Background

Reconstructing the morphology of neurons at the single-cell level provides new insights into various neuroscience issues, including neural cell classification, elucidation of information transmission among different neuron types, and clarification of the potential functions of distinct brain regions [1–3]. Due to the diversity and complexity of neuron morphology, establishing an efficient procedure for single-neuron reconstruction that ensures both the integrity and accuracy of the results is essential for obtaining large-scale single-neuron reconstruction data and conducting subsequent analysis [4]. In recent years, advancements in sparse labeling techniques [5,6] and advanced imaging systems, such as fluorescence micro-optical sectioning tomography (fMOST) [7], have enabled the acquisition of continuous section imaging results for labeled neurons. However, there is still a need for a systematic approach and optimization strategies for reconstructing single-neuron morphologies across the entire brain from raw imaging data. In this protocol, we first preprocess the fMOST imaging results and establish quality control standards to assist researchers in selecting high-quality imaging samples suitable for continuous tracing of single-neuron morphologies across the entire brain. Next, we developed a neuron soma localization tool that allows for selective tracing within the region of interest, mitigating interference from labeled neurons in other regions—an ongoing challenge in dense labeling approaches [8,9]. For the successfully traced neurons, by using the semi-automated tracing software Fast Neurite Tracer (FNT) [10], we significantly enhanced the accuracy of the tracing results by increasing quality control steps. Additionally, we distinguished labeled axonal and dendritic structures, enabling the analysis of morphological relationships among various neuronal structures. As a final step in our reconstruction pipeline, we enhanced the spatial accuracy of the reconstructed neurons by implementing a registration method that utilizes features extracted from multiple brain subregions. The single-neuron reconstruction results obtained through these methods can be directly applied to data visualization and analysis. This procedure has been utilized in studies of single-neuron projectomes in multiple mouse brain regions, building the currently largest dataset of whole-brain projectome maps for 10,100 single neurons in the mouse hippocampus [11,12]. It provides a foundation for future cross-species brain mapping research, including non-human primate studies, and offers guidance for research on brain diseases and brain-inspired artificial intelligence.