Cellular and network mechanisms for learning and memory consolidation

Abstract

In learning, repeated experiences might be integrated individually as they occur, or they might be combined within dedicated time windows, possibly promoting quality control. Here we show that in Pavlovian, incremental and incidental learning, related information acquired within time windows of 5h (time units for learning) is combined to determine whether and what mice learn. Trials required for learning had to occur within 5h, when learning-related shared partial cues could produce association and interference with learning. Upon acquisition, cFos expression was elevated during 5h throughout specific system-wide neuronal assemblies. Time unit function depended on network activity and cFos expression. Local cFos activity was required for distant assembly recruitment through network activity and distant BDNF. Long-term consolidation of the memories depends on processes occurring many hours after acquisition. Whether this involves plasticity that is specifically required for long-term consolidation remains unclear. We found that learning-induced plasticity of local parvalbumin (PV) basket cells was specifically required for long-term, but not short/intermediate-term, memory consolidation in mice. PV plasticity, which involves changes in PV and GAD67 expression and connectivity onto PV neurons, was regulated by cAMP signaling in PV neurons. At 12–14 h, PV plasticity was required for enhanced sharp-wave ripple densities and c-Fos expression in pyramidal neurons. Our results reveal general network mechanisms of long-term memory consolidation that requires plasticity of PV basket cells induced after acquisition and sustained subsequently through D1/5 receptor signaling. Upon learning, local inhibitory parvalbumin (PV)-expressing Basket cell networks can switch to opposite configurations that either favor or interfere with further learning, but how this opposite plasticity is induced and relates to distinct learning requirements has remained unclear. Here, we show that PV Basket cells consist of hitherto unrecognized subpopulations, with distinct schedules of neurogenesis, input connectivities, output target neurons, and roles in learning. Plasticity of hippocampal early-born PV neurons was recruited in rule consolidation, whereas plasticity of late-born PV neurons was recruited in new information acquisition. This involved regulation of early-born neuron plasticity specifically through excitation, and of late-born neuron plasticity specifically through inhibition. Therefore, opposite learning requirements are implemented by distinct local networks involving PV Basket cell subpopulations specifically regulated through inhibition or excitation.