Plasticity degeneracy underlies flexible formation and reconfiguration of spatial representations in hippocampal granule cells
Learning and memory require neural representations that are simultaneously robust and adaptable. How do neural systems reconcile these opposing demands of stability and flexibility? Here, we use hippocampal granule cells, which transform spatially diffuse entorhinal inputs into sparse sharply-tuned spatial representations, to address this fundamental question using emergence, stabilization, remapping, and suppression of place-fields as ideal systems to study stability-flexibility balance. Employing an unbiased population-based approach involving different strengths of plasticity in several components, we show that sharply tuned place-cell firing does not arise from a unique plasticity mechanism. We employed a population of heterogeneous, experimentally constrained biophysical models of dentate gyrus granule cells receiving cortical grid-field and contextual inputs to demonstrate that spatially restricted synaptic potentiation is insufficient to selectively route spatial information. Instead, each of the four distinct target transitions, namely reliable emergence, stabilization, remapping, and suppression of place-fields, were achievable by coordinated plasticity spanning excitatory synapses and several intrinsic ion channels. Across each of these four transitions, we identify numerous mechanistically distinct plasticity combinations that generate equivalent functional outcomes. These solutions are non-random and highly diverse, exhibiting weak dependence among individual plasticity components despite converging onto constrained functional states, through a combination of targeted amplification and global suppression of spatial firing. Our findings reveal extensive plasticity degeneracy in hippocampal spatial coding and suggest that flexible neural representations emerge not from unique plasticity rules involving one single component, but from a repertoire of alternative plasticity routes that are capable of implementing the same computation.