Obtaining high resolution data from a large volume of brain tissue in an automated and reliable process

By Christel Genoud, Gatan

Morphometry is an important and growing discipline within neuroscience. Theoretical models of neuronal circuits require 3D information over extensive volumes. Current models are based on real data obtained from serial sectioning brain tissue and subsequent reconstruction to show components present, for example synapses, axons, dendrites. Realistic and meaningful analysis requires morphometric analysis at the ultrastructural level over large sample volumes. Large volumes are required in order to be statistically relevant and usable for model building.

Electron microscopy is key to providing information at the ultrastructural level.

Until now, the classical method to obtain such data was serial sections collected on grids and observed in the TEM. This is a long and difficult process requiring much skill. Sections are obtained as ribbons using an ultra-microtome. Ribbons must be divided manually. Multiple sections are then collected on grids using the surface tension of water near the diamond knife. Sections on grids are processed in various chemicals. All these steps are risky as the grids or sections can be broken. Only one broken grid compromises the entire series of sections because it introduces an important gap in the series.

A revolutionary solution called 3View, from Gatan, uses serial block face scanning electron microscopy (SBFSEM). This new method automates the cutting and imaging of the specimen. Grids are no longer used because sections are not cut for collection. It is the block of tissue itself that is directly introduced inside the scanning electron microscope and its surface is repeatedly shaved and scanned in order to obtain a stack of aligned images. The raw images obtained can be directly exported to software for reconstruction and quantification. Image stack alignment is due to the inherent design and stability of the system and typically does not require software post-processing. The following article will describe the main steps necessary to obtain this kind of data.

Case study: imaging ~ 10,000 synapses in a volume of mouse cortex using 3View.

1. Tissue processing

Like any microscopy technique optimum results, especially spatial resolution in 3D, are often achieved by following certain protocols. In the case of 3View specimen preparation, two key considerations are the level of contrast shown by the specimen, and the resilience of the embedding resin to the electron beam radiation. For a given specimen, higher image contrast is achieved by increasing experimental variables such as the beam energy and the number of injected electrons per pixel point. However, such higher injection conditions have other detrimental effects on the specimen and imaging quality, for example the cutting quality, additional scattering associated with the gas required to compensate the injected charge, as well as the depth resolution of the 3D data. For this reason, a specimen optimised to give the highest possible 3D resolution data whilst minimizing the acquisition timeframe, is one with high contrast in the tissue, and embedded in a medium relatively resistant to electron beam damage. In this particular case, after fixation, the tissue has been postfixed in osmium tetroxide with ferrocyanide. In a further step, uranyl acetate has been used to enhance the contrast of some organelles. Epoxy based resins (e.g. epon, durcupan) are superior and in this case the resin used was durcupan.

Figure 1 Contrast and resin optimization. The tissue has been processed following a protocol with high osmium content.. This image is 2048x2048 pixels with each pixel being 12nm square.

2. Immunohistochemistry

3View is a technique compatible with immunohistochemistry. However, only pre-embedding techniques are suitable because SBFSEM is not based on the observations of sections but on the imaging of the block face. The pre-embbeding method used in this case study of a mouse cortexis described. Once the tissue is fixed, it is cut into thick sections and a protein antibody called parvalbumine is employed. Parvalbumine is known to be specific to a subset of inhibitory neurons in the cortex and its presence is used in the classification of the inhibitory interneurons of the cortex. The secondary antibody is biotinilated and DAB is used for staining. Following the DAB staining, tissue is processed with a classical embedding protocol. It is important to ensure penetration of the antibody inside the tissue. If the aim is to follow a labelled structure, one must check that the labelling is not restricted to the surfaces of the section but has penetrated through all the section.

Figure 2 Example of pre-embedding labelling visualised with 3View. The tissue has been processed for immunochemistry with DAB staining. This image shows a section through the apical dendrite of a cortical neuron. A protruding spine making a synapse is visible.

3. Trimming of the block

Once embedded, the tissue should be trimmed before being inserted in the 3View microtome. The piece of tissue to be observed is glued on the top of a specimen holder specific to 3View. In this example, the specimen has been trimmed to obtain a rectangular or trapezoidal surface of approx 300x400 microns. Two sides are parallel and are trimmed at 45 degrees. The 2 other sides are trimmed at 90 degrees. When tissue is stained, it is important to trim the pyramid in order to centre the labelled structure on the pyramid. Specimen preparation including block trimming is advised because it can influence the stability of the stacks, especially when automated cuts of extensive acquisition times are required.

4. Setting of the microscope

Once the sample is ready to be cut, it is placed on the 3View ultramicrotome. The 3View ultramicrotome is already attached to the door of the SEM and can be operated out of the chamber. In this open configuration of the microscope, a binocular optical microscope can be adjusted on the door and the sample set relative to the diamond knife. This is made convenient via a camera fixed on the optical microscope. When the sample has been adjusted with respect to the diamond knife, the door is closed and the SEM evacuated. The face of the block is now positioned just below the pole piece of the SEM.

5. Cutting and image acquisition

All the cutting and acquisition parameters are entered into Gatan’s DigitalMicrograph software. Settings are chosen in order to fit the magnification and speed of scanning with the field of view and the resolution required. The focus is adjusted at the same time and is then valid for the entire acquisition session.
Detection is via a backscattered electron detector (BSED) , designed especially for low kV work on biological specimens. The cutting and imaging process is based on the repetition of the following steps:

This illustration is based on 800 sections obtained in approx 14 hours. The parameters have been chosen in order to have a field of view of 20 by 20 microns with a voxel size of 10x10x50 nm showing entire cell bodies of neurons as well as the synapses present on them. 800 sections x 50nm means that 40 microns of depth in the tissue has been cut and can be sequentially seen in Digital Micrograph

With this acquisition, the identification of ~ 10,000 synapses present in the volume of tissue investigated would be possible (Rodney et al, 2004).

6. Three dimensional (3D) visualisation and rendering

Alignment allows projection in all axes and visualisation in all desired planes. Gatan can provide an additional tool called the DigitalMicrograph 3D Visualisation Tool, allowing 3D visualisation as well as 3D automatic rendering from any image stack. With this tool, the stacks can be visualised as volumes and can be rotated, projected on all the different axes and sliced through. Examples are shown in figure 3.

Figure 3 Examples of 3D visualisation possible with the Gatan Visualisation Tool. This module allows visualisation in different planes

7. Image exportation and reconstruction

In this case study, images have been exported and saved in TIFF format in an ascending numerical order. The TIFF images were opened in the reconstruction software and a manual segmentation of a neuron present in the images performed (figure 4) as well as identification of the synapses. Because the tissue is labelled for parvalbumin, we can determine which of the synapses are made with inhibitory interneurons expressing parvalbumin.

Figure 4 Example of reconstruction software (Reconstruct, GNU General public licence version 2) showing an image from the stack with the neuron of interest manually segmented. The window on the right shows the superposition of all the traces done and the scaffold of the neuron. This scaffold of traces is converted by the software into a volume having a homogeneous surface allowing our eyes to visualise a structure in 3D.