Independent mosaics of large inner- and outer-stratified ganglion cells in the goldfish retina.

Goldfish retinal ganglion cells were filled with horseradish peroxidase and studied in flatmounts. Two regular mosaics of large neurons with many of the properties of mammalian alpha ganglion cells were found, differing from each other in spacing, size, and dendritic stratification. The existence of biplexiform ganglion cells with additional dendrites in the outer plexiform layer was also confirmed. One of the two alpha-like mosaics consisted of giant ganglion cells with thick primary dendrites and large, sparsely branched dendritic trees in the outer sublamina of the inner plexiform layer (IPL). In fish 55-65 mm long, about 300 formed a tessellated array across each retina. Their somata (mean area 277 +/- 6 microns 2) were displaced to varying degrees into the IPL, neighbours in the mosaic often occupying different levels. Their dendrites ramified in one stratum near the inner nuclear layer, at a mean depth of 70.8 +/- 0.5% of the IPL. The other alpha-like mosaic comprised about 900 large ganglion cells, with slightly smaller somata (mean area 193 +/- 4 microns 2) in the ganglion cell layer. Most of their dendrites lay in a narrow stratum at 41.9 +/- 0.5% of the depth of the IPL. However, deviations (usually into more vitread strata) were common, which was not true for similar cells in the distantly related cichlid fish Oreochromis. Measurements of nearest neighbour distance (NND) for 4 outer and 4 inner mosaics showed that they were at least as regular as the alpha cell mosaics of mammals: the ratio of the mean NND to the standard deviation ranged from 4.03 for the least regular outer mosaic to 6.47 for the most regular inner mosaic. The wide phylogenetic distribution of these paired, regular mosaics points to a fundamental role in vision. The presence of some variability in dendritic stratification even within the exceptionally regular inner-stratified mosaic suggests that classifications based entirely on the detailed morphology of individual neurons may not always correlate well with their primary functional roles. Where possible, neuronal morphology and spatial distribution should be studied together.