Retinal ganglion cells (RGCs) yield highly reproducible responses to a repeated stimulus. Yet investigators routinely model spike generation as a memoryless stochastic process and claim that it limits efficiency of information transfer. We sought experimentally to identify the sources of variability that most limit RGC response reproducibility. We also extended work by Krishnaswamy (showing higher RGC firing rates in response to natural images, relative to random phase images having naturalistic power spectra), by examining firing rate and reliability of responses to natural and computer-generated diffuse, but temporally modulated, stimuli.
Whole-cell recordings of ganglion cells from tiger salamander retinae were made under voltage- and current-clamp. The stimulus, delivered in darkness or in the presence of a rod-saturating background light, was either a time-varying diffuse light (16.6s long, photopic) or an injected "synaptic" current (created by recording the cells response to the diffuse light stimulus under voltage-clamp). Impulse response properties and measures of spike train reproducibility were used to quantify response variability.
We found the dominant cause of trial-to-trial spike train variability to be systematic, slow adaptational changes in the cell's input-output characteristic. Variability in pre-synaptic inputs and dendritic integration of synaptic inputs may also impact reproducibility. However, response reliability was not limited by background transmitter release, somatic V-gated channel fluctuations, or the spike generator.
Our results suggest that the RGC spike generating mechanism is appropriately modeled as a stochastic process with memory.
Ref: Krishnaswamy JT (2003) PhD Thesis UC Berkeley
Support Contributed By: NIH EY03785, T32 EY07043
T Madon, WG Owen. (2004) Limitations on the Reliability of Retinal Ganglion Cell Spike Trains. Soc. for Neurosci. Abstr 408.11