Autophagy-dependent filopodial kinetics restrict synaptic partner choice during Drosophila brain wiring

Abstract

Brain wiring is remarkably precise, yet most neurons readily form synapses with incorrect partners when given the opportunity. Dynamic axon-dendritic positioning can restrict synaptogenic encounters, but the spatiotemporal interaction kinetics and their regulation remain essentially unknown inside developing brains. Here we show that the kinetics of axonal filopodia restrict synapse formation and partner choice for neurons that are not otherwise prevented from making incorrect synapses. Using 4D imaging in developing Drosophila brains, we show that filopodial kinetics are regulated by autophagy, a prevalent degradation mechanism whose role in brain development remains poorly understood. With surprising specificity, autophagosomes form in synaptogenic filopodia, followed by filopodial collapse. Altered autophagic degradation of synaptic building material quantitatively regulates synapse formation as shown by computational modeling and genetic experiments. Increased filopodial stability enables incorrect synaptic partnerships. Hence, filopodial autophagy restricts inappropriate partner choice through a process of kinetic exclusion that critically contributes to wiring specificity.

Fig. 1. Autophagy deficiency in Drosophila photoreceptors leads to increased neurotransmission and visual attention.

a, b Newly hatched (0-day-old) genetic mosaic flies with autophagy-deficient (atg6 and atg7 mutants) photoreceptors exhibit normal eye morphology (a) and axonal projections in the optic lobe (b). Repeated three times independently. c Representative electroretinogram (ERG) traces. Repeated three times independently. d, e Quantification of ERG depolarization (d) and on-transient (e) amplitudes relative to control. Rescue of atg6 mutant photoreceptors with GMR > atg6 expression leads to overcompensation and increased autophagy (see Supplementary Fig. 1). n = 20 flies per condition. Two-tailed unpaired t-test with Welch’s correction; *p < 0.05, **p < 0.01, ***p < 0.001. Error bars denote mean ± SEM. f Buridan’s paradigm arena to measure object orientation response of adult flies, with two black stripes positioned opposite to each other as visual cues. g The parameter “stripe deviation”’ measures how much a fly deviates from a straight path between the black stripes in the arena. h Stripe fixation behavior of adult flies with atg6 mutant photoreceptors, photoreceptors with upregulated autophagy (atg6, GMR > Atg6), and their genetically matched controls are shown on the population level (heatmap) and as individual tracks. Flies with atg6 mutant photoreceptors show reduced stripe deviation, whereas increased autophagy (atg6, GMR > Atg6) leads to increased stripe deviation. i Quantification of stripe deviation. The error bars indicate the 25th percentile, the boxed area the 75th percentile, and the middle line of the boxplots indicates the median. n = 60 flies per condition, two-way ANOVA and Tukey’s HSD as post-hoc test; ***p < 0.001. Source data are provided as a Source Data file.

Fig. 2. Autophagy-deficient Drosophila photoreceptors form supernumerary synapses.

a–e’ Representative images of R1–R6 and R7 photoreceptor axon terminals with Brp^short-GFP marked active zones in wild-type (a, a’), atg7 mutant (b, b’), atg6 mutant (c, c’), atg18 mutant (d, d’), and atg6, GMR > Atg6 (e, e’). R7 axon terminals are shown from distal (top) to proximal (bottom) medulla. Relative thicknesses of medulla layers are shown in Control R7 terminals panel (a) along R7 axon terminals. Red boxes show supernumerary synapses in loss of autophagy at distal part of R7 axon terminals. Repeated five to ten times independently with similar results. f, g Number of Brp puncta per terminal in R1-R6 (f) and R7 (g) photoreceptors. n = 40 terminals per condition. Kruskal–Wallis and Dunn’s as post-hoc test; **p < 0.01, ***p < 0.001, ****p < 0.0001. Error bars denote mean ± SEM. h Number of Brp puncta in distinct medulla layers along R7 axon terminals (see Methods for the definition of medulla layers and a for relative thicknesses of medulla layers). n = 22 terminals for control, n = 30 terminals for atg7, n = 27 terminals for atg6. Kruskal–Wallis and Dunn’s as post-hoc test; *p < 0.05, ***p < 0.001, ****p < 0.0001. Error bars denote mean ± SEM. Source data are provided as a Source Data file.

Fig. 3. Loss of autophagy leads to synaptic connections with aberrant neuronal partners.

a–c Neurons postsynaptic to control (a), atg6 mutant (b), and atg18 mutant (c) R7s are labeled with trans-Tango (see Methods for full genotypes; magenta = postsynaptic neurons, green = CadN, Me = medulla, Lo = lobula, Lop = Lobula plate). Arrowheads show postsynaptic neurons labeled for autophagy-deficient R7s but not for control R7s. Repeated three to five times independently with similar results. d Number of postsynaptic neurons per optic lobe for control, atg6 mutant, and atg18 mutant R7s based on trans-Tango-labeled cell body counts. n = 10 optic lobes per condition. One-way ANOVA and Tukey’s HSD as post-hoc test **p < 0.01, ***p < 0.001. Error bars denote mean ± SEM. e Examples of aberrant neuronal partners of autophagy-deficient R7s, with individual neurons pseudo-colored in white. f Schematic of dendritic and axonal arborization of aberrant neuronal partners (redrawn and adapted based on Golgi impregnations from Fischbach and Dittrich). g Number of each aberrant neuronal partners per optic lobe from 1-week-old fly brains. Note that only ∼10% of R7s are mutant for atg6 and trans-Tango labeling is dependent on synaptic strength between partners and progressively increase through age. See Methods for detailed Drosophila genotypes used to perform trans-Tango experiments. Source data are provided as a Source Data file.

