile; they are not immune.
5. No re-exposure, no effect. ANI in the home cage 24 h after training does nothing, in fear conditioning and in the water maze. Reactivation is required.
6. Generalises to spatial memory. One probe trial plus ANI disrupts the platform memory at 24 h; ten spaced probe trials produce extinction in both groups, and ANI blocks the long-term extinction so ANI mice still search the old quadrant a day later.
7. Different chemistry. CB1 and L-type calcium channel blockade impair extinction but not consolidation or reconsolidation; NMDA blockade and ANI impair all three. Reconsolidation and extinction are dissociable in time, anatomy (amygdala/hippocampus vs medial prefrontal cortex) and molecules.
8. Blocking protein synthesis during a long re-exposure, when both reconsolidation and extinction would run, leaves the original memory unaffected; the two processes interact.

Authors' interpretation: after retrieval, behaviour is governed by two competing processes; re-exposure duration, memory strength and memory age jointly determine which dominates; older memories may sit in a distributed cortical trace that takes longer to destabilise; reconsolidation may be the mechanism by which new information is integrated into old memories.

Limits

• Mouse fear and spatial memory, days to weeks old, one-trial or two-day learning. No human data, no autobiographical or declarative content.
• The disruption is pharmacological. The paper does not test what happens to a memory that is reactivated and then reconsolidates normally, which is the human elicitation case.
• Systemic drug; the anatomical claims are from other work.
• The 3-minute and 10-minute thresholds are protocol-specific; the ordering (longer for stronger or older) is the transferable result.

What the brief uses it for, and whether it holds

Brief section 4: "Whether retrieval destabilises a memory depends on re-exposure duration, memory strength and age." Limit: "Mouse data; the human autobiographical case is not characterised." Section 6.3: "Retrieval can render a memory labile under some conditions [8] ... The claim 'every retrieval destabilises' is too strong; the practical conclusion is not." Section 11: "'Retrieval makes memories labile' is too categorical. Boundary conditions cited [8]; design rules retained."

• Holds exactly. The paper is the right citation for the boundary conditions and the brief's limit is the right one.
• What the paper adds to the design that the brief does not yet say:
  • Re-exposure duration is the controllable variable. Very brief cueing did not open the reconsolidation window; extended cueing did, for every memory tested, including old and strong ones. An elicitation session that dwells on one episode for minutes is, in this model, the long-re-exposure condition. The perturbation audit (section 7) should log dwell time per memory as a covariate, and the agent's cue policy could treat dwell time as a cost.
  • Prolonged re-exposure produced new learning (extinction), not erasure. The human analogue is not forgetting but the creation of a "telling" trace that competes with or updates the original. That is the mechanism behind the brief's rule never to show a model reconstruction and count later agreement as confirmation: reconsolidation "may be used to update or integrate new information into long-term memories".
  • Old memories are not safe. The intuition that decade-old episodes are consolidated and inert is what this paper overturns at 10 minutes of re-exposure. Bonnici's [6] participants retrieved each ten-year-old memory 21 times.
• Caveat for the brief's framing: without a drug, a reactivated memory reconsolidates. The paper is evidence that retrieval opens a window, not that ordinary retrieval degrades content. The risk the brief should name is distortion and updating during the window, with [11] Slotnick & Schacter and the wider false-memory literature as the human side.

Cross-references

• [7] Ryan 2015 cites this paper (ref 24) and shows that reconsolidation-blockade amnesia is optogenetically rescuable, i.e. in the mouse the window affects access more than the trace.
• [1] Abdou 2018 uses retrieval plus anisomycin plus autophagy induction to go beyond this and erase.
• [6] Bonnici 2012 is the human protocol in the evidence base with the most repeated retrievals of the oldest memories.