a, b show projections of two-photon (2P−) laser z-stacks of an injured M-axon one, four, and seven days after either distal (a) or proximal (b) injury. Outlines of the M-cell and its axon are shown on the left. Systemic injuries were set using a broken glass microelectrode (top panels). Rectangular boxes and zoomed in images below highlight differences in axonal regrowth after distal and proximal injury. Asterisks indicate most caudal end of the axon. Orientation is shown in b (c = cranial; l = lateral; d = dorsal). Scale bar = 100 μm. a Example of a distally injured M-axon. Left axon remained intact, the right axon was injured but did not grow past the injury site. b Example of a proximally injured M-axon. The injured axon showed robust regeneration through the injury site. c, d show examples of aberrant regeneration after distal (c) and proximal (d) systemic spinal cord injury. Scale bar = 100 μm. c After distal injury the axon changed direction (U-bend, see arrow) and regenerated cranially. d After proximal injury aberantly regrown axons always pass the site of injury and grow in the appropriate direction, but occasionally switch side (arrow). Note that the aberrant axon still shows robust regeneration. e Quantification of axon regeneration over time. Axon length was monitored one, two, three, four and seven dpi after distal (grey) or proximal (blue) injury. Data are shown as mean ± SEM. Nonlinear sigmoidal fits (logistic function) are shown. Distal: R² = 0.24; N = 28; relative steepness s = 0.7; maximal slope at x₀ = 3.1 days. Proximal: R² = 0.73; N = 24; s = 0.58; x₀ = 3.7 days. f Axon length of siblings after systemic injury 1365 ± 38 μm and 1760 ± 38 μm from the soma either directly post-injury (0 dpi) or 4 dpi (N = 4). Significant difference is highlighted with an asterisk (p = 0.04, Mann–Whitney-U test) between axon lengths at 4 dpi.

a Linear regression between median angular speed (i.e., bending angle divided by bending duration) and axon length 1 day before escapes were recorded. While escape latency had been completely restored to preinjury levels when axons had regenerated to the state marked as ‘long axon’, median angular speed of the escapes stayed constant, i.e., the slope of the regression line does not differ from zero (P = 0.88; n = 68 stages in N = 16 larvae). b Boxplots show angular speed before injury and when the axons had regrown to short (<1200 μm) or long axon length (>1900 μm). Circles show medians of angular speed for each axon length. Note absence of significant differences (highlighted with different letters; Kruskal–Wallis with Dunn’s multiple comparison post-hoc test) in angular speed of the different states during axon regrowth and difference at both stages from the angular speed level seen before injury (P < 0.0001). Data were obtained from the same escape responses (and same regeneration states) as in Fig. 7 that showed complete restoration of escape latency.