We evaluated the two-dimensional distribution of superior colliculus (SC) neurons visualized after retrograde transneuronal transport of rabies virus injected into the lateral rectus muscle of rhesus monkeys to test whether the density of projection neurons might play a role in the spatiotemporal transformation and vector decomposition. If this were the case, the number of horizontal eye movement-related SC neurons should increase with their distance from the rostral pole of the SC and decrease with their distance from the representation of the horizontal meridian. Labeled neurons of the intermediate SC layers were counted inside a 1-mm-wide band that matched the horizontal meridian of the collicular motor map. Local areal densities were plotted against distance from the rostral SC pole. At 2.5 days after inoculation, there was no labeling in the SC. At 3 days, moderate labeling appeared on both sides, mostly in the intermediate layers. At 3.5 days, cell numbers substantially increased and the laminar distribution changed as cells appeared in the superficial SC layers. At 3 days, rostrocaudal density profiles were unimodal, with peaks at locations near 50 degrees (contralateral SC) and 25-30 degrees (ipsilateral SC) horizontal eccentricity. At 3.5 days, distributions were bimodal due to the appearance of a second high-density region near the rostral pole of the SC. The distribution of SC neurons influencing the abducens nucleus, thus, was nonuniform. Caudal sites contained more neurons, but the experimentally observed density gradients were shallower than the theoretically predicted ones that would be necessary to fully account for the spatiotemporal transformation. Similarly, we studied the distributions of cell densities in the intermediate SC layers along an isoamplitude line (representing saccades of equal amplitudes but different directions). Consistent with theoretical estimates of the density gradients required for vector decomposition, we found that the concentrations of labeled cells were highest in the vicinity of the horizontal meridian but their decrease toward the periphery of the motor map was steeper than predicted. We conclude that SC cell density gradients cannot fully account for the spatiotemporal transformation and vector decomposition in the absence of an additional mechanism such as the previously demonstrated (Grantyn et al.,  Soc. Neurosci. Abstr. 23:1295; Moschovakis et al.,  J. Neurosci. 18:10219-10229) locus-dependent weighting of the strength of efferent projections to the saccade generators.
Copyright 2002 Wiley-Liss, Inc.