At P9, all BC types examined contained synapses at about half their appositions
with G10 dendrites (p > 0.5, ANOVA test). By P21, three distinct patterns emerged (p < 0.001, ANOVA test). B7 cells maintain a constant connectivity fraction (P9: 0.51 ± 0.11 synapses/apposition, n = 9; P21: 0.46 ± 0.15 synapses/apposition, n = 13; p > 0.6). RB cells, which keep appositions with G10 RGCs, lose all synapses from them (P9: 0.79 ± 0.41 synapses/apposition, n = 7; P21: 0 ± 0 synapses/apposition, n = 14; p < 0.001), and B6 cells increase the rate of conversion and typically form more than one synapse per apposition at P21 (P9: 0.60 ± 0.13 synapses/apposition, n = 14; P21: 1.35 ± 0.10 synapses/apposition, this website n = 35; p < 0.001). Because synaptic rewiring from P9 to P21 appeared independent of changes in the number of appositions between BCs and RGCs (Figure 2), we wanted to test whether the formation and elimination of individual synapses was likewise uncoupled from the dynamics of axo-dendritic appositions. To address this question we generated transgenic mice in which nearly all ON BCs express YFP (Grm6-YFP; Figure S3) and biolistically labeled RGCs with tdTomato and PSD95-CFP. Time-lapse imaging every 2 hr for up to 18 hr revealed very little synaptic beta-catenin inhibitor turnover at P21, supporting the notion that circuits are mostly mature at this age (data
not shown). By contrast, at P9 we frequently observed synapse formation (1.3% ± 0.41% of synapses/hr, n = 8 cells) and elimination events (0.77% ± 0.31% of synapses/hr)
distributed throughout the RGC dendrite ( Figures 3A–3C). When we analyzed the presence of axo-dendritic appositions relative to the timing of synapse formation and elimination events at P9 (Figures 3D and 3E), it became evident that synaptic dynamics of BC-RGC pairs are not tightly linked to changes in appositions. Every one of 40 synapse formation events we observed during time-lapse imaging occurred not at BC-RGC appositions that were present at the first time point of the series, 2–8 hr before synapse formation (Figures 3B and 3D). Similarly, most appositions persisted for many hours after synapse elimination (Figures 3C and 3E). While some appositions dissociated after synapse elimination, these events did not occur more frequently than expected by chance given the average stability of appositions without synapses (data not shown). Together, these results argue that converging excitatory inputs establish cell type-specific patterns of connections by differential synaptic conversion of relatively stable axo-dendritic appositions. In our analysis of connectivity patterns we distinguished synaptic and nonsynaptic appositions based on the presence and absence, respectively, of PSD95-YFP puncta from sites of axo-dendritic overlap. In support of this distinction (Kerschensteiner et al., 2009 and Morgan et al.