Abstract
Temporal associations between sensory stimuli separated in time rely on the interaction between the hippocampus and medial prefrontal cortex (mPFC). However, it is not known how changes in their neural activity support the emergence of temporal association learning. Here, we use simultaneous electrophysiological recordings in the hippocampal CA1 region and mPFC of mice to elucidate the neural dynamics underlying memory formation in an auditory trace conditioning task. We found that in both areas conditioned (CS+/CS−) and unconditioned stimuli (US) evoked similar temporal sequences of neural responses that progressively diverged during learning. Additionally, persistent CS representations emerged in mPFC after learning, supported by CS+ coding states whose transient reactivation reliably predicted lick onset and behavioral performance on single trials. These results show that coordination of temporal sequences in CA1 and persistent activity in mPFC may underlie temporal association learning, and that transient reactivations of engrams in mPFC predict the animal behavior.
Highlights
Temporal representation of CS+, but not CS− and US, strengthens in CA1 after learning.
Similarity between stimulus and reward temporal representations decrease with learning, but is recovered in error trials.
Representation of stimulus identity is strong and stable in PFC since stimulus onset, while it only emerges in CA1 during trace period.
Neural states defined on faster time-scales reveal the emergence of CS coding states in PFC whose onset predicts lick times and task performance.
PFC-CA1 states do not increase coordination during late CS+ stimulus and trace.
Competing Interest Statement
The authors have declared no competing interest.
Footnotes
Beyond a more concise title, the results previously depicted in Figures 6 to 8 and their associated supplementary figures are now revised to account for a new criterion of state activation in the HMM analyses. Briefly, state activation is now defined as the most likely state instead of using a threshold, to account for the lower probabilities in CA1 states. With this correction we were able to capture CS coding states not only in PFC but also in CA1, allowing us to extend some of our previous findings.