Information flow in cultured neuronal networks by transfer entropy

Neurons form functional connections in neuronal networks of the brain. Neurotransmission refers to the information flow of electrical signals between neurons, and changes in the strength of this flow characterize brain functions. An evaluation method for neuronal information flow strength is necessary to elucidate the basic principles of brain function. To analyze the strength of information flow, identifying information transmission between time series of electrical spikes in neuronal networks, such as spike trains, is required. In this study, we evaluated the information flow strength in neuronal networks using transfer entropy (TE), an analysis method based on information theory, to elucidate causal relationships between two spike trains. Cultured rat hippocampal neuronal networks were used as living brain models, and extracellular monitoring of action potentials (expressed as spikes) was performed using microelectrode arrays. Spike trains of spontaneous activity and stimulus-evoked responses were measured in multiple neurons, and the causal relationship between electrode pairs was evaluated as an index of information flow between neurons using TE. From the results of spontaneous activity and evoked responses, it was suggested that TE values between electrodes increased with culture days and that the information flow strength was enhanced by network maturation. Furthermore, the correlation between TE values and neuronal population distance was evaluated using TE analysis. We found that neurons that received strong inputs slightly overlapped in their spontaneous activity and evoked responses. These findings suggest that spontaneous activity and evoked responses exhibit similar patterns in neuronal networks. In the graph theoretical analysis based on TE values, the network topology exhibited changes that strengthened information flow over 70 days in culture. Therefore, TE analysis is an effective tool for estimating the information flow between neurons based on neuronal electrical activity recorded at multiple sites.