Correlative In Vivo 2-Photon Beam Scanning Electron Microscopy: 3D Analysis of Neuronal Ultrastructure

Bohumil Maco, Anthony Holtmaat, Anne Jorstad, Pascal Fua, Graham KnottInfoscience 2014 · DOI

Abstract

This paper details a robust protocol for Correlative Light and Electron Microscopy (CLEM), specifically bridging in vivo 2-photon imaging with Focused Ion Beam Scanning Electron Microscopy (FIB-SEM). By utilizing Near-Infrared Branding (NIRB) to create fiducial marks, the authors enable precise 3D reconstruction and morphometric analysis of specific neuronal structures first identified in the living brain. This workflow addresses the critical challenge of locating microscopic regions of interest within the complex volume of the brain, allowing for a direct link between dynamic cellular behavior and static ultrastructural detail.

Introduction: The Resolution Gap in Neuroscience

Neuroscientists often face a fundamental trade-off between temporal and spatial resolution. While 2-photon microscopy is ideal for tracking the life history of neurons—such as the growth or retraction of dendritic spines over weeks in a living animal—it lacks the resolution to visualize individual synapses or internal organelles. Conversely, Electron Microscopy (EM) provides nanometer-scale detail but is restricted to fixed, static tissue.

Bridging these two modalities, known as Correlative Light and Electron Microscopy (CLEM), has traditionally been a "needle-in-a-haystack" problem. Finding a specific 10-micron stretch of a dendrite within the vast volume of the brain is labor-intensive and error-prone using manual serial sectioning. This paper presents a modernized pipeline that automates this process and improves reliability.

The CLEM Pipeline: From Living Brain to Nanometer Resolution

The authors outline a multi-step workflow designed to modernize the CLEM approach:

  1. In Vivo Imaging: Neurons are first imaged in a living mouse using 2-photon microscopy to identify dynamic regions of interest (ROIs).
  2. Near-Infrared Branding (NIRB): After tissue fixation, the same 2-photon laser is used to burn identifiable fiducial marks (such as squares or rectangles) into the tissue near the ROI. These marks survive chemical processing and serve as a GPS for the EM.
  3. FIB-SEM (Slice-and-View): Instead of manual cutting, a Focused Ion Beam (FIB) automatically mills away nanometer-thin layers of tissue, while a Scanning Electron Microscope (SEM) images the newly exposed face. This results in isotropic voxels, where resolution is identical in X, Y, and Z planes—a necessity for reconstructing complex 3D structures like axons.
Overview of the correlative workflow. (A, B) Identification of the region of interest (ROI) and surrounding neurite architecture. (C, D) High-magnification imaging showing dendritic spines and the laser-induced branding marks (white arrowheads in D) that serve as fiducial points for subsequent electron microscopy.

Key Technological Advancements

  • Label-Free Identification: The NIRB technique allows researchers to find their target without electron-dense labels that might obscure fine details. This ensures optimal tissue fixation and high-quality ultrastructure preservation.
  • Automation and Efficiency: The shift from manual Transmission Electron Microscopy (TEM) to automated FIB-SEM transforms a highly skilled task into a repeatable engineering process. Automated segmentation tools like Ilastik can reduce the time required for 3D reconstruction from days to hours.
  • 3D Morphometry: By importing EM reconstructions into software like Blender (via the NeuroMorph plugin), researchers can treat neurons as CAD models, allowing for precise measurements of volume, surface area, and synaptic contact length.

Results and Morphometric Analysis

The protocol enables the creation of 3D models of dendritic spines and axonal boutons with a voxel size of approximately 5nm. This resolution is sufficient to clearly identify individual synaptic vesicles and mitochondria. The NIRB technique ensures that ROIs can be located within a few microns, even after the tissue undergoes the shrinkage and distortion typical of resin embedding.

Limitations and Future Outlook

Despite its advantages, FIB-SEM is slower than some alternative EM methods and is generally limited to relatively small volumes (tens of microns). Additionally, the ion milling process generates heat, requiring stabilization periods to prevent imaging drift. While machine learning has improved segmentation, manual proofreading remains a bottleneck for complex neuronal architectures. Future developments in more efficient detectors and advanced computer vision are expected to further automate the reconstruction stage of the pipeline.

Reference: Maco, B., et al. (2014). Correlative In Vivo 2-Photon Imaging and Focused Ion Beam Scanning Electron Microscopy: 3D Analysis of Neuronal Ultrastructure. Methods in Cell Biology, Vol 124.

Summary updated

BibTeX
@inproceedings{maco2014correlative,
  author = {Bohumil Maco and Anthony Holtmaat and Anne Jorstad and Pascal Fua and Graham Knott},
  title = {{Correlative In Vivo 2-Photon Beam Scanning Electron Microscopy: 3D Analysis of Neuronal Ultrastructure}},
  booktitle = {Infoscience},
  year = {2014},
  doi = {10.1016/b978-0-12.801075-4.00016-1},
}
Figure