| Abstract |
It is generally accepted that the brain stores memories in distributed neuronal representations. The long-term storage of memories is believed to take place through the strengthening and weakening of synaptic connections between neurons. Recent research has begun to probe the mechanisms that underlie these changes in synaptic connections and to identify the correlates of specific memories in the brain, which are known as memory engrams or traces. These studies have identified a number of different factors, which determine memory storage, from molecular processes to network electrophysiological phenomena. Despite this progress in identifying pieces of the puzzle of the mechanism of memory, we are still lacking a unified framework, which can explain the experimental findings with regards to memory, and can predict the structure of memory traces.
In this thesis we present a novel modeling approach to memory acquisition which combines multiple experimentally observed phenomena. We apply this model in specific well-studied memory-related experimental protocols and we use it to predict the structure of the resulting memory traces in multiple spatial scales, from the synaptic to the neuronal network level. In order to create a unifying framework for studying memory formation we incorporate mechanisms related to the consolidation of long-term memories in the brain. These include the mechanisms for plasticity-related protein capture according to the synaptic tagging and capture model, the localized modulation of excitability, as well as the effects of homeostasis and inhibition. Importantly, we study the effect of dendritic compartmentalization, which is known to affect memory, by incorporating dendritic phenomena in the description of neurons.
Using this model, we show that the sub-cellular structure of memories, which consists of the distribution of synapses to specific neurons and specific dendritic branches is dependent on the ability of neurons to synthesize plasticity proteins in dendrites, and on the activation history of the neuron, which affects its excitability. In addition, we find that synapses tend to potentiate in groups, a phenomenon known as synapse clustering. We extend the model to the case of multiple memories in order to examine their possible interactions and find that the neuronal populations and the synapses that represent time-related memories are intertwined. Using the model, we predict the overlapping components of different memories as a function of time. Finally, we examine the role of dendritic spine reorganization, which occurs constantly in the brain, in the storage of memories.
|
Dendritic contributions to neuronal and synaptic memory allocation
