A new strain of genetically engineered mice has allowed researchers to pinpoint, for the first time, the precise cellular connections that form as a memory is created. By tracing a protein tagged to glow fluorescent green as it migrates through individual neurons, from the cell body out through the branching dendrites, the researchers could see exactly which synapses–connections to other neurons–were involved when the mice learned to fear an electric shock.

“It’s a first step in visualizing the synapses that encode memories,” says Stephen Maren, director of the neuroscience graduate program at the University of Michigan, who was not involved with the research. “We really haven’t had a tool like this to see memory encoding at a synaptic level. It’s an exciting paper.”

“We are developing techniques that allow us to focus on the actual physical sites that are changing in the brain with learning, at finer and finer resolution,” says the study’s lead investigator, Mark Mayford, associate professor of cell biology at the Scripps Research Institute.

Neuroscientists believe that in order for a memory to form, individual synaptic connections must be strengthened in response to a memory-generating stimulus. This strengthening is likely the result of a specific set of proteins migrating to synapses in a precisely choreographed pattern, but it remains a mystery which proteins are involved and how they are targeted to their destinations. The new study, which appears in today’s issue of Science, is the first to trace a particular protein as it makes its way to particular synapses.

The studied protein is a receptor for glutamate, a neurotransmitter previously implicated in memory formation. The researchers engineered a strain of mice so that the glutamate receptor would glow green under extremely specific, manipulable circumstances.

The genetically modified mice were then trained to expect an electric shock to their feet whenever they were placed in a certain box. The resulting fear is “a very long-lasting, very robust memory,” says Mayford. Presumably, he says, the neurons activated as the mice learned to fear the box were those responsible for forming the aversive memory.

The fluorescently tagged glutamate receptor was modified so that neurons would only manufacture it when they became active. This allowed the group to identify which neurons contributed to the memory formation by following the green glow.

In addition, the researchers could completely turn off the entire tagged protein system by administering the drug doxycycline. The mice were fed doxycycline throughout their lives–right up until the learning task, and again when the task was over. In this way, the tagged protein was manufactured only during the formation of this particular memory.

“You’re capturing only the events surrounding the learning episode,” says Craig Powell, assistant professor of neurology and psychiatry at University of Texas Southwestern Medical Center, who was not involved with the research.

Mayford’s group followed the glowing glutamate receptor as it migrated through neurons in a region called the hippocampus by examining brain slices at several time points after the learning task. They found that after the protein was manufactured in the nucleus, it traveled outward through the cell’s many branching dendrites and eventually settled in far-flung synapses.

Surprisingly, the protein preferentially lodged itself in one kind of synapse. Synapses come in a few flavors, depending on whether they’re formed by so-called thin, stubby, or mushroom spines protruding from the cell. The tagged glutamate receptor migrated primarily into the mushroom-type synapses.

“I think the most important thing about this study is that it suggests that a specific type of spine may be more important for learning and memory processes than other types of spines,” says Powell.

The receptor’s “preference” for mushroom-type synapses suggests that, at least in the process of forming a fear-related memory, there is a specialized trafficking system to direct synaptic proteins to their targets. “But what sort of molecular flag gets waved to say, ‘Come up here and make your home at my type of synapse,’ is not really clear,” says Maren.

Another mystery is why the tagged receptor disappears from the synapses after 72 hours, when the memory persists much longer. Other proteins and other brain areas are almost certainly involved in forming and maintaining the memory. The amygdala in particular probably plays a key role. While the hippocampus is critical for encoding information about place–in this case, the box where the shocks were administered–the amygdala seems to tie that information to the fear response produced by the shocks.

“The hippocampus is probably not the final storage site,” says Maren. “If you really wanted to see where the long-term memory was encoded for this type of learning, you probably want to look at the amygdala.”

In previous investigations of the amygdala using similarly engineered mice, Mayford’s group showed that the same neurons are activated both when a memory is formed and when it is later retrieved. In future studies, the researchers may apply the new finer-scale approach to probe memory formation in the amygdala.

Mayford also hopes to use the new technique to elucidate the precise structure of a memory encoded by the hippocampus–in particular, a memory of the box. He plans to determine whether he can teach a mouse that’s never been shocked inside the box to fear it nonetheless. To do so, he would activate the hippocampal neurons that encode the memory of the box, and then give the mouse a shock.

If the experiment is successful, it could help explain how the box is represented within the mouse’s brain. “One of the big questions in neuroscience,” says Mayford, “is, what does it take to make a representation of the external environment?”