How does our brain make memories? We know that a brain region called the hippocampus is important for forming new memories. If this region is damaged or inhibited in mice or humans, memory formation is impaired.
It has been hypothesized recently that a relatively small population of hippocampal neurons becomes activated during learning and that they act as a form of memory storage. During memory retrieval these same neurons will become reactivated in such a way that the organism will “remember” the original stimulus. If there were some way to tag these neurons during memory formation and then artificially reactivate them, you could test if these neurons are indeed acting as memory storage.
Two papers came out recently that used the same technique to tag specific neurons associated with a particular memory. One paper by Garner et al published in Science, investigated if an artificially stimulated memory would impair or be incorporated into a naturally stimulated memory. The other paper was rushed out by Nature the next week, probably because they realized the two papers are fairly similar and they didn’t want to be scooped. The Nature paper, by Liu et al., asked if their artificially stimulated memory could induce a behavioral response on its own. In other words, if you stimulate a certain set of hippocampal neurons associated with a memory, will the mouse act as if it’s really remembering the original stimulus?
Both papers used mice, which learned to associate a particular cage with a negative stimulus. When mice are put into a cage, they usually walk around and aren’t too afraid. However, if the base of this cage is electrified and the researchers give short electric shocks to the mice, they start to form a memory that this cage is a mini torture chamber. (I use the term “torture” facetiously; I am sure this research was approved by the universities to make sure no animals were ever in serious pain.) If they put the mice back into a normal cage, the mice recover and go on about their business. The next day, the mice are put back into the torture cage and they recognize how it’s decorated and remember that they’ll get shocked. The mice are afraid and freeze in anticipation of being shocked. This is called “fear conditioning” and is a way of testing if mice have learned this association and can remember it days later. The amount of freezing is measured as the behavioral output of memory formation and recall.
Memory formations of different cage decorations activate different sets of neurons in the hippocampus. The researchers wanted some way of tagging these particular neurons so they could be activated at a later date using an artificial stimulus. Okay, what follows is a complicated molecular biology scheme, but I think it’s really ingenious.
Artificial activation of hippocampal neurons
I will describe how Liu et al. were able to artificially activate particular hippocampal neurons, but the strategy used by Garner et al. is very similar. When mice are placed into a new cage, they notice the color of the walls, the shape of the cage, etc. This activates a certain subset of neurons in the hippocampus. When neurons are highly activated, they start expressing a gene that makes a protein called c-Fos. This protein is what we call a transcription factor, which can activate expression of other genes. Normally, it would activate genes associated with making the neuron more easily activatable for the future. This is what we call “cellular memory”. The neurons “remember” being activated once and will be more likely to fire in the future.
The mice used in these experiments were also expressing some genes that were made by the researchers and put into the mice. During memory formation, C-Fos binds all its normal target genes, but it also binds a construct made by the researchers; it activates expression of the tTA gene, which makes another transcription factor. tTA activates expression of another gene called Channelrhodopsin (ChR2). This gene makes an ion channel that opens in the presence of blue light. When it opens, it lets positive sodium ions into the neuron, which activates the cell. Feeding the mice a chemical called doxycycline can inhibit the action of tTA.
The left vertical pathway is what happens normally during memory formation. The horizontal pathway is what happened, in addition, in the transgenic mice used by Liu et al.
That’s complicated, but here’s the important part:
1) Neurons that are activated when the mouse is in a new cage will make C-Fos and turn on this molecular system.
2) The activated neurons will end up expressing a light-activated channel (ChR2).
3) Researchers can then turn on these specific neurons with light whenever they choose. A neuron that was not originally activated by the new cage will not have the light activated channels.
4) Researchers can thus label active cells and turn them on or off with light. This whole expression system can also be turned off with doxycycline, so neurons can be labeled only during the experimental time period.
In the other paper, instead of expressing ChR2, the mice expressed a different receptor that can be activated with another chemical called CNO. Chemical activation isn’t as fast as light activation, so this was more for long-term activation (hours vs. seconds).
Liu et al. wanted to see if they could induce the freezing behavior of mice, just by activating the subset of neurons with light. Here’s the set-up: the mice were put in Cage A and given electrical shocks. If they were put back in Cage A, they would freak out and freeze. If they were instead put in Cage B, which looked different, they were not afraid and did not freeze. Remember that these mice also were expressing ChR2 in the neurons that had been originally activated by Cage A. When the mice were in neutral Cage B, the researchers shined blue light onto their hippocampus (via an implanted fiber optic cable). The blue light activated ChR2, activated the neurons associated with Cage A, and the mice froze! The neurons that were activated by the Cage A context had formed an association with fear in the brain, so when they were activated artificially later, the mice became afraid. Wow! Artificially induced memory recall.
The experiments done by Garner et al. were interesting too, but a bit more complicated. They put the mice in Cage A without an electric shock and labeled these neurons with their receptor. Then they moved these mice to Cage B and gave them an electric shock. So the mice were fear conditioned for Cage B, but their neurons for Cage A could still be artificially activated by the researchers. Next they put the mice in Cage B again and they were scared and froze. Then they artificially activated the Cage A neurons (by giving them the CNO activator) while the mice were in Cage B. This “memory” of the neutral Cage A was able to override the fear memory of Cage B and the mice did not freeze anymore. In other words, the internal retrieval of a neutral memory interfered with the mouse’s fear memory. In some distant future, a similar technique potentially could be used by humans trying to face a fear, but that’s pure speculation.
In another experiment done by Garner et al. (see figure below), they put the mice in neutral Cage A and labeled these neurons. Then they did fear conditioning in Cage B, but this time they also artificially activated the neurons for Cage A at the same time. At this point neurons for Cage B were naturally being activated, while the Cage A neurons (which are different) were also active. One day later, when they tested the memory of the mice in Cage B, the mice no longer froze; they weren’t afraid of Cage B. But if they were in Cage B and the Cage A neurons were activated by the drug CNO at the same time, the mice froze. So what’s going on with this? During fear conditioning, the mice were seeing Cage B but were also having Cage A neurons activated. The fear memory incorporated both Cage A and Cage B neurons into the memory trace in the brain. Wow! They created a synthetic hybrid memory. This implies that if you’re thinking about one thing while a memory of something else is being formed, the two things may merge together into some other non-realistic memory. In the words of the authors “the internal dynamics of the brain at the time of learning contribute to memory encoding”.
I love both of these papers. They used new techniques in molecular biology to really address new questions about how our brain encodes memories. Both demonstrated that a discrete population of neurons in the hippocampus is responsible for forming new spatial memories (about the visual nature of the cages). Liu et al. showed that they could elicit an output of a memory just by reactivating the neurons that had been active during learning. Garner et al. demonstrated that the internal self-generated activity of neurons in the brain can have an impact on memory formation. Kudos to both groups and there’s no need to feel scooped, Authors, because the two papers are complimentary in the experimental questions that were answered.