This abstract was presented today at the Association for Research in Vision and Opthalmology (ARVO) meetings in Ft. Lauderdale, Florida by Robert E. Marc, J. Scott Lauritzen, Bryan W. Jones, Carl B. Watt and James R. Anderson.
Purpose: Transitions between scotopic and photopic switch seem smooth, but psychophysical dissection reveals that it is underpinned by mutual rod-cone suppression processes. Further, rod and cone signals mix to generate shifted spectral percepts. The neural architecture supporting these processes has resisted discovery.
Methods: Multiple amacrine cell (AC) networks connecting 70 rod and >100 cone bipolar cells (BCs), 30 A-II and 20 A-I ACs were traced in the ultrastructural rabbit retinal connectome RC1, annotated with the Viking viewer, and explored by 3D rendering and graph visualization of connectivity (Anderson et al. 2011. The Viking Viewer. J Microscopy). RC1 contains embedded small molecule signals, enabling complete cellular classification independent of network identity.
Results: Multiple GABAergic AC pathways connect rod and cone BCs. (1) Certain wide-field GABAergic ACs are reciprocal feedback elements at every ON cone BC they encounter, but also collect rod BC input enabling rod suppression of cone signals. (2) Conversely, A-I(S2) ACs are reciprocal feedback ACs at rod BCs but are also presynaptic to some ON cone BCs, also enabling rod suppression of cone signals. (3) Every rod BC receives inhibitory input from GABAergic ACs driven directly by ON or OFF cone BCs. Instances of ON glycinergic AC > rod BC inhibition also exist. (4) Every A-I AC receives massive inhibition via OFF cone BC > OFF cone AC > A-I chains (Fig. 1). (5) Every A-II AC receives inhibitory input from multiple cone BC driven ACs. (6) A-II amacrine cells collect ribbon input signals from wide-field ON cone BCs.
Conclusions: The mammalian retina appears to use ACs to create a winner-take all architecture for rod and cone bipolar cells. When rod responsivity exceeds cones, multiple inhibitory networks further suppress the cone pathway output, and vice versa.
Four primary synaptic chains support this process:
Rod BCs > GABAergic ACs > ON and OFF Cone BCs
ON and OFF Cone BCs > GABAergic ACs > Rod BCs
ON and OFF Cone BCs > GABAergic ACs > A-II ACs
OFF Cone BCs > GABAergic ACs > A-I ACs > Rod BCs
Rod signals directly mix with cone signals via A-II cells potentially driving color shifts. If the wide-field BCs of the rabbit retina are blue-dominated, this may explain blue-biased hue shifts near rod threshold.