Recruitment of Retinal Ganglion Cell Coherent Spiking is Mediated by Electrically Coupled Networks and not by Single Coupled Neighbors

Purpose : Retinal ganglion cells (RGCs) are extensively coupled via gap junctions (GJs) made with RGC neighbors and/or amacrine cells. These electrical circuits underlie different patterns of correlated spike activity. For example, cross correlation profiles (CCPs) between coupled RGCs often show two peaks, suggesting reciprocal interactions in which spike activity in one RGC can induce activity in the coupled neighbor. Here, we examined directly the ability of spikes evoked in RGCs to induce correlated spike activity in electrically coupled neighbors.

Methods : Simultaneous whole-cell and loose-patch recordings were made from alpha RGCs in Kcng4-YFP transgenic mice. Extrinsic current was applied to induce spike activity in one alpha RGC, which was compared to spike activity in the coupled neighbor by computing CCPs. To examine recruitment of RGC activity by electrically coupled networks, channelrhodopsin (ChR2) was delivered to the retina via AAV infection and recordings were made with a multielectrode array.

Results : In patch recording experiments, we found that spikes evoked in single alpha RGCs using extrinsic current injection could not reliably produce correlated spike activity in coupled RGC neighbors. In ChR2 experiments, blue light produced correlated activity in pairs of ChR2-expressing RGCs regardless of their inter-somatic distance. As expected, these correlations were due to independent expression of ChR2 and survived blockade of synaptic transmission. In contrast, light-evoked correlated activity derived from photoreceptors was seen for RGC pairs separated up to 360 µm and was abolished by the cocktail of antagonists. Under synaptic blockade, ChR2-induced activity could generate spikes in coupled non-ChR2-expressing RGC neighbors at an efficiency rate of nearly 60%. These correlations were unaffected by application of synaptic blockers, but were abolished by addition of GJ blockers.

Conclusions : Our results show that single RGCs cannot reliably evoke correlated spike activity in coupled RGC neighbors. Rather, a network of RGCs with correlated spike activity is needed to further recruit RGCs with the correlated spiking pattern. While our data indicate that a large correlated network activated with ChR2 can recruit cells with high efficiency, the size of correlated RGC electrical networks produced under physiological conditions needs to be determined.