#SfN12 Biasing memory storage in neurons


This is part of a blog series for the 2012 Society for Neuroscience annual meeting. You can read the rest of it here.

BBB19 200.06 Memory allocation mechanisms for trapping and activating emotional memories


While neurons may seem like similar entities, at any point in time, they’re actually quite variable in their inner protein makeup. Just like your experiences defines a unique YOU, past activations or inhibitions of a neuron can also set it in a particular “state”. In particular, some neurons in the amygdala (the “emotion centre”) may have relatively higher levels of a protein called CREB.

When it comes to storing a new emotional memory, not all neurons are equal. Neurons with more CREB seemed to be preferentially “selected” by the amygdala to encode this memory. As the old saying goes, “neurons that fire together wire together”, hence the selected high-CREB neurons should form their own little niche that creates a specific activation pattern in the network, what we call a “memory trace”.

That’s the main idea. So how to test this?

In comes the super-mouse with neurons that are light-activated and have artificially enhanced CREB levels in them. Light-activated? There is a special type of receptor, ChR2, which can respond to light shown in the brain, and activate whichever neuron it’s on. As a clever twist, not ALL amygdala neurons have these traits – some were left normal as is. The researchers also generated slightly less-cool mice with only light-activated neurons. Both types of mice were trained to associate a tone with a sharp electric foot shock. Very soon, just hearing the tone alone is enough to worry the mouse – and they demonstrate this by freezing in place. This type of fear memory is encoded by the amygdala, and if the theory holds true should be preferentially stored in the special CREB+++/light-sensitive cells.

After a single training session, the researchers used light to activate neurons in the brain. In the super-mice, light activation caused a lot more freezing than in mice that only have the light-activated neurons. Why? For the super-mice, since ALL light-activated neurons have increased CREB levels, it’s conceivable that the memory is almost completely stored in these cells. When they’re activated, the memory can be strongly retrieved. On the other hand, for mice that only have light-activated channels, the fear memory is stored randomly in a subset of neurons – some of which may have ChR2 channels, others which may not. Hence, only part of the memory trace is recalled when ChR2 neurons are activated by light. This strongly supports the idea that neurons with increased CREB are preferentially used to store memory.

So how is CREB biasing a memory towards certain “privileged” neurons for storage? CREB is known to make neurons more excitable – could this be the reason? To test this, the researchers equipped mice with TWO types of light-activated channels. The first responds to blue light, and allows the neuron to be activated for a long time. They then separated these mice into three groups. The other is sensitive to orange light, and activates the neuron. In the first group, the researchers turned on blue light BEFORE any training, making a population of neurons hyper-active during training, then turned on orange light. In another, they didn’t bother with blue light, trained the animals as usual before activating neurons with orange light. In the last group, researchers trained the animal, then activated neurons with orange light, THEN shown blue light to excite the same neurons.

Without pre-excitation with blue light, none of the neurons in the amygdala particularly stood out; hence the fear memory was randomly distributed. So when neurons were activated with orange light, some –but not all – of them contained a bit of the memory trace, so the last two groups did show some freezing. However, because much of the memory trace is stored in non-light responsive neurons, the mice didn’t get the full memory back (in terms of strength and information). In contrast, when a sub-population of neurons were put in a hyperactive state by blue light BEFORE the training, these neurons were selected in encoding the memory – hence when the same neurons were later activated by orange light, nearly ALL of the memory trace was recalled – behaviorally, they showed much more freezing than the other two groups.

This is a super cool study showing that CREB overexpression can lead to excitation of certain neurons; and if this excitation overlaps with memory encoding, the memory will be preferentially stored in these neurons.  Of course, if you’re thinking “next step: Eternal Sunshine of the Spotless Mind!”, most neurons don’t just store one single memory, so killing off neurons involved in one bad memory will likely wipeout multiple other memories.