Memory storage and sleep consolidation run on separate tracks
In mice, fear memories endured after most hippocampal synapses vanished, while thalamic support cells set how well sleep locked learning in place. Storage and stabilization look like separable processes.
Science··Night
The fear chamber is still known after synapse loss
New Scientist reports that researchers at the Okinawa Institute of Science and Technology found a fear memory still intact in mice after most of the hippocampal connections that carried it were lost. The animals learned to fear an alcohol-scented chamber; about half then spent two days in drug-induced artificial hibernation that slowed metabolism. Imaging showed that more than half of the synapses on the hippocampal engram neurons disappeared in that period. Five days after waking, the mice again froze in the chamber where they had been shocked. Synapses sitting in clusters on those neurons were retained while scattered ones vanished, and the scattered contacts regrew within a day of arousal. Kazumasa Tanaka, who led the work, reads clustered contacts as holding memory content and the others as helping reach it. The study appeared in Science. It was done in mice, the hibernation was induced rather than natural, and the molecular route that protects the clusters remains unidentified. Those limits block any direct transfer to human memory, and the report does not claim a cross-species generalisation. Even so, the mouse hippocampus offers a concrete laboratory picture in which storage and access can come apart: content can survive inside a tight spatial arrangement of contacts.[1]
Thalamic support cells set sleep-time consolidation
Medical Xpress reports that deleting the NFIX gene from mature astrocytes changed cell shape in the mouse brain in only one place, the thalamic reticular nucleus. The animals kept much of their sleep pattern yet failed working memory, object recognition and spatial tasks. Astrocytes were long treated as housekeeping cells for neurons; here the loss of NFIX left them shorter and less branched in that hub and lowered MAOB and P2RX7, proteins the cells need to make and release the inhibitory messenger GABA. Restoring either protein brought back part of the normal signalling and part of the memory performance. The reticular nucleus gates traffic between thalamus and cortex during sleep, when the day's experience is replayed and stabilised. The work was published in Neuron. It is a mouse study built on deleting a gene from birth in one cell type; the report does not give effect sizes, and whether the same route operates in other brain regions or other species is untested. Weakened consolidation of learning while sleep architecture holds suggests that stabilisation can depend on a thalamic gate separate from the hippocampal storage pattern, still inside mouse data only.[2]
Storage and stabilisation are discussed side by side
Two mouse studies light different moments of memory, and neither result is enough to carry into the human brain. On the hippocampal line, a fear memory survives the loss of more than half its synapses through clustered contacts, while scattered ones form again after arousal. On the thalamic line, when astrocyte shape and the GABA path fail in the reticular nucleus, sleep largely holds while consolidation of learning weakens. The shared reading is that where content is held and how it is locked during sleep are processes laboratory interventions can touch separately. An alternative reading is that artificial hibernation or lifelong gene deletion produces each outcome through idiosyncratic side effects, so storage and consolidation may not part so cleanly in nature. The New Scientist and Medical Xpress reports do not validate each other; both are mouse models, and both show that a memory trace is not merely the strength of a single cellular contact. What lands on the night desk is concrete: memory research is setting a loss-resistant storage pattern beside a sleep-dependent stabilisation gate on the same day, and holding both claims inside their species limit.[1], [2]
Related columns
For more information on this topic, you can read the related columns.
