|
The trial-to-trial variability of neuronal response patterns is a prominent feature of sensory systems and has a profound impact on subsequent sensory signal processing. Fluctuations of the underlying ionic currents represent a major intrinsic noise source that causes neuronal response variability. To characterize these intrinsic stochastic properties of a neuron, direct somatic recordings are well suited. In many systems, however, such recordings are difficult to achieve without severely damaging the sensory transduction machinery. In this study, we therefore introduce an indirect approach to assess the stochastic dynamics of sensory neurons based on interspike interval statistics of the spike train responses. Spike responses of receptor cells were recorded intracellularly from auditory nerve fibres of Locusta migratoria during simultaneous acoustic stimulation with pure tones of various intensities. As many other neurons, the auditory receptors exhibit spike-frequency adaptation. In the steady state, the interspike intervals (ISIs) show high variability with CVs up to 0.9 depending on sound intensity. With increasing spike frequency the shape of the ISI histograms changes from an inverse Gaussian to a peaked probability density. Additionally, the ISI correlations exhibit a shift from slightly negative values to positive coefficients with increasing spike rate. By means of simulations of a perfect integrate-and-fire as well as a more realistic Hodgkin-Huxley-type model, Schwalger et al. demonstrated that positive serial correlations and peaked ISI densities can arise from stochastic ion channels mediating spike-frequency adaptation while models with a deterministic adaptation current and an additional white noise input revealed negative serial correlations and an inverse Gaussian ISI density. Using simulations of single-compartment conductance-based models we tested different assumptions of possible channel noise sources which could account for the observed transitions of the locust ISI histogram shape and correlation. In a mixed channel noise model with both fast ion channel fluctuations, like from the mechanosensory receptor channels, and slow stochastic adaptation currents we were able to replicate all the features seen in the interspike interval statistics of the spike train responses of locust auditory receptor cells. This indicates that higher-order statistics can be used to distinguish different kinds of noise sources. |
|