Abstract :
[en] Our brains enable us to perform complex actions and respond quickly to the external world, thanks to transitions between different brain states that reflect the activity of interconnected neuronal populations. An intriguing example is the ever-present switch of brain activity that occurs while transitioning between periods of active and quiet waking. It involves transitions from small-amplitude, high-frequency brain oscillations to large-amplitude, low-frequency oscillations, accompanied by neuronal activity switches from tonic firing to bursting. The switch between these firing modes is regulated by neuromodulators and the inherent properties of neurons. Simultaneously, our brains have the ability to learn and form memories through persistent changes in the strength of the connections between neurons. This process is known as synaptic plasticity, where neurons strengthen or weaken connections based on their respective firing activity.
While it is commonly believed that putting in more effort and time leads to better performance when memorizing new information, this thesis explores the hypothesis that taking occasional breaks and allowing the brain to rest during quiet waking periods may actually be beneficial. Using a computational approach, the thesis investigates the relationship between the transitions in brain states from active to quiet waking described by the neuronal switches from tonic firing to bursting, and synaptic plasticity on memory consolidation.
To investigate this research question, we constructed neurons and circuits with the ability to switch between tonic firing and bursting using a conductance-based approach. In our first contribution, we focused on identifying the key neuronal property that enables robust switches, even in the presence of neuron and circuit heterogeneity. Through computational experiments and phase plane analysis, we demonstrated the significance of a distinct timescale separation between sodium and T-type calcium channel activation by comparing various models from the existing literature.
Synaptic plasticity is studied to understand learning and memory consolidation. The second contribution involves a taxonomy of synaptic plasticity rules, investigating their compatibility with switches in neuronal activity, small neuronal variabilities, and neuromodulators.
The third contribution reveals the evolution of synaptic weights during the transition from tonic firing in active waking to bursting in quiet waking. Combining bursting neurons with traditional synaptic plasticity rules using soft-bounds leads to a homeostatic reset, where synaptic weights converge to a fixed point regardless of the weights acquired during tonic firing. Strong weights depress, while weak weights potentiate until reaching a set point. This homeostatic mechanism is robust to neuron and circuit heterogeneity and the choice of synaptic plasticity rules. The reset is further exploited by neuromodulator-induced changes in synaptic rules, potentially supporting the Synaptic-Tagging and Capture hypothesis, where strong weights are tagged and converge to a high reset value during bursting.
While burst-induced reset may cause forgetting of previous learning, it also restores synaptic weights and facilitates the formation of new memories. To exploit this homeostatic property, an innovative burst-dependent structural plasticity rule is developed to encode previous learning through long-lasting morphological changes. The proposed mechanism explains late-stage of Long-Term Potentiation, complementing traditional synaptic plasticity rules governing early-stage of Long-Term Potentiation. Switches to bursting enable neurons to consolidate synapses by creating new proteins and promoting synapse growth, while simultaneously restoring efficacy of postsynaptic receptors for new learning. The novel plasticity rule is validated by comparing it with traditional synaptic rules in various memory tasks. The results demonstrate that switches from tonic firing to bursting and the novel structural plasticity enhance learning and memory consolidation.
In conclusion, this thesis utilizes computational models of biophysical neurons to provide evidence that the switches from tonic firing to bursting, reflecting the shift from active to quiet waking, play a crucial role in enhancing memory consolidation through structural plasticity. In essence, this thesis offers computational support for the significance of taking breaks and allowing our brains to rest in order to solidify our memories.
These findings serve as motivation for collaborative experiments between computational and experimental neuroscience, fostering a deeper understanding of the biological mechanisms underlying brain-state-dependent memory consolidation. Furthermore, these insights have the potential to inspire advancements in machine learning algorithms by incorporating principles of neuronal activity switches.
Commentary :
When memorizing new information, it is commonly assumed that the more effort and time we invest, the better our performance will be. However, it turns out that taking occasional breaks associated with a restful brain may be precisely what we require.
In my thesis, we investigate this hypothesis using a computational approach. We developed a biophysical neural network capable of simulating various brain states, including an active learning state and a quiet waking state. This network can modify its connections to encode new memories. Our results suggest that during periods of quiet waking, neurons collectively engage in bursting activity, which facilitates the consolidation of memories.
