Purpose:
In 1991 we hypothesized that optimal neuronal information encoding requires higher order coherence in the timing of action potentials (spikes)[1]. That is, the time any spike occurs is determined by the exact timing of multiple previous spikes. We now present evidence of such spike coherence in the visual code of the retina. RGCs are an ideal substrate to test this, since they have no input from other brain structures. A mechanistic understanding of how the retina encodes visual stimuli can help restore sight effectively.
Methods:
We used in vitro multielectrode techniques to record RGC activity in 10 wt mouse retinas with or without visual stimulation (full field flashes or random checkerboard sequences). Recorded spike trains were analyzed for first and second order coherence by determining the probability of a spike occurring, given that one or more spikes occurred at specific preceding delays. Thus, we plotted P(s0 | Δs1) and P(s0 | Δs1, Δs2), where Δs1 (or Δs2) is the delay with which spike1 (or 2) occurs before s0.
Results:
Plots of P(s0 | Δs1) and P(s0 | Δs1, Δs2) revealed coherence patterns that are statistically distinct from incoherent spike trains such as a Poisson distribution of spike intervals. These patterns are consistent for individual cells but different among cells. They differ between unstimulated, checkerboard and flash stimulus conditions (Figure).
Conclusions:
This is the first demonstration of higher order spike coherence in the activity of RGCs. This temporal structure is distinct from that previously reported for interspike intervals (ISIs). Further analysis of RGC spike coherence may help 1) evaluate the effectiveness of sight-restoring therapies, and 2) design more effective stimulation with a retinal prosthesis to elicit meaningful visual perceptions in the brain.
Coherence patterns (Δs1 vs , Δs2) with different stimuli in the same cell (high coherence: arrows; low coherence: triangles).1. Abramoff (Braamhof), M. Spatiotemporal correlation in the cerebellum. in Neural Networks, 1991., IJCNN-91-Seattle International Joint Conference on. 1991.
Keywords: electrophysiology: non-clinical • ganglion cells • computational modeling