ИСТИНА

The advent of neuromorphic technologies brings new tasks for the AIoriented neuroscience research. Our work explores neurobiological principles of memory encoding that can vastly surpass capabilities of existing neuromorphic hardware. In the first place, neurobiological memory is very selective in respect to number of neurons involved. Here we studied this selectivity in a model of associative memory formation in mice. Typically, associative conditioning uses reinforcement delivered immediately after a new cue. However, this does not allow to separate neuronal mechanism for cue memory encoding from processes of its association with reinforcement. In the present work we used the context preexposure facilitation effect paradigm (Fanselow, 1990; Rudy and O'Reilly 2001) to study allocation of associative memory into the neurons of previously established contextual memory. Mice were first allowed to explore a new context A for 5 min and following a 3-day delay received a 2 sec footshock immediately after being replaced in this context. Resulting contextual fear persisted for at least 30 days after this engram conditioning and was specific for the context A, it did not develop if immediate shock was delivered in a context B and there was no freezing in a novel context C. We next studied how this association is allocated in the mouse brain neuronal circuits. For this we used activity-inducible transcription factor c- Fos as a cellular marker of memory allocation during conditioning. Furthermore, we combined this c-Fos neuroimaging with c-Fos TRAP (Targeted Recombination in Active Populations; Guenthner et al., 2013) technique to reveal overlap in cellular populations involved in the initial acquisition of contextual engram and its later associative conditioning. Our data suggest that only a limited population of neurons in the mouse associative neocortex, mainly pyramidal neurons, are involved in the contextual engram. Furthermore, it is in this selective ensemble of neurons the subsequent associative memory for negative reinforcement is allocated. We discuss this data in respect to principles of neuromorphic associative memory encoding. Supported by RSCF 14-15-00685, RFBR 17-00-00215.