One of the most influential ideas in considering the function of the hippocampus is that reactivation of hippocampal memory traces drives the cortex back to the state it was in during the encoding, thus allowing for a re-experience of past events. However, until Tanaka et al.’s 2014 study, it appears that there was never direct evidence that this indeed occurs.
Tanaka et al. used a similar Tet-Off system used by Liu, Ramirez et al. to express GFP and ArchT in neurons of dorsal CA1 that were active in encoding. ArchT, by the way, inhibits neurons when activated by light. They fear conditioned mice while labeling the active population in CA1, and later showed that inhibiting these neurons impaired fear memory recall. Interestingly, they demonstrated that in control mice, memory performance was strongly correlated with the degree of overlap between neurons active during encoding, and those active during retrieval (awesome analysis in checking out individual differences).
In one set of experiments, the authors demonstrated that if a trace for a context that is highly similar, yet still different from the fear context, is labeled with ArchT, then inhibiting that trace in CA1 will begin to impair recall. This isn’t the case for a highly dissimilar labeled context.
When it comes to the cortex, inhibiting CA1 reduces the overlap between the encoding and decoding populations in every cortical area studied. Except for the lateral entorhinal cortex, overall levels of c-fos expression at retrieval were the same. In the LEC, the group with inhibition showed less overall activity. Additionally, CA1 inhibition did not reduce activity in the amygdala at retrieval, but did reduce overlap between encoding and retrieval populations in the central nucleus, but not the basal lateral amygdala.
In summary, inhibiting CA1 populations active during encoding at the time of recall reduces the overlap of activity between encoding and retrieving populations in downstream regions. This overlap is necessary for, and correlates with, expression of contextual memories.
The results demonstrated here of CA1 necessity in recall are made all the more interesting when we remember that artificial activation of CA1 is not sufficient for recall. Perhaps breaking the CA1 bridge to the cortex makes it impossible to reinstate that experience, while making a labeled memory trace compete with a naturally occurring memory trace is not enough to influence behavior. As Ramirez, Liu et al. noted, it may be that CA1 communicates with a rate code not imitated by their stimulation protocol and that the natural processing of CA3 is absolutely necessary for CA1 to communicate with the cortex. And as shown in Denny et al., CA3 activity at the time of retrieval must overlap with the activity at encoding for retrieval to work. And what of the dentate gyrus? Some argue that the dentate gyrus is strictly necessary for encoding, while others suggest that dentate granule cells may also play a role in recall. This is a debate that I will go as deep into as possible at some point in the future.