Transient excitability and synaptic consolidation support stable memory despite neural drift

Abstract

Long-term memories remain behaviorally stable despite turnover in the neuronal populations that encode them. This representational drift differs across brain regions: hippocampal representations can reconfigure within hours, whereas cortical ensembles remain comparatively stable over days to weeks. This asymmetry poses a challenge for systems consolidation, in which hippocampal activity is thought to instruct the formation of persistent cortical memory traces. We developed a two-region recurrent network model of the hippocampus (HPC) and anterior cingulate cortex (ACC) incorporating region-specific synaptic plasticity, excitability-dependent neuronal allocation, hippocampal-to-cortical coupling, and activity-dependent intrinsic plasticity. When the regions evolved independently, faster synaptic turnover in HPC produced pronounced drift, whereas persistent ACC connectivity preserved a more stable cortical ensemble. In the intact circuit, hippocampal input recruited ACC neurons but also propagated hippocampal variability into cortex, destabilizing the emerging cortical engram and impairing memory expression. A transient increase in the intrinsic excitability of recruited ACC neurons counteracted this instability by promoting repeated reactivation during an early consolidation window, stabilizing cortical ensemble membership without preventing hippocampal drift. Simulated erasure of learning-induced potentiation further reproduced the early dependence of memory on hippocampal, but not ACC, plasticity. Together, these results support a sequential mechanism in which hippocampal recruitment, transient cortical intrinsic plasticity, and persistent cortical synaptic plasticity transform a dynamic hippocampal representation into a stable cortical memory trace.