Morphological determinants of dendritic arborization neurons in Drosophila larva

Abstract

Pairing in vivo imaging and computational modeling of dendritic arborization (da) neurons from the fruit fly larva provides a unique window into neuronal growth and underlying molecular processes. We image, reconstruct, and analyze the morphology of wild-type, RNAi-silenced, and mutant da neurons. We then use local and global rule-based stochastic simulations to generate artificial arbors, and identify the parameters that statistically best approximate the real data. We observe structural homeostasis in all da classes, where an increase in size of one dendritic stem is compensated by a reduction in the other stems of the same neuron. Local rule models show that bifurcation probability is determined by branch order, while branch length depends on path distance from the soma. Global rule simulations suggest that most complex morphologies tend to be constrained by resource optimization, while simpler neuron classes privilege path distance conservation. Genetic manipulations affect both the local and global optimal parameters, demonstrating functional perturbations in growth mechanisms.

Keywords: Computational modeling; Confocal microscopy; Molecular neurogenetics; Morphological reconstructions; Neuronal development.

Fig. 1:. Reconstructions of dendritic arborization (da) neurons, superimposed on the image stacks: three wild type (WT) cell classes and three RNA-silenced phenotypes of ddaC.

(a) Class I (ddaD); (b) Class II (ldaA) (c) Class III (v’pda); (d) Class IV (ddaC); (e) RpS2-IR (ddaC): reduction in complexity; (f) SkpA-IR (ddaC): Increase in complexity; (g) Ank2-IR (ddaC): Overall similar to WT ddaC, with moderate reduction in branch numbers close to the soma and tufted branching at distal dendritic terminals. All scale bars are 100 μm.

Fig. 2:. Quantitative morphometry of the four wild type da neuron cell classes and of the genetically altered phenotypes of Class IV subtype ddaC.

(a) Four da neuron cell classes are compared by the distribution of dendritic length against path distance from soma. Inset: comparison of all Class IV subtypes, with the blue dotted line indicating the ddaC data newly reconstructed for this study. (b) Comparison of WT ddaC with all RNAi silenced (IR) ddaC phenotypes analyzed in the study. Inset 1 (green arrow): three IR ddaC phenotypes that are only slightly less complex than WT ddaC, but distally shifted. Inset 2 (purple arrow): three IR phenotypes with significantly reduced arbor complexity.

Fig. 3:. Morphological homeostasis in da neurons.

(a) Reconstruction of a Class I (ddaD) neuron, with two individual stems (green and purple) of slightly different sizes, emerging from the soma. (b) Average length and average number of terminals for each dendritic stem as a function of the total number of stems in individual Class I neurons. (c) Distribution of standard deviations of the total length of artificial Class I cell groups, created by shuffling individual trees among real neurons. Trees were only shuffled among neurons with the same number of stems, and then pooled together and randomly grouped. The red bar is the standard deviation of total dendritic length in real Class I neurons. (d) Reconstruction of a Class IV (ddaC) neuron, with four individual stems (green, purple, sky-blue, and red) of different complexities, emerging from the soma. (e) Same as (b) for Class IV neurons. (f) Same as (c) for Class IV neurons.

Fig. 4:. Generation of artificial da neurons using local rule-based simulation.

(a) Distribution of the number of tips against branch order for real Class I neurons (blue line) and for artificial Class I neurons (red line) generated using the optimal local model variant. (b) Distribution of dendritic length against path distance from soma, for real Class I neurons (blue) and for artificial Class I neurons (red). (c) Dendrogram of a real Class I neuron. (d) Dendrogram of a simulated Class I neuron. (e) Same as (a) for Class IV neurons. (f) Same as (b) for Class IV neurons. (g) Dendrogram of a real Class IV neuron. (h) Dendrogram of a simulated Class IV neuron. The local model variant names in this figure and for all following figures describe the index of the model variant, followed by the series of FDs that constrained the five BPs (taper rate, daughter ratio, parent daughter ratio, branch path length, bifurcation probability). PD: path distance from soma; RD: radius; BO: branch order. The terminal spacing for all dendrograms in this and subsequent figures is 1 μm; thus, X-axis approximates the total number of terminals. The dendrogram Y-axis represents path distance from soma.

Fig. 5:. Generation of artificial da neurons using global rule-based simulation.

(a) Example of a real Class I da neuron. (b) Example of an artificial Class I neuron generated using a balancing factor (bf) of 0.6. (c) Distribution of number of tips against branch order and (d) of dendritic length against path distance in real Class I neurons (blue) and artificial Class I neurons generated with a bf of 0.6 (red). (e) Example of real Class IV da neuron. (f) Example of an artificial Class IV neuron generated with a balancing factor of 0.3. (g) Distribution of the number of tips against branch order and (h) of dendritic length against path distance in real Class IV neurons (blue) and artificial Class IV neurons generated with a bf of 0.3 (red).

Fig. 6:. Comparing the morphology of two RNA-silenced ddaC phenotypes (RpS2-IR: reduced complexity; SkpA-IR: increased complexity) using local and global rule-based simulations.

(a) Example of a real RpS2-IR neuron. (b) Example of an artificial RpS2-IR neuron generated at the balancing factor of 0.2. (c) Dendrogram of a real RpS2-IR neuron. (d) Dendrogram of an artificial RpS2-IR neuron generated using the optimal local parameters (204_BO_RD_RD_RD_BO). (e) Distribution of tips against branch order and (f) of length against path distance for real RpS2-IR (blue), and optimum-local (red) and optimum-global (green) artificial RpS2-IR neuron groups. (g) Example of a real SkpA-IR neuron. (h) Example of an artificial SkpA-IR neuron generated at the balancing factor of 0.3. (i) Dendrogram of real SkpA-IR neuron. (j) Dendrogram of an artificial SkpA-IR neuron generated using the optimal local parameters (27_PD_PD_BO_BO_BO). (k) Distribution of tips against branch order and (l) of length against path distance from soma for real SkpA-IR (blue), and optimum-local (red) and optimum-global (green) artificial SkpA-IR neuron groups.