Dynamic Control of Network Level Information Processing through Cholinergic Modulation

Abstract

Acetylcholine (ACh) release is a prominent neurochemical marker of arousal state within the brain. Changes in ACh are associated with changes in neural activity and information processing, though its exact role and the mechanisms through which it acts are unknown. Here I show that the dynamic changes in ACh levels that are associated with arousal state control informational processing functions of networks through its effects on the degree of Spike-Frequency Adaptation (SFA), an activity dependent decrease in excitability, synchronizability, and neuronal resonance displayed by single cells. Using numerical modeling I develop mechanistic explanations for how control of these properties shift network activity from a stable high frequency spiking pattern to a traveling wave of activity. This transition mimics the change in brain dynamics seen between high ACh states, such as waking and Rapid Eye Movement (REM) sleep, and low ACh states such as Non-REM (NREM) sleep. A corresponding, and related, transition in network level memory recall is also occurs as ACh modulates neuronal SFA. When ACh is at its highest levels (waking) all memories are stably recalled, as ACh is decreased (REM) in the model weakly encoded memories destabilize while strong memories remain stable. In levels of ACh that match Slow Wave Sleep (SWS), no encoded memories are stably recalled. This results from a competition between SFA and excitatory input strength and provides a mechanism for neural networks to control the representation of underlying synaptic information. Finally I show that during the low ACh conditions, oscillatory conditions allow for external inputs to be properly stored in and recalled from synaptic weights. Taken together this work demonstrates that dynamic neuromodulation is critical for the regulation of information processing tasks in neural networks. These results suggest that ACh is capable of switching networks between two distinct information processing modes. Rate coding of information is facilitated during high ACh conditions and phase coding of information is facilitated during low ACh conditions. Finally I propose that ACh levels control whether a network is in one of three functional states: (High ACh; Active waking) optimized for encoding of new information or the stable representation of relevant memories, (Mid ACh; resting state or REM) optimized for encoding connections between currently stored memories or searching the catalog of stored memories, and (Low ACh; NREM) optimized for renormalization of synaptic strength and memory consolidation. This work provides a mechanistic insight into the role of dynamic changes in ACh levels for the encoding, consolidation, and maintenance of memories within the brain.