Denny et al. used a different method of memory trace labeling as do the Garner et al. and Liu, Ramirez et al. studies. However, it’s the same principle. Here, instead of removing Dox from the diet to label memory traces, the authors inject ArcCreERT2 mice with an estrogen receptor antagonist (in this case, it was tamoxifen) which in turns induces DNA activation. This only happens, however, in those cells that are actively producing the protein in the promoter region of the transgene: Arc. Arc is an immediate early gene expressed in cells that have recently been highly active. Thus, the authors targeted these recently active cells using this system and were able to express in them a protein of interest, e.g. eYFP. After tagging hippocampal neurons involved in memory encoding with eYFP, the authors then performed immunohistochemistry to tag IEG’s with a second fluorophore after memory recall to be able to distinguish those neurons active during memory encoding to those active during memory retrieval. So now you can map the encoding population to the retrieval population, which is just beautiful (Figure 1).
Five days after conditioning animals to fear a context A, animals were either re-exposed to context A, where animals showed a heightened fear response, or exposed to a novel context B, where animals displayed less fear. As was demonstrated by Garner et al., this method of memory trace labeling resulted in largely overlapping activation during during re-exposure to a specific context, while different contexts activated largely separate hippocampal populations in both the DG and area CA3. However, when repeating this experiment with a longer delay (30 days) after encoding, animals were equally afraid of both contexts and showed more overlapping activation during exposure to different contexts.
In the next set of experiments, instead of eYFP, the ArcCre system was used to drive the expression archaerhodopsin (Arch), which can be used to inhibit neurons via light stimulation, in encoding cells. Mice were fear-conditioned to context A after a tamoxifen injection and two weeks later, were re-exposed to A. During optogenetic inhibition of the encoding population, mice froze significantly less in context A than controls. No difference of inhibition was seen while in a novel context B, nor when the encoding population corresponded to a third, non-fearful context C.
Lastly, the effect of reducing hippocampal neurogenesis on memory and memory trace structure was examined. Using a one-shock CFC protocol, mice with reduced neurogenesis showed impaired contextual fear memory. Meanwhile, both the reduced neurogenesis and control groups exhibited high overlap between the encoding and recall populations of the DG for this same context and yet, very little overlap in CA3. Using a 3-chock CFC protocol restored overlapping encoding/decoding CA3 activity as well as contextual fear recall.
While Garner et al. and Liu et al. demonstrated that activating hippocampal populations active during contextual fear encoding is sufficient to elicit fear memory recall, Denny et al. show here that these populations are also necessary for fear memory recall. Interestingly, ablation of neurogenesis impairs contextual fear memory and CA3 memory trace formation on a one-shock, but not three-chock, CFC protocol. This indicates that without neurogenesis, the hippocampus is a slower learning system that requires more time to sculpt the appropriate representation in CA3. Over time, the less plastic DG population of older, mature granule cells can catch up to systems which have neurogenesis. However, would this result hold for an organism with rich daily experience, such as social interaction among conspecifics, food foraging and caching, or escape behavior.
You may notice in this paper that in the DG, Arc activity during encoding is much, much higher than activity during recall. This may be in part due to degradation of the trace over time. Cazzulino et al. show that perhaps, tamoxifen injection isn’t the best way of labeling encoding populations after all. When comparing tamoxifen to 4-OH (another estrogen receptor ligand), the ArcCreERT2 had a much more specific label of the encoding population as 4-OH has a shorter half-life and produces less background signal.
So, after reading, Liu, Garner, and now Denny, I’m now aggressively pitching ArcCre experiments to my advisor. One thing I keep fantasizing about, as mentioned by Tonegawa et al. is to label a memory trace and then use CLARITY to render the tissue translucent. Unfortunately, I don’t have the scope or time to carry out such a project, so I’ll just have to wait and see.