(A) (B) Images of a MU branch on two sessions (A) with on-focus high-magnification images of 4 NMJs (B) (dashed boxes in A), where one NMJ remained innervated (#1, B’), one was partially lost (#2, B’’) and two others were newly innervated by this MU through axonal sprouting (#3 and #4, B’’’; heterologous reinnervation) (Green: YFP-labeled axon; Red: nAChR). (C) Histogram showing the average proportion of NMJs from the initial pool which are innervated (green), re-innervated (purple, ‘autologous reinnervation’) or lost (red) by the MU and the proportion which are gained through expansions (blue, ‘heterologous reinnervation’). Note that the overall size of MUs increased over time despite the loss of NMJs from their initial pool. Numbers in brackets represent the number of remaining MU arbors over the total number of MU arbors observed (N = 11). Figure 4—figure supplement 1 illustrates how these opposing changes could alternate within the same MU over time. Controls showing that heterologous reinnervation (MU expansions) occurs on NMJs lost by the imaged MU are presented in Figure 4—figure supplement 2. (D) Correlation between MU expansions (blue) or asynchronous complete NMJ losses (red) and their branch order (n = 253; 19 arbors; N = 11) showing the inverse relationship between the propensity of axonal branches to expand (logistic regression, p=0.001) or to be lost (logistic regression, p=0.010). Each dot represents a data point while the solid and the dotted lines represent the logistic regression curves. Individual results for all MUs from SOD1 mice (N = 11), including graphical representations, are included in Figure 4—source data 1. The raw values for the graph in panel (D) are presented in Figure 4—source data 2. Data are presented as mean ± SEM. Scale bar, low magnification: 100 μm; high magnification: 25 μm.