| Activity-dependent modifications of synaptic strengths are thought to underlie memory formation and learning. However, the precise cellular mechanisms involved in such modifications are still unknown.;We investigate the properties of a biophysical model of synaptic plasticity. The Calcium-Dependent Plasticity model (CaDP) uses the postsynaptic calcium mediated by N-methyl-D-aspartate receptors (NMDARs) as the associative signal for Hebbian learning. At the synapse, the information on the local, presynaptic activity is provided by the amount of neurotransmitter bound to NMDARs, while the global, postsynatic activity is signaled by the back-propagating action potentials. Analysis show that the model can explain the rate-, voltage-, and spike timing-dependent plasticity as consequences of, respectively, the temporal integration of calcium transients, the voltage-dependence of NMDAR conductances, and the coincidence-detection property of these receptors. We propose that different experimental procedures evoke similar intracellular responses that lead to long-term potentiation (LTP) or depression (LTD).;Learning stabilization is provided by a mechanism of calcium homeostasis dependent on the activity-driven regulation of NMDAR conductances. A neuron thus implemented is stable, and displays properties analogous to previously described metaplasticity and synaptic scaling. We suggest that these are related processes, sharing a common underlying mechanism. In addition, simulations show that the model displays synaptic competition, allowing for structure formation in the synaptic space. Inhomogeneous distribution of synaptic weights emerges in response to variations in the statistics of input spike-trains. We compare these results with known properties of developing neurons, such as the formation of input specificity. We conclude that this model is suitable for describing a variety of single-cell properties, and can serve as a biological basis for a more general theory of synaptic plasticity. |
