Revealing synaptic nanostructure distribution through automatic dendritic spine segmentation and single-molecule localization microscopy

Relating dendritic spine morphology to synaptic organization in brain tissue is essential for understanding excitatory synaptic transmission and plasticity. Single-molecule localization microscopy (SMLM) offers the spatial precision needed to study the synaptic protein distribution at the nanoscale. However, the widefield setup required for SMLM produces diffraction-limited images with poor contrast and resolution in thick brain slices (> 30 μm), making accurate segmentation of dendritic spines challenging. To overcome this challenge, we developed an automated 3D segmentation approach tailored to this condition by combining two existing machine-learning models. We integrated this strategy with SMLM-based localization of synaptic proteins to map post-synaptic protein PSD-95 within spines at nanoscale resolution. This framework, named ISEPLA (Integrated Spine Extraction and Protein Localization Analysis), revealed a hierarchical organization of synaptic proteins: spines contain multiple nanomodules, each composed of smaller nanodomains. Larger spines contain more nanomodules, and larger nanomodules comprise more nanodomains. Therefore, our method enables precise morphological and molecular analysis under physiologically relevant imaging conditions, providing new insights into the synaptic organization in spines.