Skip to main content

Brain computer interface to study and manipulate mamories of aversive experience during sleep

Periodic Reporting for period 3 - MNEMOSYNE (Brain computer interface to study and manipulate mamories of aversive experience during sleep)

Reporting period: 2020-10-01 to 2022-03-31

We all have places we like to go back to, either because the food there is particularly good, or because we have memories of wonderful evenings with friends. In contrast, on our way home, there are streets we tend to avoid, sometimes without remembering exactly why. The Mnemosyne project aims to understand the relationship between internal representation of space – the neuronal code of where we are in the environment – and the emotional valence associated with location in the environment. Building upon our initial demonstration that appetitive place association can be artificially elicited during sleep (by taking advantage that neurons coding for space are “reactivated” during sleep), we have planned to now demonstrate that aversive associations can also be artificially created during sleep and, unlike our initial manipulations based on the activity of single neurons, to extend the read-out of the brain activity during sleep at the population level. To this end, the goal is to use advanced signal processing and machine learning tools that can decode neuronal activity in real time.
We want to reverse an aversive experience during wakefulness by a rewarding conditioning during sleep by using our brain-computer interface. This would pave the way to treat post-traumatic brain disorders during sleep by using spontaneous reactivation and avoiding the exposure to the stressful situation that is classically used in exposure therapy. We postulate that sleep reactivations represent the best recall-like strategy to observe the reactivation of the exact same memory trace that led to the development of PTSD.
During the first phase of our project, we have made major progress towards the achievement of our overarching goals. First, we have designed a new behavioral task that allows us to disentangle animal’s location, aversive and appetitive behavior so that neuronal activity can be reliable associated with each behavioral feature. One challenge is to design an environment simple enough that animal’s location can be decoded from neuronal population activity (e.g. open environment requires extremely large number of neurons for decoding), as well as offering the animal a ‘safe’ zone that can be easily identified during behavioral analysis. We have found that a U-shape maze enables exactly this, and comprehensive testing of this environment under various conditions has proved the robustness of this approach. We also encountered several challenges that we managed, for most of them, to mitigate. Second, we have put in place a neuronal population analysis technique based on deep-learning techniques that enable real time neuronal activity detection processing and decoding. This is a major achievement of our project and now allows us to routinely perform the tasks of our work-packages. Last, we have shown that physiological parameters such as breathing rate and hearth beat are not only reliable markers of emotional states (as long known) but also play an active role in the expression of emotions. We can therefore extend our initial goals of linking internal (i.e. neuronal) representations with emotion to the whole-body physiological features.
We also made an unexpected observation during fear learning in the UMaze. We observed freezing in the shock zone (that was expected) but we also observed that the animal froze in the opposite extremity of the UMaze (i.e. a safe zone). That was not expected. But we observed that those freezing periods were full of ripples and correspond to reactivation of the shock but in the hippocampus and the prefrontal. Moreover, we showed that the two freezing n the shock zone and the safe zone were not identical but associated with different breathing and heart rate. These two types of freezing are reminiscent of what is considered as panic versus anxiety whose characterization is still lacking in rodents.
In conclusion, this project is on right tracks to achieve the desired aims.
Several of the contributions states above go way beyond state-of-the-art. Although the focus of several projects by different labs worldwide, teal-time decoding of neuronal activity in vivo remains extremely challenging and no reliable solutions have been printed yet. Although additional work and refinement are still necessary, our current solution seems extremely promising. Second, we have demonstrated that aversive conditioning can be artificially created with our close-loop device, which has not been previously reported. Further experiments are still needed to confirm this result.
By the end of the project, we will answer to the following questions:
- Is it possible to create an artificial memory during sleep with aversive instead of rewarding electrical stimulations?
- Is there any difference between awake and sleep replays for aversive and appetitive learning?
- Is the body reaction similar for real experience and for the replays?
- Is body reaction an index of the emotional reaction or an actor?
- Is it possible to reverse an aversive memory acquired during wakefulness by a positive memory created during sleep? It would be a proof of concept for a potential treatment of Post-traumatic stress disorder.
Summary of the key experiment of the project