MICROCIRCUITRY OF PARALLEL CHANNELS IN VISION

人类的视觉表现主要取决于神经回路的效率

Human visual performance depends critically on the efficiency of circuits that connect foveal cones to ganglion cells. Our broad goal is to discover the basis for this efficiency. In the standard view all information destined for cortex is "multiplexed" by two arrays: P/midget (95% of foveal ganglion cells) and M/parasol (5%). However, our current studies suggest that non-midget cells are more prevalent than suspected (>25%) and comprise four types. Thus, we hypothesize five ganglion cell arrays: a dense array of spectrally nonselective cells (midget); two sparser arrays of spectrally opponent cells (R/G & B/Y bistratified); a moderately dense array of linear, high contrast gain cells (beta-like parasol); and a sparse array of non-linear, motion-sensitive cells (alpha-like parasol). If confirmed, such arrays suggest that signal are not multiplexed but segregated early, thus permitting maximum amplification. We now propose: 1) Test the "five array" hypothesis by determining quantitatively the circuits for all four types of non-midget ganglion cell. This should suggest possible correspondences to psychophysical channels because the number and weighting of cones connected to a ganglion cell type, plus its sampling frequency, set the information capacity of the array. 2) Test hypothesis that R and G midget ganglion cells in human receive different numbers of synapses. This will link our current studies on macaque to human circuitry and test our identification of spectrally opponent inputs to bistratified cells. 3) Determine amacrine circuits to ganglion cells. Amacrine synapses are numerous (e.g., 50% of the synapses to midget cells), but their circuits, which probably serve nonlinear mechanisms such as gain control, are completely unknown. 4) Determine which members of the ionotropic and metabotropic families of glutamate receptor are expressed on specific types of bipolar and ganglion cell. A circuit's coding capacity (i.e., signal/noise, gain, temporal bandwidth) depends critically on molecular properties of its postsynaptic receptors (e.g., binding constant, channel conductance and open time). Localizing these known properties in identified circuits will provide data essential to AIM 5. 5) Assess how identified factors (circuit structure, sampling frequency, and postsynaptic receptor properties) affect coding efficiency. Incorporate neural factors into "ideal observer" models for comparison to preneural factors and psychophysical performance. Circuits will be studied by electron microscopy and by intracellular dye injection followed by digital light microscopy; glutamate receptors will be identified by immunocytochemistry and amplification of mRNA in identified cells; compartmental models will be used to assess how circuits optimize information transfer given their constraints: to be small (few synapses; prone to saturate) and to employ noisy mechanisms for signal transfer.