Origins of Activity Patterns in Self-Organizing Neuronal Networks in Vitro.

Nerve cell ensembles growing in culture on microelectrode arrays allow long-term, multisite monitoring of spontaneous and evoked action potential (spike) activity. These ensembles generate complex spatio-temporal spike patterns and constitute evolving non-linear and non-station-ary dynamical systems. They provide effective platforms for observations of the "internal dynamics" of neuronal networks, especially pattern generation and the evolution of network organization. Nerves cells in cultures form many small neuronal clusters (termed nacelles) in which substantial membrane contact is achieved. We propose that such morphological clusters generate subthreshold interactions that can lead to noise-induced membrane potential oscillations with frequent ignition to spiking. Recruitment of several nacelles and dispersed neurons then leads to dynamically organized circuits capable of reliably generating and maintaining a specific burst pattern. The observed coexistence of different burst patterns suggests network substructures (circuits) with autonomous activity. Entrainment of all circuits to a global pattern is thought to represent a network "decision state" with automatic fault-tolerant output. Such phenomena suggest that not only the redundancy of connections but also the form of the message are important factors in the creation of fault-tolerance.