日本冲绳科学技术大学院大学（OIST）的研究团队透过人工冬眠技术研究小鼠的大脑记忆机制。实验发现，小鼠在进入类似冬眠的低体温与低代谢状态后，海马体中超过半数的突触消失且神经活动显著下降，但小鼠在苏醒恢复后依然能精准记住先前的恐惧制约与迷宫任务，这项发现挑战了传统认为记忆必须仰赖整个突触网络维持不变的观点。（关键数字：70）

研究人员进一步利用光学与电子显微镜观察记忆印迹（engram）细胞间的连接，发现散落的突触数量大幅减少，但聚集在同一树突分支上的「群聚突触（clustered synapses）」却以正常密度完整保存下来。有趣的是，这些留存的突触突起并非特别粗大，甚至比邻近突起更小，推翻了过去认为记忆主要依靠单个突触结构增大来强化的理论。

实验亦显示单纯的突触再生并不足以挽救记忆，唯有群聚突触完整留存的小鼠才能成功回忆。研究团队指出，群聚突触与记忆留存之间存在高度相关，未来的关键研究将聚焦于透过专门破坏或重建这些突触群聚，以确认其是否为维持长期记忆的关键物理架构，该研究成果已正式发表于《科学》（Science）期刊。

A research team at the Okinawa Institute of Science and Technology (OIST) utilized artificial hibernation to investigate memory mechanisms in mice. The study revealed that although mice lost more than half of their hippocampal synapses and exhibited a 70% reduction in neural firing during hibernation, they fully retained their previously learned fear conditioning and maze navigation upon warming up, challenging the classical view that memory stability requires the entire synaptic network to remain intact.

By examining connections specifically between memory-tagged engram cells with light and electron microscopy, researchers discovered that while scattered synapses were lost, clustered synapses packed tightly on dendritic branches survived at normal densities. Intriguingly, these spared clustered spines were smaller than neighboring ones rather than enlarged, contrary to the longstanding hypothesis that long-term memory relies primarily on the physical enlargement of individual synaptic spines.

The findings also demonstrated that the structural regrowth of lost spines alone is insufficient to preserve memory if the clustered engram configuration is disrupted. The team noted that while clustered synapses correlate strongly with memory recall, future decisive tests involving targeted disruption and reconstruction of these clusters will determine if they serve as the fundamental physical architecture for enduring memories; the full study was published in the journal Science.