Regulation of retinal ganglion cell axon arbor size by target availability: mechanisms of compression and expansion of the retinotectal projection.

The ability of pre- and postsynaptic populations to achieve the proper convergence ratios during development is especially critical in topographically mapped systems such as the retinotectal system. The ratio of retinal ganglion cells to their target cells in the optic tectum can be altered experimentally either by early partial tectal ablation, which results in an orderly compression of near-normal numbers of retinal projections into a smaller tectal area, or by early monocular enucleation, which results in the expansion of a reduced number of axons in a near-normal tectal volume. Our previous studies showed that changes in cell death and synaptic density consequent to these manipulations can account for only a minor component of this compensation for the population mismatch. In this study, we examine other mechanisms of population matching in the hamster retinotectal system. We used an in vitro horseradish peroxidase labeling method to trace individual retinal ganglion cell axons in superior colliculi partially ablated on the day of birth, as well as in colliculi contralateral to a monocular enucleation. We found that individual axon arbors within the partially lesioned tectum occupy a smaller area, with fewer branches and fewer terminal boutons, but preserve a normal bouton density. In contrast, ipsilaterally projecting axon arbors in monocularly enucleated animals occupy a greater area than in the normal condition, with a much larger arbor length and greater number of boutons and branches compared with normal ipsilaterally projecting cells. Alteration of axonal arborization of retinal ganglion cells is the main factor responsible for matching the retinal and tectal cell populations within the tectum. This process conserves normal electrophysiological function over a wide range of convergence ratios and may occur through strict selectivity of tectal cells for their normal number of inputs.