Cortical oscillations and synaptic plasticity : from a single neuron to neural networks

Abstract

Cortical oscillations have been widely observed in the brain cortex and play a crucial role for understanding temporal correlations between neurons in cortical circuits. Synchronous activities of neuronal populations, which produce cortical oscillations, have close relationship with synaptic connectivity of neural networks. In this thesis, two kinds of cortical rhythms - beta oscillation (10 ~ 30 Hz) andgamma oscillation(30 ~ 80 Hz) - together with the evolution of neural network by synaptic plasticity are investigated. Firstly, the influence of distal-dendritic inhibitory oscillations on the somatic firing rate of a biologically detailed layer V pyramidal neuron model is examined. An optimal release from the impact of inhibitory inputs oscillating around beta frequency is observed at the apical tuft dendrites. This preferred frequency range for the maximal firing rate is free from the influence of many parameters, including the frequencyof pre-synaptic inhibitory spikes, the peak conductance of the GABA receptor synaptic conductance and NMDA receptors. However, removing the hyperpolarization-activated current eliminated this resonance. There is also a critical dependence of resonant frequency on the existence of bursts. The optimal frequency could be remarkably shifted by varying the degree of synchronization of the pre-synaptic spikes and the peak conductance of voltage-gated calcium channels, which are curial for the generation of bursts. When the oscillatory inhibition targets the soma, the resonance frequency is decreased to a lower range due to the strong and direct somatic inhibition. Moreover, gain modulation by top-down excitatory inputs is shown to be able to significantly affected by the distal-dendritic inhibition, suggesting that oscillatory distal inhibition of varying frequencies could provide a strategy for the gain modulation of pyramidal neurons. Secondly, the generation of gamma oscillations in a fast-spiking interneuron network is investigated. Simulations show that gap junctions remarkably improve the robustness of synchrony against heterogeneity. With only synaptic inhibition, but lacking gap junctions, the network is not able to reach synchronization no matter how the inhibitory synaptic conductance is increased. Using the network of interneurons oscillating synchronously at gamma frequency, we reproduce the experimentally observed regulation of gamma inhibition on pyramidal neurons' response. The rhythmically gating effect of the gamma-oscillated interneuron networks on pyramidal neurons’ signal transmissionis further reflected by the interactions of this interneuron network and a single pyramidal neuron. Prominent frequency power at gamma-range can also be observed from the excitatory-inhibitory loop. Besides, synaptic delays have substantial effect on the regulation of feedback inhibition to the pyramidal cell. Finally, in order to understand how synchronous activity emerges from self-organized neural networks, we propose a novel network refined from spike-timing dependent plasticity (STDP). Due to the existence of heterogeneity in neurons which exhibit different degrees of excitability, the network finally evolves into a sparse and active-neuron-dominant structure. That is, strong connections are mainly distributed to the synapses from active neuronsto inactive ones. We argue that this self-emergent topology essentially reflects the competition of different neurons and encodes the heterogeneity. This structure is shown to significantly promote synchronization and enhance the coherence resonance and stochastic resonance of the entire network, indicating its high efficiency in information processing. Based on this work, we further develop another network organized from two stages of learning process, including STDP and another burst-based plasticity, i.e., burst-timing dependent plasticity (BTDP). The likely relationship between the learning rules with different timescales and the formation of architecture with different special scales is explored. The final network exhibits a two-level hierarchical structure after the synaptic refinement. This self-organized network shows higher sensitivity to afferent current injection compared with alternative archetypal networks with different neural connectivity. Statistical analysis also demonstrates that it has the small-world properties of small shortest path length and high clustering coefficient. Thus the selectively refined connectivity enhances the ability of neuronal communications and improves the efficiency of signal transmission in the neural network.