Genetic Single Neuron Anatomy reveals fine granularity of cortical interneuron subtypes

Xiaojun Wang, Jason Tucciarone, Siqi Jiang, F Yin, Bor‐Shuen Wang, Dingkang Wang, Jia Yao, Xueyan Jia, Yuxin Li, Tao Yang, Zhengchao Xu, Masood A. Akram, Yusu Wang, Shaoqun Zeng, Giorgio A. Ascoli, Partha P. Mitra, Hui Gong, Qingming Luo, Z. Josh HuangbioRxiv 2017 · DOI

TL;DR

This paper introduces a new platform, "genetic Single Neuron Anatomy" (gSNA), that combines genetic labeling with high-resolution, whole-brain imaging to reconstruct and analyze individual neuron morphologies in the mouse brain. They used this platform to reveal that a specific type of inhibitory neuron, axo-axonic cells (AACs), are not a single type but rather a collection of fine-grained subtypes with distinct anatomical features that likely dictate their specific roles in neural circuits.

What Problem Does This Paper Address?

Understanding how neural circuits are wired and function requires classifying the diverse types of neurons that make them up. While morphology (the shape and structure of neurons) has long been a key feature for classification, the sheer complexity and variability of neuronal shapes in the mammalian brain make it difficult to define discrete cell types. This "continuum" of morphologies makes it challenging to resolve the appropriate level of detail needed to understand circuit organization. Existing methods often struggle with specific, sparse, and complete labeling of individual neurons, high-resolution imaging of entire brains, and accurate reconstruction and analysis of complex neuronal structures. This paper aims to overcome these challenges to enable a more granular understanding of neuronal diversity.

Methods

The core of this work is the genetic Single Neuron Anatomy (gSNA) platform, which integrates several key technologies:

  • Genetic and Viral Labeling: A sophisticated system using Nkx2.1-CreER mice crossed with Flp reporter mice, combined with Flp-dependent AAV vectors, allows for specific, sparse, and complete labeling of individual neurons. This strategy allows for precise targeting of specific cell populations.
  • Whole-Brain High-Resolution Imaging: The study utilizes dual-color fluorescence Micro-Optic Sectioning Tomography (dfMOST). This technique enables simultaneous imaging of labeled neurons (GFP channel) and brain cytoarchitecture (propidium iodide, PI channel) at submicron resolution across the entire mouse brain. The PI channel acts as a self-registered atlas, crucial for spatial registration and identifying cortical layers.
  • Neuron Reconstruction: A multi-step process involving resin embedding, ultra-thin sectioning, and automated image processing (using Amira software) is used to reconstruct the complete morphology of individual neurons from the massive image datasets.
  • Computational Analysis: A pipeline is employed to register the reconstructed morphologies within a standardized spatial coordinate system, allowing for quantitative analysis of features like dendritic and axonal arborization patterns, and their distribution across cortical layers. This includes specialized analyses like length density analysis and hierarchical clustering to identify subtypes.

Datasets

The study utilized whole-brain imaging datasets from genetically modified mice. Specifically, they reconstructed and analyzed 62 axo-axonic cells (AACs) from the medial prefrontal cortex (mPFC), primary motor cortex (MC), and whisker-barrel somatosensory cortex (SSC) of mice. No new benchmarks were introduced, but the study established a new dataset of complete AAC reconstructions.

Key Insights and Ideas

  • AACs are not monolithic: The most significant finding is that axo-axonic cells (AACs), previously considered a single type of interneuron, are actually composed of multiple distinct subtypes. This challenges the existing classification and suggests a more nuanced view of neuronal diversity.
  • Morphology as a proxy for connectivity: The study strongly emphasizes that detailed morphological analysis, when registered within the brain's anatomical framework, provides crucial insights into a neuron's input-output connectivity. The precise alignment of AAC axon terminals with pyramidal neuron axon initial segments (AIS) makes them an ideal model for this approach.
  • Laminar organization is key: The distribution of AAC dendrites and axons across cortical layers is not random. This laminar stratification appears to be directly related to which input streams they receive and which pyramidal neuron populations they innervate, suggesting functional specialization based on location.
  • A platform for future discovery: The gSNA platform itself is a major contribution, providing a robust and scalable method for high-throughput single neuron anatomy. This opens the door for similar detailed analyses of other neuron types and for building comprehensive atlases of neuronal connectivity.

Key Results

  • The study successfully reconstructed 62 complete AACs from three cortical areas, representing the first comprehensive dataset of its kind.
  • AACs were found to be distributed across multiple cortical layers, with a significant proportion in the supragranular layers (L2/1 border) and infragranular layers (L5 and L6).
  • Hierarchical clustering and clique analysis revealed at least 8 distinct AAC subtypes based on their laminar distribution of dendrites and axons, and their cell body positions.
  • These subtypes showed distinct patterns of dendritic and axonal arborization, with some exhibiting translaminar projections, suggesting differential connectivity.
  • Analysis of axon terminal geometry revealed correlations with the postsynaptic AIS orientation, further supporting the idea that AAC subtypes are specialized for specific targets.

Limitations and Future Work

  • Reconstruction bottleneck: The authors acknowledge that manual reconstruction of neurons is time-consuming and a bottleneck for scaling up the gSNA platform. They suggest that future innovations in machine learning-based automatic reconstruction algorithms will be crucial.
  • Saturation screening: While the approach demonstrated unprecedented specificity, the study did not achieve "saturation screening" for all AAC subtypes, meaning some rare subtypes might have been missed due to limited sample size.
  • Subtype validation: The division of AACs into subtypes based on anatomy needs to be substantiated by orthogonal datasets, including physiological connectivity, gene expression profiles, and developmental genetic basis. The authors suggest future work should integrate these approaches.
  • Broader application: The gSNA framework is expected to be applicable to projection neurons as well, enabling similar detailed analyses of their connectivity and diversity.

Summary updated

Abstract

Abstract Parsing diverse nerve cells into biological types is necessary for understanding neural circuit organization. Morphology is an intuitive criterion for neuronal classification and a proxy of connectivity, but morphological diversity and variability often preclude resolving the granularity of discrete cell groups from population continuum. Combining genetic labeling with high-resolution, large volume light microscopy, we established a platform of genetic single neuron anatomy that resolves, registers and quantifies complete neuron morphologies in the mouse brain. We discovered that cortical axo-axonic cells (AACs), a cardinal GABAergic interneuron type that controls pyramidal neuron (PyN) spiking at axon initial segment, consist of multiple subtypes distinguished by laminar position, dendritic and axonal arborization patterns. Whereas the laminar arrangements of AAC dendrites reflect differential recruitment by input streams, the laminar distribution and local geometry of AAC axons enable differential innervation of PyN ensembles. Therefore, interneuron types likely consist of fine-grained subtypes with distinct input-output connectivity patterns.
BibTeX
@misc{wang2017genetic,
  author = {Xiaojun Wang and Jason Tucciarone and Siqi Jiang and F Yin and Bor‐Shuen Wang and Dingkang Wang and Jia Yao and Xueyan Jia and Yuxin Li and Tao Yang and Zhengchao Xu and Masood A. Akram and Yusu Wang and Shaoqun Zeng and Giorgio A. Ascoli and Partha P. Mitra and Hui Gong and Qingming Luo and Z. Josh Huang},
  title = {{Genetic Single Neuron Anatomy reveals fine granularity of cortical interneuron subtypes}},
  year = {2017},
  doi = {10.1101/219485},
  howpublished = {bioRxiv},
}
Figure