Skip to Main content Skip to Navigation

Hippocampal reactivations after aversive or rewarding experience : classical and deep learning approaches

Abstract : Hippocampus is required for encoding and consolidation of declarative memories in humans. In animals, hippocampus-dependent memories are often studied in the paradigm of spatial navigation.Place cells are hippocampal neurons that exhibit location-specific firing. In calm states, such as sleep and low-speed wakefulness, place cells are reinstating patterns that were observed during active exploration - a phenomenon termed 'reactivations'. There is ample evidence that hippocampal reactivations are neural correlate of spatial memory consolidation. However, it is still not clear whether hippocampus codes for purely spatial information, or it also encodes non-spatial parameters such as emotional valence.It is known from the literature that rewarded locations are reactivated more often in sleep following active behavior than the others. There are two competing hypotheses that could explain this fact. According to one of them, locations that are associated with certain motivational and emotional values are reactivated more. Other hypothesis postulates that increase in reactivation rate of rewarded environment is caused simply by increased time spent in the reward location.To tackle this question, we performed two parallel experiments employing aversive and rewarding learning. In one of them, we used aversive intracranial stimulation of periaqueductal gray matter to create aversive spatial association, and in the second one, we used rewarding stimulation of medial forebrain bundle to create the appetitive spatial association.Behaviorally, emotionally important locations had massive difference in terms of occupancy after learning. However, in sleep following learning sessions we have found reactivations both for aversive and for rewarding experience, which confirms the hypothesis that motivational and emotional variables of the task affect hippocampal coding. We have demonstrated that neuronal patterns active during aversive learning are reactivated during sleep following the task stronger than the activity recorded during free exploration. Given that cognitive maps in our task were predominantly stable, we concluded that hippocampal activity during the task and in reactivations registered after the task does not code for pure spatial information but also for motivational salience associated with space. Supporting this claim, we have found that the strength of sleep reactivations correlated with the magnitude of avoidance behavior. In addition, we have shown that after aversive learning, despite significantly decreasing time spent in the aversive zone, reactivations mostly consisted of representations of the aversive zone and adjacent locations, possibly reflecting avoidance behavior.In addition, we have attempted to reverse aversive spatial association by using rewarding intracranial stimulation in the sleep that follows aversive learning. To achieve this goal, we have developed brain-computer interface that is based on the stack of convolutional and recurrent artificial neural networks. In other words, we have designed the position decoder, that will be able to decode which position is reactivated from hippocampal activity online. This decoder does not require spike sorting as most published decoding methods, and more importantly has a confidence measure that allows the user to filter out spuriously decoded positions.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Thursday, December 2, 2021 - 2:56:07 PM
Last modification on : Friday, December 3, 2021 - 3:47:58 AM
Long-term archiving on: : Thursday, March 3, 2022 - 7:39:12 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03463521, version 1



Dmitri Bryzgalov. Hippocampal reactivations after aversive or rewarding experience : classical and deep learning approaches. Neuroscience. Université Paris sciences et lettres, 2021. English. ⟨NNT : 2021UPSLS065⟩. ⟨tel-03463521⟩



Record views


Files downloads