Spike timing and information transmission at the retinogeniculate synapse

Previous studies have demonstrated that the probability of a spike in the lateral geniculate nucleus of the thalamus (LGN) immediately following an incoming retinal spike is inversely proportional to the interspike interval (ISI) preceding that retinal spike. To understand the role of this phenomenon in information processing, we first examined the visual information encoded by retinal spikes that followed different ISIs by recording from the cat optic tract under binary white-noise stimulation. We found that stimulus/response correlation was highest for retinal spikes that followed short ISIs and decreased with increasing ISI, demonstrating that encoded visual information varies with ISI. We next recorded from monosynaptically connected retinogeniculate cell pairs under white-noise stimulation to compare responses in the retina and LGN directly. Using cross-correlation analysis, we sorted retinal spikes into those which were relayed by the LGN and those that were not relayed. As expected, relayed spikes had higher stimulus/response correlation than non-relayed spikes. Furthermore, relayed retinal spike events have lower spike count variability and lower spike timing variability, yielding more encoded visual information (bits/spike). Furthermore, relayed spikes had slightly longer response latencies, suggesting a trade-off between coding fidelity and transmission speed. To determine how stimulus contrast influences information transmission, we compared the responses of retinogeniculate cell pairs to drifting sine-wave gratings. We found that contrast-dependent phase advance (CDPA), a measure of contrast gain control, was greater in LGN neurons than their retinal inputs, reflecting an influence of contrast gain control on retinogeniculate processing. Finding that CDPA was also greater for relayed retinal spikes than non-relayed spikes, we implemented a model of retinogeniculate filtering to test the hypothesis that ISI-based spike filtering leads to increased CDPA. In accordance with our hypothesis, the model produced greater CDPA, suggesting that ISI filtering may be one mechanism underlying contrast gain control