To be fair, one of these papers reviews the other as well as the Liu, Ramirez, et al. paper I’ve already reviewed. But A) I read both, and B) I need to make up ground in advance for the 6 months I’ll need to re-read, understand, and review Marr’s 1971 paper.
I highly recommend playing the theme music to Inception while reading this paper and/or blog post.
Anyways, Ramirez, Liu et al. expanded upon their 2012 study by using optogentic/engram technology to create an artificial associative cue. Again, using the Tet-Off system, they labeled the dentate gyrus population that encoded context A with ChR2. They then activated this population while fear conditioning in context B. So now, trace B consists of the cells activated by sensory info B + activated trace A. Exposing these animals to a new context C didn’t scare them, as they didn’t need to be. However, in the case of context A, even though they were never shocked here, the linking of this population with a fear memory cause animals to freeze much more than controls. The authors have transformed context A in a fear memory retrieval cue. Interestingly, reactivating the CA1 population that encoded A during conditioning the memory had no effect on freezing in A.
Imagine working in a restaurant and a customer asks for a napkin. You get them a napkin. Now imagine a customer asks for a napkin, but at the same time, a customer at another table asks for a menu. The simultaneous requests may interfere with each other and instead of getting right to their request, you may need a second longer to process both (Or you do that thing where you step back and look at both customers to see which one asks again first, and they look at each and both say, “Go ahead” at the same time…awkward…) That’s what happened in the next set of experiments.
One group of mice had their A traces incorporated into trace B via opto stim, while the other group did not. Then they were reintroduced to B. Animals who had the stimulation during conditioning froze more when trace A was again stimulated, as has been shown. But animals without the stim became “confused” while trace A was stimulated in context B as this population was not involved naturally, nor artificially in the B trace. It’s like presenting them both with the fearful and the safe context at once. Then, using the same groups, the authors did the experiment in a novel context D. Animals whose B trace was independent of the A trace did not freeze much at all while in D, regardless of whether or not opto stim was occurring. However, mice whose A trace was activated while in D did freeze more.
Full recall of a fear memory that has artificially included trace A in it requires trace A to be activated in order to reach basal levels of memory expression. Following up on this result, the authors looked at c-fos activation in the amygdala. As one might predict, activating trace A results in as much amygdala expression as context B re-exposure without opto stim.
I don’t know if some would see the amygdala c-fos expression follow-up experiment as obvious, but to me, it’s what makes these papers so awesome. For someone who is early on in their grad career, these are the kind of experiments that I likely wouldn’t think to include in a project pitch to my advisor, but that I need to develop the skill for. It’s very much like a crash course in logic and experimental design.
Lastly, the authors put mice into an arena with two chambers blocked off from each other. They labeled the trace for one chamber with ChR2 and then activated this trace during fear conditioning. Upon re-exposure to the apparatus with the choice to be in either chamber, the mice spent much more time in the unlabeled chamber, the trace for which was not activated during conditioning. Again though, doing this same experiment, but labeling cells in CA1 instead of the DG produced no apparent behavior difference over controls.
So essentially, artificially activating a population of DG neurons during CFC can turn that population into a CS to elicit a fear response. This is not true for CA1 where the authors found that though the traces for different contexts were largely separate, there was significantly more overlap than in the DG. In reviewing this experiment, the authors suggest that perhaps CA1 has a temporal code for recall rather than a population code. It would be interesting (though I don’t think practically possible) to test this by using an optetrode to read a rate code for a particular context and then mimic it with opto stim. However, I don’t know how many units the optetrode records from and rather this is just a field recording or whether it can resolve individual units. Further, which unit do you choose? Do you average? Does anyone want to try it and let me know how it goes? Can you cite blogs?
What just occurred to me while reading this paper, is that while Yiu et al. showed that neurons that are relatively excitable get incorporated into the memory trace, optogenetically activating trace A during encoding of B seemed to make trace B a combination of the natural trace B and activated trace A. I would’ve have thought that exciting the cells of A during exposure to B, would mean they win the trace B sweepstakes and the memories for both contexts get mashed into one confusing memory trace that would either be competitive between the two contexts or would contain elements of both contexts (resulting in impaired natural recall, which did happen) without the recruitment of a new population of neurons. Perhaps neurons that have recently encoded have a set of encoding brakes that get put on until their content is consolidated. Maybe these brakes were switched off during opto activation. Arc perhaps? OR the sensory info for different contexts activates such a different hippocampal ensemble, that there is no competition between encoding cells and those who have recently encoded. Thank you, dentate gyrus.