Putting mice into hibernation causes a major loss of synapses
Researchers at OIST induced a hibernation-like state (QIH) in mice, which erased over half of their hippocampal synapses, yet all tested memories remained completely intact after arousal 82% of eliminated synapses reappeared at the exact same location on the same dendrite, far above chance levels, suggesting memory traces survive synaptic turnover through structural reconstitution Engram synapses arranged in tight spatial clusters on dendrites were selectively preserved, while isolated engram sy
Analysis
TL;DR
- Researchers at OIST induced a hibernation-like state (QIH) in mice, which erased over half of their hippocampal synapses, yet all tested memories remained completely intact after arousal
- 82% of eliminated synapses reappeared at the exact same location on the same dendrite, far above chance levels, suggesting memory traces survive synaptic turnover through structural reconstitution
- Engram synapses arranged in tight spatial clusters on dendrites were selectively preserved, while isolated engram synapses were eliminated during hibernation
- Approximately one-third of clustered engram synapses were attached to multisynaptic boutons—rare structures where one presynaptic terminal connects to multiple postsynaptic spines—offering a potential mechanistic explanation for memory resilience
- The findings challenge the dominant hypothesis that memory is stored solely in individual synaptic strengths and point toward clustered, structurally reinforced synaptic architectures as the true substrate of long-term memory
Why It Matters
This research fundamentally challenges the long-held synaptic efficacy hypothesis of memory storage, suggesting that the brain employs redundancy and structural clustering to protect memories against the inevitable turnover of synaptic connections. For AI and computational neuroscience, it raises important questions about how artificial systems might achieve robust, long-term memory storage in the face of continual plasticity and structural change—insights that could inform more resilient memory architectures in neural networks.
Technical Details
- QIH protocol: Artificial activation of Q neurons in the hypothalamus induces a hibernation-like state in mice, reducing body temperature to ~20°C, decreasing heart rate and breathing significantly, and suppressing hippocampal neuronal activity by ~70% for 48 hours. The state is fully reversible on demand.
- Synaptic quantification: Serial block-face scanning electron microscopy was used to image hippocampal tissue before, during, and after QIH, revealing that more than 50% of synapses were eliminated during hibernation.
- Memory assays: Two hippocampus-dependent tasks were used—contextual fear conditioning (association of a box with a mild shock) and a plus-maze navigation task. Post-hibernation performance was indistinguishable from non-hibernating controls, and post-training lesions of the hippocampus confirmed the memories were hippocampus-dependent.
- Engram mapping via eGRASP: The eGRASP technique labeled synapses formed between neurons co-activated during learning, revealing that isolated engram synapses were eliminated while spatially clustered engram synapses survived the synaptic purge.
- Multisynaptic boutons: ~33% of clustered engram synapses were found on multisynaptic boutons (one presynaptic terminal connecting to multiple postsynaptic spines), compared to only 3.3% in randomly sampled non-engram synapses, suggesting this rare structure may serve as a protective anchor for memory traces.
Industry Insight
- The discovery that memory survives massive synaptic turnover through clustered, structurally reinforced connections suggests that redundancy and spatial organization—not just individual connection strength—are critical for durable information storage, a principle that could inspire more fault-tolerant memory systems in neuromorphic and AI architectures.
- The role of multisynaptic boutons as engram protectors highlights the importance of studying rare or unconventional neural structures; similarly, AI research may benefit from exploring underutilized architectural patterns (e.g., multi-head attention, shared weights) that provide robustness beyond standard designs.
- The QIH technique itself represents a powerful experimental tool for dissecting the relationship between synaptic dynamics and memory persistence, and analogous "stress test" approaches could be valuable for evaluating the durability of memory mechanisms in artificial systems under extreme perturbation.
Disclaimer: The above content is generated by AI and is for reference only.