We analyzed and quantified the distribution and levels of DN and VN signals across the optic tract (Figures 1G–1I) and generated a missorting index (MI) measuring the proportion of missorted DN axons (Figure 1J). While sorting of VN axons was accurate at all stages observed, a significant number of DN axons were clearly missorted at 48 hpf (MI 23.3%). Missorting of DN axons progressively

decreased at later stages (MI 10.8% at 54 hpf) and was no longer observed at 72 hpf. These results indicate that pretarget sorting of retinal axons is not precisely established during initial growth cone guidance along the optic tract but rather is achieved by correcting missorted DN projections. How are missorted DN axons corrected? Missorted axons could respond to a specific cue and retract,

as has been shown during pathfinding PLX4032 mouse error correction at the chiasm (Hutson and Chien, 2002). Alternatively, they could “shift” posteriorly and reach the ventral branch of the tract through selective fasciculation. Finally, they could be removed through selective degeneration. To identify the mechanism involved, we observed the behavior of DN and VN axons as they elongate LY294002 mouse along the tract by confocal time-lapse imaging (Figure 2, Movies S1 and S2). We topographically injected dyes intraocularly at 48 hpf, began imaging at ∼54 hpf, and collected confocal z series every why 15 min for up to 12 hr. Based on axon morphology and behavior, no evidence of phototoxicity was observed. We also injected different combinations of dyes in various conditions to ensure that the dyes we used

did not have any toxic effect (data not shown). In all our time-lapse analyses (n = 12), axons were very dynamic, elongating rapidly and sometimes pausing during their navigation. During these pauses, growth cones of correctly sorted axons harbored many dynamic filopodia that extended and retracted quickly. Interestingly, missorted DN axons had a different behavior. Although initially as dynamic as axons that were correctly sorted, they then stopped their elongation and became more quiescent. As they stopped, blebbing started to appear uniformly along their length (Figure 2B). This blebbing phase was rapidly followed by an abrupt and uniform fragmentation (Figures 2C–2F). Axonal fragments were of different sizes and appeared disconnected. Fragmentation of missorted DN axons was a rapid process, occurring over a period of 30 min (Movie S2, Figures 2J and 2K). Fragments became smaller over time and eventually disappeared (Figure 2G). This last clearance phase was slow and not always observed in our time-lapse conditions, even after 14 hr of imaging. Our observations thus indicate that topographic sorting of retinal axons along the optic tract is achieved through the selective degeneration of missorted DN axons.