Mapping the Architecture of Ferret Brains at Single-Cell Resolution

Ben Long, Tao Jiang, Jianmin Zhang, Siqi Chen, Xueyan Jia, Xiaofeng Xu, Qingming Luo, Hui Gong, Anan Li, Xiangning LiFrontiers in Neuroscience 2020 · DOI

TL;DR

Researchers have developed a novel pipeline for creating high-resolution, whole-brain cytoarchitectonic atlases of large brains, overcoming previous limitations in staining and imaging. This method was applied to the ferret brain, revealing detailed information about giant pyramidal neurons, including their morphology, neurochemical properties, and distribution, which could advance our understanding of mammalian brain organization.

What Problem Does This Paper Address?

Mapping the intricate cellular architecture (cytoarchitecture) of the brain is crucial for understanding how neural systems are organized and function. However, creating detailed, whole-brain maps at a single-cell level has been a significant challenge, particularly for larger, more complex brains like those of ferrets or primates. These "gyrencephalic" brains, characterized by folds (gyri) and grooves (sulci), present substantial hurdles in tissue processing, uniform staining, and high-resolution imaging. Traditional methods often struggle with deep tissue penetration and consistent staining across large volumes, leading to low-resolution data that obscures fine cellular details. This paper aims to bridge this gap by developing an integrated approach to generate a comprehensive, single-cell resolution cytoarchitectonic atlas of a large mammalian brain.

Methods

The authors introduce a multi-stage pipeline designed to overcome the challenges of large-brain cytoarchitectonic mapping:

  • Modified En-Bloc Nissl Staining: A key innovation is a modified Nissl staining protocol adapted for large-volume, intact ferret brains. This involves immersing the brains in a 2.5% thionine solution for an extended period (60 days) in a slightly acidic environment (pH 5.0), followed by thorough rinsing. This extended protocol ensures uniform staining throughout the entire large tissue volume, which is critical for subsequent imaging.
  • Resin Embedding: Following staining, the brains are embedded in Spurr resin. This process, involving a graded series of ethanol and acetone dehydration and resin infiltration, prepares the tissue for precise mechanical sectioning at the micron level.
  • Whole-Brain Imaging (MOST System): The embedded brains are then imaged using a Micro-Optical Sectioning Tomography (MOST) system. This advanced technique allows for continuous, high-resolution (0.33 µm x 0.33 µm x 1 µm voxel resolution) scanning of the entire brain as it is sectioned. The system captures terabytes of data, generating a detailed 3D representation.
  • Image Preprocessing: The massive raw image data undergoes significant preprocessing, including stitching of individual image strips, removal of periodic noise, and calibration for non-uniform illumination. This ensures data consistency and prepares it for analysis.
  • Automated Soma Segmentation (3D U-Net): To identify and analyze individual neurons, a deep learning approach using a 3D U-Net convolutional neural network is employed. This model is trained on manually segmented ferret brain cells and is capable of automatically segmenting neuronal cell bodies (somata) at single-cell resolution.
  • Data Analysis and Visualization: The segmented data is then analyzed using specialized software (NeuroGPS, Imaris, Amira) to quantify cellular features, map their distribution, and visualize the entire brain architecture in 3D.

Datasets

  • Training Data: Manually segmented giant pyramidal neurons (n = 286 cells) from the ferret brain were used to train the 3D U-Net deep learning model for soma segmentation.
  • Evaluation Data: The study utilized whole-brain datasets from three adult female ferrets. The primary dataset for the atlas construction comprised two brains processed for Nissl staining and whole-brain imaging, resulting in over 11,000 horizontal sections and exceeding 22.0 terabytes of data. A third ferret brain was used for immunolabeling to characterize neurochemical phenotypes.

