Dynamic and selective engrams emerge with memory consolidation

Douglas Feitosa Tomé, Ying Zhang, Tomomi Aida, Olivia Mosto, Yifeng Lu, Mandy Chen, Sadra Sadeh, Dheeraj S. Roy, Claudia ClopathNature Neuroscience 2024 · DOI

This paper, published in Nature Neuroscience by Douglas Feitosa Tomé, Ying Zhang, and colleagues, challenges the classical view of the "memory engram" as a static snapshot of neuronal activity. Instead, it presents a model where memories are dynamic, evolving through a process of "inhibitory sculpting" to become more precise over time.

TL;DR

Memory engrams are not fixed; they are dynamic ensembles where neurons "drop in" and "drop out" during the hours following learning. This study demonstrates that inhibitory synaptic plasticity—specifically from CCK+ interneurons—is the essential mechanism that transforms a broad, unselective memory into a sharp, selective one during consolidation.

What Problem Does This Paper Address?

In neuroscience, a "memory engram" is the physical trace of a memory in the brain, typically defined as a population of neurons activated during learning that are necessary and sufficient for recall. However, a persistent mystery remains: when researchers compare neurons active during initial learning to those active during recall a few days later, the overlap is often surprisingly low (only 10–40%).

This raises a fundamental question: Is this "turnover" just noise, or does it serve a functional purpose? Furthermore, why do memories often start out "blurry" (generalizing to similar environments) and become more "specific" (selective) over the first 24 hours? The authors sought to bridge the gap between dynamic neuronal composition and the emergence of memory selectivity.

Methods

The researchers employed a "dual-track" approach, combining computational modeling with advanced in vivo mouse experiments:

  • Computational Modeling: They developed a Spiking Neural Network (SNN) of the hippocampus. This model incorporated multiple forms of plasticity: excitatory (Triplet STDP, heterosynaptic, and transmitter-induced) and, crucially, inhibitory synaptic plasticity.
  • Cal-Light Labeling: To bridge the gap between activity and gene expression, they used the "Cal-Light" system. This allows for the labeling of neurons that are active only during a specific window (like fear conditioning) with a fluorescent marker (EGFP).
  • Longitudinal Calcium Imaging: Using GRIN lenses and head-mounted miniscopes, they tracked the activity of the same individual neurons in the mouse Dentate Gyrus (DG) across multiple days.
  • Optogenetics & Chemogenetics: They used light-sensitive proteins (optogenetics) and designer drugs (chemogenetics) to selectively silence specific types of inhibitory interneurons (CCK+ and PV+ cells) or block their ability to change their synaptic strength (plasticity) during the consolidation period.

Datasets

As a biological and computational study, the "datasets" consist of:

  • Behavioral Data: Freezing levels of mice in "Training" vs. "Neutral" contexts to measure memory selectivity.
  • Imaging Data: Calcium traces ($\Delta F/F$) from hundreds of individual neurons in the mouse Dentate Gyrus, tracked longitudinally.
  • Simulation Data: Synthetic spike trains and synaptic weight distributions generated by their SNN model.
  • Histology: Confocal imagery and in situ hybridization to verify the identity and labeling of CCK+ and PV+ interneurons.

Key Insights and Ideas

  • Engrams are "Drop-in/Drop-out" Clubs: The engram is not a fixed set of cells. During consolidation, some neurons that were active during training leave the ensemble, while new ones are recruited. The model predicted—and experiments confirmed—that neurons with higher initial firing rates are more likely to stay in the engram.
  • Selectivity is Earned, Not Given: Immediately after training, mice (and the model) were unselective, showing fear in both the training context and a similar neutral context. Selectivity emerged roughly 12 hours later.
  • Inhibition as the Sculptor: While excitatory plasticity helps form the memory, it is inhibitory plasticity that makes it selective. Specifically, CCK+ interneurons (but not PV+ interneurons) increase their "grip" on the network during consolidation. This suppresses "off-target" neurons, effectively performing a biological version of pattern separation.
  • Functional Representational Drift: The study suggests that "representational drift" (the changing of neural codes over time) isn't a bug of a biological system; it is the mechanism by which the brain refines and specifies memories.

Key Results

  • Selectivity Emergence: Behavioral experiments showed a significant "discrimination index" (the ability to tell contexts apart) appearing at the 12-hour and 24-hour marks, but not at 1 or 5 hours.
  • Causal Role of CCK+ Cells: Optogenetically inhibiting CCK+ interneurons during recall completely abolished memory selectivity, causing mice to freeze in neutral environments.
  • Plasticity is Mandatory: Chemogenetically blocking the plasticity of CCK+ cells specifically during the consolidation window (the first few hours after training) prevented the memory from ever becoming selective, even though the memory itself was still stored.
  • Model Validation: The SNN model's predictions regarding engram turnover and the necessity of inhibitory plasticity were consistently validated by the biological results.

Limitations and Future Work

  • Sample Size: The authors acknowledge that the longitudinal calcium imaging was performed on a relatively small number of mice ($n=4$), making those specific results preliminary.
  • Regional Specificity: The study focused on the Dentate Gyrus. How these dynamic engrams interact with other hippocampal regions (CA3, CA1) or the cortex during long-term "systems consolidation" (weeks to months) remains an open question.
  • Sleep: While the paper mentions that reactivation often happens during sleep, it did not explicitly monitor sleep stages to see if the "drop-in/drop-out" process is sleep-dependent.

This work provides a compelling new framework for understanding memory: it’s not just about which neurons fire together, but how the brain’s inhibitory network "prunes" that activity to ensure our memories remain distinct and useful.

Summary updated

Abstract

Episodic memories are encoded by experience-activated neuronal ensembles that remain necessary and sufficient for recall. However, the temporal evolution of memory engrams after initial encoding is unclear. In this study, we employed computational and experimental approaches to examine how the neural composition and selectivity of engrams change with memory consolidation. Our spiking neural network model yielded testable predictions: memories transition from unselective to selective as neurons drop out of and drop into engrams; inhibitory activity during recall is essential for memory selectivity; and inhibitory synaptic plasticity during memory consolidation is critical for engrams to become selective. Using activity-dependent labeling, longitudinal calcium imaging and a combination of optogenetic and chemogenetic manipulations in mouse dentate gyrus, we conducted contextual fear conditioning experiments that supported our model's predictions. Our results reveal that memory engrams are dynamic and that changes in engram composition mediated by inhibitory plasticity are crucial for the emergence of memory selectivity.
BibTeX
@inproceedings{tome2024dynamic,
  author = {Douglas Feitosa Tomé and Ying Zhang and Tomomi Aida and Olivia Mosto and Yifeng Lu and Mandy Chen and Sadra Sadeh and Dheeraj S. Roy and Claudia Clopath},
  title = {{Dynamic and selective engrams emerge with memory consolidation}},
  booktitle = {Nature Neuroscience},
  year = {2024},
  doi = {10.1038/s41593-023-01551-w},
}
Figure