We have used the anterograde axon tracer 1,1'-dioctodecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate (Dil) to characterize the development of topographic order in the rat retinocollicular projection. Retinal axons were labeled by Dil injections covering 0.15-2% of peripheral temporal, nasal, superior, or inferior retina, or more central retina, in rats ranging in age from embryonic day 20 to postnatal day (P) 19. At P11-P12 and later, such injections label retinal axons that form overlapping arbors restricted to a topographically correct terminal zone covering about 1% of the superior colliculus (SC) area. At perinatal ages, though, axons labeled from each retinal site are distributed in the SC over much of its medial-lateral axis and extend caudally well beyond the rostral-caudal location of their correct terminal zone; some continue caudally into the inferior colliculus. Axons typically form side branches and often arborize at topographically incorrect positions throughout the SC; however, they appear to branch preferentially in a region that includes, but is much larger than, their correct terminal zone. The mature, retino-topically ordered projection emerges during an early postnatal remodeling period through the rapid remodeling of the early, diffuse projection. This process involves the large-scale removal of axons, axon segments, branches, and arbors from topographically inappropriate positions concurrently with a dramatic increase in branching and arborization at topographically correct locations. Quantitative measurements show that elimination of aberrant branches without loss of the primary axons contributes substantially to the development of order. By P6, fewer mistargeted axons persist, but those that do persist tend to branch or arborize more extensively in topographically inappropriate regions. By P8, the labeling patterns begin to approximate those seen at maturity. Further refinement leads to an adultlike topographic ordering of axonal arborizations by P11-P12. At maturity, some axons take very indirect routes to reach their correct terminal zone. However, such trajectory changes typically correct only small positional inaccuracies, indicating that axons and axon segments that make larger targeting errors do not survive the remodeling phase. Previous retrograde labeling studies indicate that some retinal axons make topographic targeting errors (O'Leary et al., 1986; Yhip and Kirby, 1990), but none have suggested the degree of diffuseness revealed by anterograde labeling with Dil. Our findings show that directed axon growth is inadequate as a mechanism to develop the topographic ordering of retinal axons in the rat SC. Rather, mechanisms that control the removal of mistargeted axons and promote the arborization of correctly positioned axons are critical for the development of retinotopic order. These data characterizing the time course of the development of topographic order in the rat retinocollicular projection provide a frame-work for interpreting the potential roles that putative position-encoding molecular cues in the retina and SC might play in this process.
