Stability and Adaptability in Balance: A Dual Mechanism for Metaplasticity in Cortical Networks

Metaplasticity dynamically adjusts how synaptic efficacy and connectivity change, helping neural circuits adapt to experience. However, the interaction between changes in synaptic weight (W) and connection probability (P) remains poorly understood. We explored their interaction using a biologically-inspired, multi-layer spiking neural network. We found that while W controls network excitability, P exerts layer-specific and time-dependent control, crucial for network stability. Simultaneous changes in W and P, i.e. metaplasticity, revealed complex, non-additive interactions, shaping response timing and neural recruitment, resulting in the emergence of functionally distinct neuronal subtypes: input-invariant neurons maintaining responsiveness and variant neurons enabling adaptation, based on differential E-I dynamics. This interaction allows the network to achieve functional homeostasis in the input layer while preserving flexibility in superficial layers. We provide a novel framework for understanding how metaplasticity balances the competing demands of stability and adaptability in cortical circuits, with significant implications for learning, memory, and neural coding.