Fig. 4. Synaptic connections between autophagy-deficient R7s and aberrant postsynaptic partners are functional based on activity-dependent GRASP.

a–c’ Activity-dependent GRASP between control R7s and Mi1s (a, a’), C2s (b, b’), and Mi4s (c, c’) show that wild-type R7s very rarely form synaptic connections, if any, with Mi1, C2, and Mi4 neurons. d–f’ Activity-dependent GRASP between atg6 mutant R7s and Mi1s (d, d’), C2s (e, e’), and Mi4s (f, f’) show widespread active synaptic connections between autophagy-deficient R7s and aberrant postsynaptic partners. g, i’, Activity-dependent GRASP between atg18 mutant R7s and Mi1s (g, g’), C2s (h, h’), and Mi4s (i, i’) show less frequent active synaptic connections compared with atg6 mutants. Note that Atg18 loss-of-function does not block autophagosome formation as effective as Atg6 loss-of-function (see Supplementary Fig. 1). Regions inside yellow rectangles are shown in close-up images as single greyscale GRASP channels. See Methods for Mi1, Mi4, and C2-specific LexA drivers and detailed Drosophila genotypes used to perform GRASP experiments. Repeated three times independently with similar results.

Fig. 5. Autophagy regulates the stability of synaptogenic filopodia at axon terminals.

a Live imaging of GFP-Atg5-expressing R7 axon terminals in intact, developing Drosophila brain shows formation of autophagosomes at the bulbous tips of synaptogenic filopodia followed by the collapse of filopodia (P + 60%). Repeated three times independently with similar results. b–e Live imaging of R7 axon terminals at P + 60% (during synaptogenesis) revealed increased stability of synaptogenic filopodia in autophagy-deficient R7 terminals (c, d) and decreased stability in R7 terminals with upregulated autophagy (e) compared with control (b). Yellow arrowheads: stable synaptogenic filopodia; white arrowheads: unstable bulbous tip filopodia. Repeated five to ten times independently with similar results. f Number of concurrently existing bulbous tip filopodia per R7 axon terminal per time instance. g Total number of synaptogenic filopodia per R7 axon terminal per hour. Autophagy-deficient R7 terminals exhibit significantly more stable synaptogenic filopodia (>40 min), whereas upregulated autophagy leads to filopodia destabilization. n = 8 terminals per condition. One-way ANOVA and Tukey’s HSD as post-hoc test; *p < 0.05, **p < 0.01, ***p < 0.001. Error bars denote mean ± SEM. Source data are provided as a Source Data file.

Fig. 6. Loss of autophagy increases the number of synaptogenic filopodia through defective synaptic seeding factor degradation, leading to increased synapse formation throughout development.

a–c Quantification of filopodia numbers (a), synaptogenic filopodia numbers (b), and Brp puncta numbers (c) during synaptogenesis (P40–P90) per R7 axon terminal based on fixed data. n = 40 terminals per condition. d–f Markov State Model simulation based on data in (a) and live data at P + 60% (Fig. 5) for filopodia numbers (d), synaptogenic filopodia numbers (e), and Brp puncta numbers per R7 axon terminal (f). g The mechanistic model: accumulation of synaptic seeding factors stabilizes synaptogenic filopodia; autophagic degradation of synaptic seeding factors destabilizes filopodia. h Measured (solid bars) and simulated (striped bars) synaptogenic filopodia numbers at P + 60% (the simulated data are based on synaptic seeding factor availability, see Supplementary Fig. 6). n = 8 axon terminals from independent live-imaging sessions. i Representative images of synaptic seeding factors (Syd-1 and Liprin-α) localizing to synaptogenic filopodia. Repeated three times independently with similar results. j, k Quantifications of the number of Liprin-α (j) and Syd-1 (k) positive synaptogenic filopodia. n = 30 terminals per condition. Kruskal–Wallis and Dunn’s as post-hoc test; *p < 0.05, ***p < 0.001. Error bars denote mean ± SEM. Source data are provided as a Source Data file.

Fig. 7. Loss of autophagy recruits incorrect synaptic partners by lowering an axon terminal-wide threshold for kinetic restriction of synapse formation.

a Representative R7 axon terminals at P + 70% with medulla layer information. Note that the edge of medulla (M0) is defined as 0 and the end of M6 layer is defined as 100 to calculate relative positions of all filopodia and bulbous tip filopodia, and distributed to medulla layers (M1–M6) using the relative thickness of medulla layers defined by Fischbach and Dittrich. Repeated five to ten times independently with similar results. b–d Relative frequency (solid lines) and absolute numbers (dotted lines) of all filopodia at P + 70% (b), synaptogenic filopodia at P + 70% (c), and synapses at 0-day-old adult (d). M1–M6 denote medulla layers. n = 40 terminals per condition. e Model: loss of autophagy during synaptogenesis increases the probability distribution (yellow area) compared with wild-type (gray area) of forming connections with postsynaptic partners through increased filopodial stability. Note that cells with projections at medulla layers where R7s form most of their synapses (Mi1, Mi4, C2, C3) incorrectly synapse with R7s, with higher probability than the cells with projections at medulla layers where R7s form a few, if any, synapses (Mi8, Tm1) (see Fig. 3e, f). Redrawn and adapted based on Golgi impregnations from Fischbach and Dittrich. Source data are provided as a Source Data file.