Natural neural networks for quantitative sensing of neurochemicals: an artificial neural network analysis

In cell culture, mammalian neurons form fault tolerant, spontaneously activ e systems with gnat sensitivity to their chemical environment and generate response profiles that are often concentration and substance specific. In t he experiments to be discussed, the response of spontaneously active murine spinal cord cultures coupled to an array of 64 transparent microelectrodes to the glycine receptor blocker strychnine was evaluated. Strychnine relia bly generated increased multichannel bursting at 5-20 nM and regular, coord inated bursting above 5 mu M. In comparison bicuculline was applied which s hows a similar network response but at higher concentrations. By principal component analysis a good discrimination of signals produced by strychnine and by bicuculline was possible. By artificial neural network analysis a qu antitative interpretation of the strychnine data was shown for the first ti me. In particular it could be demonstrated that for subsequent strychnine e xpositions a quantitative analysis of an unknown strychnine concentration i s possible. These results indicate that cultured neuronal networks are prac tical systems that can be used as a biosensor system for the characterizati on of a great variety of neuroactive substances. (C) 2000 Elsevier Science S.A. All rights reserved.