Instant Replay: Study Finds Potential Mechanism For Building Long-Term Memory

Researchers led by neuroscientist Joe Tsien found that the brain appears to have a system of repeatedly replaying and reinforcing the same cellular event that led to the initial formation of a memory. The reinforcement is critical for creating the cell-to-cell connections that constitute long-term memories, the researchers found.

"It's really surprising to find out we need to reactivate this initial learning event," said Tsien. "It's like learning something again in your brain, only this time it's due to some kind of spontaneous reactivation mechanism."

This observation could yield insights into the much broader question of how the brain maintains a continuity of knowledge and memories over a lifetime despite the constant turnover of molecules and proteins.

The insight also may one day help with the understanding of human diseases such as schizophrenia, said Tsien. If the process of consolidating new experiences into long-term memories goes wrong, it could result in the incorrect association of a real memory with a mentally created experience, thereby leading to delusions, he said.

Tsien's study, published in the Nov. 10 issue of Science, also is interesting in its development of a cutting-edge genetic engineering technique that allowed him to control the function of a gene not only in a very specific region of the brain, but also at specific points in time. He created a strain of mice that had a modified version of a gene critical for memory formation. He constructed this gene so that it would stop functioning when the animals received a dose of the common antibiotic doxycycline in their drinking water. Taking the drug away restored the gene to its normal function.

The study focuses on a feature called the NMDA receptor, which studs the surface of brain cells and which Tsien and others have shown to be critical for the initial formation of memories. The NMDA receptor has been called a "coincidence detector" because it receives signals from neighboring cells, but only triggers a reaction when it receives two closely timed signals. This characteristic allows the brain to associate two sensory inputs, such as a voice and a face.

In earlier studies, Tsien impaired learning and memory in mice by removing the NMDA receptor, and enhanced those abilities by augmenting the receptor. (The NMDA-enhanced mice have often been referred to as "smart mice.")

In the new study, the researchers compared learning and memory in normal mice and in those engineered to have an NMDA receptor that would turn off in a specific region of the brain with a dose of doxycycline. In one test, the mice were taught to find their way out of a tray of water by finding a hidden platform. During seven training sessions, the mice all learned at the same rate. They were then treated with doxycycline for a week and plain water for another week. Retesting showed that the genetically modified mice that had received the drug were significantly slower at finding their way out of the water, even though they had initially been equally adept at it. Experiments testing emotional memory showed similar results.

The scientists also conducted tests in which they waited longer - until after long-term memory was established - to administer the doxycycline. Those tests showed no impairment of the ability to retrieve information, indicating that NMDA receptor function in the hippocampus, a tiny region inside the brain, is important for what scientists call "memory consolidation," but not retrieval.

Current theory suggests that the shift from short- to long-term memory is driven by a cascade of biochemical reactions unleashed by the original act of learning. These reactions, including the creation of new proteins, build connections between neurons, thus creating new permanent memories.

That theory seemed flawed, said Tsien, because the cascade of reactions lasts only a few hours or couple of days, and the consolidation of long-term memory in the mammalian brain can take weeks, months or even years. "The time scale simply doesn't match." The answer may be that something very similar to the initial learning event happens again and again, a process Tsien calls synaptic reentry reinforcement, or SRR. He believes SRR could be a model for understanding memory over longer time scales.

"Scientists are always mystified by how brain stays connected, how information is stable for most of our lives," said Tsien. "This SRR process could be a crucial mechanism to ensure the stored information is continually consolidated and remains largely stable, despite of the constant turnover of the nuts and bolts of our brains."