|
Ulrike Ziehm1*, Jan Benda2 1 Theoretical Neuroscience/Neuroinformatics, Institute of Biology Freie Universitaet Berlin and Bernstein Center for Computational Neuroscience Berlin, Germany 2 Department of Biology and Bernstein Center for Computational Neuroscience Muenchen, Ludwig-Maximilians-Universitaet Muenchen, Germany The adjustment of neuronal sensitivity to the mean intensity of a stimulus is a common feature in many sensory systems and practically all levels of sensory processing. Little, however, is known about the cellular mechanisms causing distinct adaptive response properties of neurons within a network. We present a biophysically plausible model of early processing levels in the auditory system of the cricket. The cricket's auditory interneuron AN1 receives convergent input from ipsilateral auditory receptor neurons sensitive in the low frequency regime. Individual receptors have sigmoid rate-level-functions with narrow dynamic ranges. Receptors within the considered population have similar sound-frequency tuning and adaptation properties, but differ in response threshold so that the large range of physiologically relevant intensities is partitioned across the population. The AN1's onset rate-level-functions have a dynamic range comparable to that of an individual receptor. Unlike receptors its rate-level-function can shift over a considerable range to match the current mean stimulus intensity. The adaptive performance to background intensity observed in the AN1 cannot be explained by simple summation of receptor responses and additional adaptation currents in a single compartment model of this interneuron. Central features of rate-level-functions and adaptation in the AN1 can be reproduced by extending the model by a dendritic compartment in which synaptic inputs can saturate at the synaptic reversal potential. The AN1 responses (model and experiment) become strictly intensity invariant only at high intensities. We also constructed a phenomenological model of the network which shows true intensity invariance over a almost the whole range of relevant intensities. A comparison of both models with respect to information transmission (quantified by Fisher-information) revealed that the biophysically realistic model with synaptic saturation performs comparably good. |
|