Predicting the responses of mechanoreceptor neurons to physiological inputs by nonlinear system identification.

Abstract

The nonlinear dynamic properties of action potential encoding were studied in mechanosensory neurons innervating the slits of a slit-sense organ in the tropical wandering spider, Cupiennius salei. The organ contains two types of neurons that are morphologically similar but have different dynamic properties. Type A neurons produce only one or two action potentials in response to a mechanical or electrical stimulus of any suprathreshold amplitude, while type B neurons can fire prolonged bursts of action potentials in response to similar stimuli. Neurons were stimulated with pseudorandomly modulated intracellular current while recording the resultant fluctuations in membrane potential and action potentials. A parallel cascade method was used to estimate a third-order Volterra series to describe the nonlinear dynamic relationship between membrane potential and action potentials. Kernels measured for the two types of neurons had reproducible forms that showed differences between the two neuron types. The measured kernels were able to predict the responses of the neurons to novel pseudorandomly modulated inputs with reasonable fidelity. However, the Volterra series did not adequately predict the difference in responses to step depolarizations.