CNS*2025 Florence

Non-homogeneous axonal bouton distributions constrain whole-brain single-cell network topology

Penghao Qian*¹, Linus Manubens-Gil*², Shengdian Jiang², Hanchuan Peng^1,3,†
1 New Cornerstone Science Laboratory, Institute for Brain and Intelligence, Fudan University, Shanghai, China.
² New Cornerstone Science Laboratory, SEU-ALLEN Joint Center, Institute for Brain and Intelligence, Southeast University, Nanjing, China
³ Shanghai Academy of Natural Sciences (SANS), Fudan University, Shanghai, China.

Presenting author: linusmg@seu.edu.cn
^† Corresponding author: h@braintell.org
* Contributed equally

Introduction
We asked whether axonal bouton distributions, often assumed to be uniform, influence brain-wide wiring. Using 1,891 fully reconstructed mouse neurons [1] with putative bouton annotations [2], we investigated how bouton placement varies across brain regions and cell types, and how this variability impacts the topology of structural networks at single-cell resolution [3]. We further tested how morphological perturbations (on axons, dendrites, or both) affect large-scale wiring, hub structure, and cost-efficiency tradeoffs.
Methods
We divided the mouse brain into 30 µm cubes and computed putative synaptic connectivity based on co-localized boutons and dendrites. We weighted connection strengths by bouton counts and dendritic coverage to construct a whole-brain single-cell connectivity matrix. We compared this matrix to random, small-world and scale-free networks as well as a network obtained using uniform bouton placement throughout axons. We computed degree distributions, triad motifs, community structure, and connector hub scores. To explore the impact of morphological details on circuit topology, we perturbed neuronal morphology (scaling, pruning, or deleting axonal/dendritic trees) and recalculated network metrics.
Results
We found bouton distributions were not homogeneous and varied systematically by cell type and region. VPM thalamic neurons placed boutons distally, while somatosensory cortical neurons showed flatter distributions. Networks built from single neuron data showed scale-free-like degree distributions, and both modularity and triad composition differed significantly when comparing putative bouton distributions to uniform bouton distribution networks. Perturbing dendritic or axonal trees reshaped network topology. Tree span and bouton density emerged as primary drivers. Dendritic pruning preserved connectivity better than equivalent axonal loss, due to convergence across axons.
Discussion
We showed that bouton placement is not random and critically shapes wiring at the single-cell level. Morphological perturbations (especially to tree span) translate into large-scale network changes. Our results highlight the need to include bouton-level structure in mesoscale brain models and suggest that morphology disruptions seen in disease can imply substantial topological changes.





We thank Zhixi Yun, Feng Xiong, and Lijun Wang for comments on the figures. This work was mainly supported by a Southeast University (SEU) initiative of neuroscience awarded to H.P. H.P. was also supported by a Zhejiang Lab BioBit Program visiting grant (2022BCF07).

1. https://doi.org/10.1038/s41586-021-03941-1
2. https://doi.org/10.1007/s12021-022-09569-4
3. https://doi.org/10.1016/j.celrep.2024.113871