The influence of corticofugal feedback on the temporal structure of visual responses of cat thalamic relay cells.

1. Visually driven single-unit activity was recorded in the dorsal lateral geniculate nucleus (dLGN) of the anaesthetized cat while inactivating or stimulating the corticofugal feedback from area 17/18 by means of cortical cooling or application of GABA (inactivation), or application of glutamate or quisqualate (Glu, Quis; stimulation) to layer VI. 2. Manipulations of the corticofugal feedback primarily affected the multimodal interspike interval pattern previously reported to be present in the tonic component of visual responses elicited by spot-like stimuli. 3. Sixty-three per cent of all neurons could be influenced, and temporally localized interspike interval distributions were measured which commonly consisted of one fundamental interval peak (leftmost peak) and integer multiples thereof (higher order peaks). During blockade of the corticofugal feedback, interspike intervals were redistributed into the higher order peaks in about 70 % of the cases, accompanied by a reduced mean firing rate. During stimulation the reverse effect occurred in 69 % of cases. 4. Increased synchronization of the EEG (increased power in the delta-wave range, 1-4 Hz) had an effect similar to cortex inactivation. The specificity of corticofugal effects was verified by consideration of these EEG effects and by dLGN double recordings with one dLGN cell topographically matched with the cortical inactivation/activation site and the second cell outside this area. Clear effects due to manipulation of the corticofugal feedback were found only for the matched dLGN site. 5. In addition we observed that the peaks of the interval distributions were narrower during active corticofugal feedback, such that the temporal dispersion of the signal transmission to the cortex was reduced. 6. The mechanisms underlying this effect were further analysed in a biophysically realistic model demonstrating that the timing of the spikes in the dLGN is improved as soon as the cortical feedback is active. The high degree of convergence/divergence between neurons along the closed feedback loop thereby leads to a temporal averaging effect which reduces the interval dispersion and also introduces synchronization between dLGN cells. 7. Such a mechanism may thus counteract the deterioration of spike timing accuracy which would otherwise occur as a consequence of synaptic noise and other uncorrelated sources of activity at a given neuron.