OK, so if I’ve learned one thing from the papers reviewed so far, it’s that the population of neurons active during the encoding of a new memory must be reactivated for that memory to be successfully remembered. But, does synaptic plasticity need to occur within the engram circuit for recall?
In the first experiment reported here, the authors study how engram cells in the dentate gyrus are potentiated as a result of their involvement in a memory trace. Further, they study how anisomycin, which prevents LTP and results in amnesia, impacts their responsive properties. First they express ChR2 in perforant path fibers (which project to the dentate gyrus) active during encoding, and the report mCherry in dentate granule cells also active in encoding. By recording from mCherry expressing (engram) cells or mCherry negative (non-engram) cells during opto stim of the perforant path, they determined that engram synapses have a higher AMPA:NMDA ratio as compared to non-engram synapses. This is indicative of LTP, as the process of potentiation results in the insertion of AMPA receptors into the postsynaptic density. This increase in abolished by anisomycin administration (amnesic mice). Further, dentate mCherry+ cells have a higher density of dendritic spine density than do the mCherry- cells. This difference is also abolished by anisomycin. Though it isn’t characterized in the paper, and it’s difficult to tell from a single figure, it appears from the confocal images in the paper that the spines on engram dendrites (without anisomycin) are large, mature mushroom spines, whereas the spines in other groups are much smaller. Amnesic engram cells displayed similar responses to input as did non-engram cells. Without anisomycin, engram cells have a larger EPSC than non-engram cells. (It appears that EPSC’s in the amnesic mice are much larger. No idea what to make of that.) Lastly, the connectivity between DG and CA3 cells was examined by potentiating DG and checking mCherry+ and mCherry- CA3 cells. Engram cells in CA3 were much more likely than non-engram cells to respond to engram cell stimulation in DG. Thus, engram circuit connectivity is preserved when potentiation is blocked.
Next, the authors examined how connectivity and potentiation might be dissociable in relation to memory processes. Animals were fear conditioned with the engram for this context tagged with ChR2. Stimulating these cells in control animals resulted in expression of a fear memory in a novel environment. Likewise, stimulating these cells in animals administered anisomycin also resulted in fear memory expression. During re-exposure to the fearful environment without opto stim, only control animals would express the fearful memory. This demonstrates that memories apparently lost in amnesic mice can be rescued by artificial stimulation of engram cells.
Probing this phenomenon further, the authors found that they could replicate this using a place avoidance task, where the animal uses a more active means of memory expression, rather than just freezing in a given context. Reactivating a trace involved in a tone-fear association (rather than context-fear) also resulted in freezing in amnesics. And, they were able to replicate these findings by stimulating CA1 instead of DG. (It was found previously that a 20Hz stimulus delivered to CA1 engram cells was insufficient for recall. Here, however, the authors reduced this to 4Hz. Perhaps more similar to population oscillations in CA1?)
The authors tested how engram reactivation may rescue disrupted reconsolidation, by administering anisomycin after mice exhibited natural fear memory recall. Re-exposure to the fearful context resulted in impaired fear memory expression in anisomycin-treated animals, however, optogenetic stimulation of the fear engram could still function as a fear-inducing stimulus.
Lastly, the authors tested how engram cells downstream from the DG respond to consolidation interference. Though natural recall resulted in reduced activity (c-fos) in amnesic mice, optogenetic stimulation of DG engram cells resulted in comparable activity between controls and amnesics in both the amygdala and CA3.
All in all, it appears that when potentiation between the cells in an engram circuit fail, the memory is still stored in the connectivity of the circuit. Thus, connectivity is necessary for storage, while potentiation is necessary for natural recall.
I keep thinking about this in terms of the chicken and the egg. Are there preexisting circuits that are optimal for the storage of particular memories which then become potentiated when sufficiently active? If this were the case, we could expect to see that better learners (or rememberers) have a more efficient baseline connectivity where new memories can find their paths easily and undergo consolidation. Further, memories that can’t quite find a good fit in the circuitry are more likely forgotten than those that seem to naturally fit. How might this be impacted by previous learning?
Or do memories forge their connective paths, requiring processes other than protein synthesis to clear the way for a new memory? This might be better addressed after reviewing Kandel and the phenomenon of short-term potentiation.
I think it’s likely to be both. I also think it’s very likely that I’ll revisit this paper and again.
Either way, it’s interesting to consider how memory disorders may independently impact the connectivity versus potentiation of engram circuits.