Figures and data
Connectome and contactome of the nociceptive circuit in Drosophila larva at developmental stages L1 and L3.
(A) Illustration of the escape behavior triggered by nociceptive or damaging stimulisensation. A-anterior; P-posterior; D-dorsal; V-ventral. (B) Circular arrangement of the nociceptive dendrites (ddac, v’ada and vdaB) and their projections to LNs (A02m, A02n, A09a, A09c, A09l and A10a). Panels A-B created with BioRender.com. (C) Comparative visualization of the larva body (top), larva nervous system (middle) and nociceptive circuit (bottom) in the L1 and L3 developmental stages. A-anterior; P-posterior; L-left; R-right. (D) Postsynaptic density (PSD) segmentation pipeline. Left: dashed lines outline distinct PSD areas segmented from ssTEM images used in the analysis. Right: schematic showing cross-sectional views of a neurite, highlighting pre- and postsynaptic sites with color-coded representations of the neurite and its synaptic partners.
Connectomic mapping of synaptic size distributions across development.
(A) Schematic illustration of expected shifts in synaptic sizes under homeostatic synaptic upscaling mechanisms for the L1 (blue) and L3 (purple) developmental stage. (B) Synaptic size (PSD area, log-scaled) distributions in L1 (N=319; blue; across all connections) and L3 (N=1562; purple; across all connections) datasets fitted with gamma curves. Dashed line shows the L1 distribution after applying a multiplicative rescaling factor (s = 0.81) to all PSD area values, aligning the L1 median to that of L3 while preserving the gamma shape. (C) Box plots of synaptic area distributions (data points) for individual LN subtypes in L1 (light blue) and L3 (purple). Synaptic size remains stable across most subtypes, with only A09c and A09l exhibiting significant differences.
Evaluating traces of correlation-based plasticity across development.
(A) Schematic illustrating the ranking of mdIV neurons connected to a single LN based on synapse count. The mdIV neuron forming the most synapses is assigned rank 1 (preferential connection), followed by those with progressively fewer synapses (ranks 2, 3, etc.). (B, C) Connectivity rank analysis of mdIV-to-LN synaps es in L1 (top, B) and L3 (bottom, C). Bars show the mean number of synapses per connection pooled across all LNs by ranking (left y-axis). Black dots represent the mean PSD area for each individual connection (right y-axis). Black line shows the regression of mean PSD area versus connection rank. (D) Schematic illustrating synaptic pairs originating from the same axon and targeting the same dendritic branch. Under correlation-based plasticity, such pairs are expected to exhibit greater size similarity. (E) Correlation of PSD areas from synaptic pairs within the same dendritic branch and sharing the same pre- and post-neuron in L1 (N=42; left) and L3 (N=172; right) datasets. Black line: linear fit of observed data; gray line: linear fit of shuffled control, with 1000 shuffled versions; red line: linear fit of sorted pairs (maximum correlation). (F) Mean coefficient of variation (CV ± SD) for PSD area pairs in observed data, shuffled, and sorted controls.
Maintenance of dendritic synaptic density throughout development.
(A) Illustration of dendritic synaptic density maintenance during dendritic cable growth for the L1 (blue) and L3 (purple) developmental stage. Same color scheme is used in the remaining panels. (B) Dendritic synaptic density of LNs (one each in left and right side of the VNC) in L1 (N = 12) and L3 (N = 12). Lines connect individual LNs of L1 with their corresponding LN in L3. (C) Summed synaptic PSD areas of all the incoming mdIV synaptic connections to each LN in L1 and L3. Bars represent LNs of each specific type from the left and right sides of the VNC. (D, E) Correlation between the number of synapses (NL1 = 319; NL3 = 1562) between connected mdIV-LN neuron pairs and their summed synaptic PSD areas in L1 (Nconnection = 48) (D) and L3 (Nconnection = 64) (E).
LN dendrite and mdIV axon scaling during postembryonic development.
(A) Visualization of all LNs in L1 (left, N = 12) and L3 (right, N = 12), with two example LNs highlighted (A09l from the left and right side of the VNC), showing dendrites (green), axons (gray), and somas plus cell body fibers (black). Incoming synaptic connections to the dendritic trees are marked with blue dots. The remaining dendritic structures are shown with reduced opacity. An approximately fivefold increase in dendritic cable length and synaptic input is observed. (B) Fold change between L1 and L3 in dendritic cable length and dendritic synaptic input for all LNs (N = 12). Lines connect the cable and input ratios for each LN type. (C) Visualization of all mdIV axons in L1 (left, N = 6) and L3 (right, N = 6), with two example mdIVs highlighted (v’ada from left and right VNC), showing axon terminals (red) and remaining axonal cable (black). Outgoing synaptic connections from the axon terminals are marked with blue dots. All other axon terminals are displayed with reduced opacity. Visualization illustrates an approximately three-fold increase in axon terminal cable length and five-fold increase in axonal synaptic output. (D) Fold change between L1 and L3 in axon terminal cable length and axonal synaptic output for all mdIVs. Lines connect the cable and output ratios for each mdIV type.
Presynaptic density compensates for reduced axon-dendrite overlap to conserve connectivity during development.
