By Lizzie Buchen
Just as a whiff of pumpkin pie can unleash powerful memories of holiday dinners, the stimulation of a tiny number of neurons can evoke entire memories, new research in mice suggests.
Memories are stored in neurons distributed across a host of brain regions. When something triggers a memory, that diffuse information is immediately and cohesively reactivated, but it's unclear how the circuit gets kicked into full gear. Over the past few years, a handful of studies have suggested that a small number of neurons -- perhaps even single neurons -- can trigger sensations. But this idea remains controversial and has never been demonstrated with memory.
Now, Michael Häusser and his colleagues at the the Wolfson Institute for Biomedical Research at University College London have developed a way to activate a small subset of the neural circuit that underlies a specific memory. They found that this handful of neurons was able to activate the rest of the circuit and get animals to recall the memory. The team's findings were presented in Chicago at the annual meeting of the Society for Neuroscience4.
"I was surprised that such a small spark could bring the memory back to life," Häusser says. "It gets at a fundamental question in memory research: what fraction of the cells in a network that are used to store information are required to reactivate that memory?"
To target the neurons involved in memory formation, Häusser's group used a combination of genetic tricks. The researchers took advantage of a gene called c-fos -- which is expressed by recently activated neurons -- and used the portion that controls gene expression to turn on the light-sensitive protein channelrhodopsin. They inserted the gene for the c-fos-controlled channelrhodopsin into cells in a portion of the hippocampus that is known to have a crucial role in memory formation.
Häusser and his team then trained mice to associate a painful electric shock with a tone. Whenever the mice heard the tone they would stop moving, bracing themselves for the imminent shock. When the neurons involved in remembering the association fired during the learning process, c-fos became activated. And for the small number of neurons that had received the gene, the light-sensitive channel was activated too.
"We basically get the brain to select the cells that express the channel," says team member Kate Powell. "The stronger the cells are activated, the stronger the channel is going to be expressed. We're really getting the cells that are most important for the memory."
Neurons that express channelrhodopsin will fire when exposed to blue light. When the researchers shined blue light onto the hippocampus using a laser and an implanted fibre-optic cable, the mice promptly froze, suggesting that the light had triggered the fearful memory. Activating about a hundred cells, and as few as 20 cells, was enough to trigger the memory.
But when the group inserted channelrhodopsin into random cells in the hippocampus, they were unable to make the mice freeze -- indicating that only the cells that were activated during memory formation can trigger the memory.
"It's a remarkably small number of neurons, when such a huge number is thought to be involved," says Powell. There are about one to two million cells in this region of the hippocampus, she says, and between 5-15% are activated during the fear-learning process.
Häusser says he wants to try to target smaller and smaller populations to find out the minimum number required to evoke the memory. "The brain is very efficient in terms of its storage and processing. There are lots of computational and energetic and wiring advantages to having a small number of cells being effective," he says.
A couple of recent papers have shown that precise, light-controlled activation of certain brain areas can affect memory formation, but these were not specific to the circuit underlying the memory.
"It's a beautiful demonstration, and a very clever, very precise technique," says neuroscientist Serge Charpak at INSERM, France's main biomedical research agency, in Paris. "I certainly wouldn't have expected so few neurons to be enough to get this behaviour."
"So far all manipulations of neural activity have targeted local clusters of neurons or certain neural cell types," adds Michael Brecht at the Bernstein Center for Computational Neuroscience Berlin, who studies neural circuitry. "If the conclusions turn out to be correct, such highly selective manipulations suggest that the brain might actually compute with small, precisely selected sets of neurons."
The latter idea started to gain support in the field only recently, and is "a major departure from the mass action views of neural activity that were taught just 10 years ago", says Brecht.
The authors note that the fear-conditioning paradigm, although robust and thoroughly studied, is a crude probe. They hope to be able to use their genetic tool to probe more subtle memories, and to search for other regions of the circuit that might be able to trigger the behaviour.
"It's really cool that you can select cells based on their activity," says Powell. "It gives us a lot of options for investigating memory in a new way."