Key Insights and Ideas

  • Overcoming Scale Limitations: The core insight is the successful adaptation of established techniques (Nissl staining, high-resolution imaging, deep learning) to address the unique challenges posed by large-volume brains, demonstrating that single-cell resolution mapping is achievable even for complex structures.
  • The Ferret Brain as a Model: The ferret, with its gyrencephalic cortex similar to primates but more manageable than human or non-human primate brains, serves as an excellent model organism for developing and validating this large-brain mapping pipeline.
  • Discovery of Giant Pyramidal Neurons (Betz Cells): The study provides the first detailed report of giant pyramidal neurons (Betz cells) in the ferret brain, identifying their presence, morphological diversity, and specific neurochemical markers. This expands our knowledge of these specialized neurons beyond rodents and primates.
  • Morphological Diversity of Betz Cells: The research reveals that ferret Betz cells exhibit distinct soma shapes (pyramidal, spindle, flat, spheroid), suggesting functional specialization within this neuronal population.
  • Neurochemical Identity: The study confirms that ferret Betz cells are excitatory pyramidal neurons, expressing NeuN (neuronal marker) and EAAC1 (excitatory amino acid transporter) while being GAD67-negative (GABAergic neuron marker).

Key Results

  • Pipeline Efficacy: The developed pipeline successfully generated a whole-brain cytoarchitectonic atlas of the ferret brain at a voxel resolution of 0.33 µm x 0.33 µm x 1 µm.
  • Giant PyN Identification: Giant pyramidal neurons were identified in layer Vb of the ferret motor cortex.
  • Quantitative Differences: Giant PyNs had a significantly larger mean cell volume (3101.31 ± 106.23 µm³) compared to other PyNs (1004.32 ± 13.49 µm³).
  • Soma Shape Diversity: Four distinct soma shapes were identified for Betz cells: pyramidal, spindle, flat, and spheroid, with spindle-shaped cells having the largest mean volume.
  • Distribution Mapping: The study mapped the brain-wide distribution of approximately 2.6 × 10⁴ giant PyNs, primarily located in the posterior sigmoid gyrus (PSG) motor cortex.
  • Localization Accuracy: The NeuroGPS algorithm demonstrated high accuracy for automatic giant PyN localization, with average recall of 87.36% and precision of 94.47%.

Limitations and Future Work

  • Single Modality: The current pipeline relies solely on Nissl staining. The authors acknowledge that this method may not fully substitute traditional section-based Nissl staining for high-quality visualization in all cases, particularly due to potential background staining from alcohol differentiation. Future work could integrate multiple modalities (e.g., myelin, tract-tracing) for a more comprehensive atlas.
  • Cellular Resolution vs. Inter-neuronal Correlation: While the pipeline achieves single-cell resolution for soma segmentation, the authors note that correlating neuronal distribution patterns with precise brain area differences in three dimensions remains a challenge not fully addressed in this study.
  • Segmentation of Small/Dense Cells: The current soma segmentation methods are primarily effective for large neurons like giant PyNs. Further optimization is needed to accurately segment smaller or more densely packed cells.
  • Broader Applicability: The authors suggest that this pipeline, with potential optimizations, could be extended to other large-scale brains, including those of tree shrews, common marmosets, and even non-human primates, facilitating comparative neuroscience research.

Summary updated

Abstract

Mapping the cytoarchitecture of the whole brain can reveal the organizational logic of neural systems. However, this remains a significant challenge, especially for gyrencephalic brains with a large volume. Here we propose an integrated pipeline for generating a cytoarchitectonic atlas with single-cell resolution of the whole brain. To analyze a large-volume brain, we used a modified en-bloc Nissl staining protocol to achieve uniform staining of large-scale brain specimens from ferret (<i>Mustela putorius furo</i>). By combining whole-brain imaging and big data processing, we established strategies for parsing cytoarchitectural information at a voxel resolution of 0.33 μm × 0.33 μm × 1 μm and terabyte-scale data analysis. Using the cytoarchitectonic datasets for adult ferret brain, we identified giant pyramidal neurons in ferret brains and provide the first report of their morphological diversity, neurochemical phenotype, and distribution patterns in the whole brain in three dimensions. This pipeline will facilitate studies on the organization and development of the mammalian brains, from that of rodents to the gyrencephalic brains of ferret and even primates.
BibTeX
@inproceedings{long2020mapping,
  author = {Ben Long and Tao Jiang and Jianmin Zhang and Siqi Chen and Xueyan Jia and Xiaofeng Xu and Qingming Luo and Hui Gong and Anan Li and Xiangning Li},
  title = {{Mapping the Architecture of Ferret Brains at Single-Cell Resolution}},
  booktitle = {Frontiers in Neuroscience},
  year = {2020},
  doi = {10.3389/fnins.2020.00322},
}
Figure