(A) Schematic illustration of axon-dendritic overlap quantification between an mdIV axon and an LN dendrite. (B) Total axon-dendrite overlap length in L1 and L3 for all mdIV-to-LN connections. Bars represent LNs of each specific LN from the left and right sides of the VNC. (C) Normalized overlap, expressed as the percentage of dendritic cable in proximity to mdIV axons for each LN (100% = complete overlap). Bars same as in B. (D) Flattened topological visualizations of a representative LN (A09l, right hemisegment) in L1 and L3 using the scalable force-directed placement (SFDP) algorithm (Hu, 2005). Overlapping mdIV axons are color-coded by identity, illustrating the reduced normalized overlap between mdIV axons and the A09l LN in L3 compared to L1. (E) Axonal synaptic density for individual mdIV axons in L1 and L3; lines link matched axons across developmental stages. (F) Relationship between fold-change in axon-dendrite overlap and fold-change in connectivity for connections present at L1 and L3 (N = 48). The dashed line indicates a 1:1 isometric relationship expected if spatial overlap alone determines connectivity. The gray line shows the observed regression. Red points and fit show corrected predictions after adjusting overlap by the fold-increase in presynaptic density. (G) Schematic summary of identified circuit scaling properties in L1 (left) and L3 (right).
Synaptic density and relative connectivity conservation stabilize voltage responses across development.
(A) Schematic representation of a pair of simulated LNs (A02m) across developmental stages. Inset highlights synapse activation with the same density and relative connectivity. Colors indicate developmental stages and presynaptic mdIV neuron identity, consistent with previous figures and applied throughout this figure. (B) Steady-state voltage responses to distributed synaptic inputs in all LN morphologies from the nociceptive circuit (Ntotal = 24, L1 = 12, L3 = 12), averaged over 1000 trials. Target voltage was set to Vsyn = 8mV, to match spike threshold of VNC neurons from Günay et al. (2015). The dashed line represents the analytical prediction, fitted using the same electrophysiological parameters as the models. (C) Relative connectivity of all mdIV-to-LN connections present in L1 and L3. The black straight line represents a linear regression fit (r = 0.94, p < 0.001). (D) Simulated voltage responses from all mdIV-to-LN connections across all neurons, averaged over 1000 trials. The black straight line represents a linear regression fit (r = 0.33, p < 0.001). Color scheme distinguishes presynaptic mdIV neurons as before. (E) Left: Simulated voltage responses from all mdIV-to-LN connections plotted against their corresponding relative connectivity as in (C). Error bars represent standard deviations. The black straight line represents a linear regression fit (r2 = 0.71, p < 0.001). Right: Similar plot, but with voltage responses plotted against the absolute number of synapses for the corresponding relative connectivity (r2 = 0.38, p < 0.001). Correlation coefficients (r) were calculated using Pearson correlation; p-values were obtained from permutation tests. Linear regression lines and coefficients of determination (r2) were included where applicable.
Supplementary to Figure 2: PSD area distributions across development with fitted log-normal and gamma models.
(A,B) Synaptic size (PSD area) distributions of L1 (A) and L3 (B) datasets, both fitted by a gamma distribution (black) and a log-normal distribution (gray).
Supplementary to Figure 4: Synaptic density conservation across development.
(A) Scatter plot comparing synaptic density (synapses per μm) in L1 and L3 across different LN subtypes. Each LN type is represented by two data points, corresponding to the left and right counterparts of the VNC. Synaptic density is strongly correlated between developmental stages (r = 0.80, p < 0.01), suggesting that dendritic synaptic density is maintained despite neuronal growth. The solid line represents the linear regression fit. (B) Mean fold change (L3/L1) of summed PSD areas from each mdIV–LN connection for different LN subtypes (open circles), along with the fold change of summed PSD areas normalized by the number of synapses within each connection for the same subtypes. The dashed line shows the median 4.95 fold increase in summed PSD areas and grey line shows the median 0.88 fold change of the normalized summed areas. Normalized values remain close to 1, suggesting that the synaptic sizes per connection are stable across development.
Supplementary to Figure 5: Dendritic synaptic input increases proportionally with dendritic growth, while axonal output increases independently of axonal growth.
(A) Scatter plot showing the relationship between dendritic cable growth ratio (L3/L1) and number of synaptic inputs ratio (L3/L1) across LN subtypes. Input ratio scales proportionally with dendritic growth (r = 0.74, p < 0.01), suggesting that synaptic input is adjusted to match dendritic expansion. (B) Scatter plot of axonal cable growth ratio (L3/L1) versus number of synaptic outputs ratio (L3/L1) for mdIV neurons. Axonal output ratio does not significantly correlate with axonal growth (r = 0.35, p = 0.49). Solid lines represent linear regression fits.
Supplementary to Figure 6: Cable overlap and synaptic density comparisons between L1 and L3.
(A) Fold change in cable overlap (L3/L1) for different LN subtypes. Open circles show absolute cable overlap change, and filled circles show normalized cable overlap. Dashed and gray lines indicate the median fold change in absolute (3.36) and normalized (0.63) overlap, respectively. (B) Comparison of synaptic density between presynaptic mdIV axons and postsynaptic LN dendrites in L3. Mean mdIV density (left, red) is significantly higher than dendritic density (right, green; p < 0.05). After correcting mdIV density by the median fold change in normalized cable overlap (0.632; middle, green), adjusted densities are not significantly different from dendritic densities (p = 0.86).
Supplementary to Figure 7: Simulated voltage responses as a function of relative connectivity or absolute number of synapses in mdIV-LN connection.
(A) Simulated steady-statSimulated steady-state voltage responses in each L1 LN neuron. Left: voltage responses across L1 LN morphologies plotted against relative connectivity (percentage of total synaptic input in 2% intervals). The dashed line shows the analytical prediction from the equation (Eq.) 
