Analysis of Brain Mitochondria Using Serial Block-Face Scanning Electron Microscopy

The complexity of the nervous system poses a significant challenge in reconstructing large tissue volumes and analyzing the morphology and distribution of organelles such as mitochondria with adequate resolution. Multiple cells including neurons, oligodendrocytes and astrocytes with numerous processes extended in three dimensions interact within the brain tissue ⁴³. Since mitochondria resides both in the soma of cells and distant processes, mitochondrial morphology is extremely pleomorphic in the nervous system (Figure 1). Adequate 3D structural information with sufficient resolution therefore cannot be acquired by conventional light microscopy techniques such as confocal or two-photon microscopy ^44,45. Electron microscopy is the only currently available technique that allows reconstruction of large volumes of neural tissue with sufficiently high resolution. Traditionally, 3D reconstruction of neural tissue has been achieved by serial section transmission electron microscopy (SSTEM) of ultrathin sections ²⁹. However, recent technological advancements have improved the quality of volume electron microscopy data acquisition and automation.

SBFSEM is an automated block-face imaging technique which combines serial sectioning inside the chamber of a scanning electron microscope, in order to reconstruct 3D tissue structure over hundreds of micrometers, with a resolution that is sufficient to follow the thinnest cellular processes and identify small organelles. This technique has opened up the prospect of routinely reconstructing both invertebrate and vertebrate nervous systems. It involves multiple steps including sample preparation, scanning electron microscope operation, data acquisition, image post-processing, and image analysis. These procedures require specialized hands-on training and certification for scanning electron microscope operation. Three factors are critical i.e. dissecting the correct brain anatomical region, obtaining high resolution images, and performing proper image analysis. After dissecting the anatomical region of interest from the vibratome slice of fixed brain tissue, it is crucial to take the picture for preserving proper orientation. It can otherwise be difficult to ascertain the correct site for imaging. Obtaining high resolution images depends on adequate fixation and proper post-fixing staining procedures. The brain tends to undergo rapid deterioration in its ultrastructure after death. Therefore, it is critical to fix the brain in a glutaraldehyde solution using proper transcardial perfusion method. This should be followed by further fixation of the brain in glutaraldehyde solution for at least 24 hr. Since the image resolution is completely dependent on the combination of a variety of heavy metal staining methods, it is critical to follow the post-fixing method as detailed in the protocol to acquire quantifiable images. Solutions should be made as described in the methods. Finally, for accurate image analysis the organelle/s to be analyzed should be identified unequivocally (e.g. by visualizing ultrastructural characteristics) before tracing, and the software guidelines must be strictly adhered to. The resolution of the acquired images should be known for conversion from pixels to nanometers for volumetric measurements.

Besides the software described in the protocol for performing image analysis, other available software's like Reconstruct and Knossos can also be used. Although the emphasis of the described protocol is on viewing mitochondria in the nervous system, it can be modified to generate high resolution 3D information on various other subcellular structures (e.g. endoplasmic reticulum, lysosomes etc.) in a variety of tissues.

While a detailed troubleshooting guide for scanning electron microscope operation is beyond the scope of this article (see instrument user manuals and training information), a few problems routinely occur during imaging experiments and understanding how to troubleshoot them may help new users or those obtaining images through collaboration/ commercial sources. A common problem is charging of the sample which occurs principally in resin regions that contain little or no stained material. In a "positive" image (i.e. cytosol appears white), the nuclei, blood vessels, and empty resin expanses may appear "blackened". In addition, charging also promotes local beam drift resulting in apparent image warping. Possible solutions vary between instruments and samples. Reducing kV settings reduces the "blackening" artifact but may promote more beam drift. Using a lower vacuum mode, and including nitrogen (N₂) gas, water vapor, etc. in the chamber also reduces charging, but at a substantial cost of resolution and signal-to-noise ratio. A second common problem is knife skipping. Slower scanning may cause beam induced damage to the block surface, which cuts unevenly or not at all for some imaging/ cutting cycles (i.e. successive images appears the same). Several solutions exist including increasing the cutting depth/ slice thickness with reduced z-resolution, increasing scan speed and consequentially accepting noisier images, reducing pixel size and accepting lower x/y-resolution, and choosing samples or areas with more intense staining (as poorly-stained areas charge more, damage more easily, and require longer beam exposure to obtain images with acceptable signal-to -noise ratios). Debris redeposition on the block-face is an occasional source of artifact in images, and if this occurs frequently it may require pausing the acquisition, and carefully cleaning the knife (blowing air), or retrimming the sample to a smaller size. Block-face charging also promotes redeposition, and steps above may help. Correction of focus and stigmation may also be required as the microscope images samples continuously for many hours to weeks, during which time the vacuum depth increases in the chamber, requiring stigmation corrections. Each scanning electron microscope differs according to age and characteristics, and experience is the best guide. Sudden dramatic changes in stigmation or focus may occur when sectioned material adheres to the microscope imaging components. The chamber must be opened and material carefully dislodged through vacuum or compressed nitrogen. This absolutely requires specialized training.

There are a few disadvantages of SBFSEM technology with regards to its utility in analyzing brain mitochondria. A major disadvantage, common to all microscopy of fixed tissues, is that it provides static images of an extremely dynamic organelle. Mitochondria undergo continuous fission and fusion cycles ²⁶, are mobile and trafficked along the neurite processes ²¹. Defects in mitochondrial trafficking and dynamics which are not purely numerical or structural can be easily missed by this approach, although, by looking at the number of mitochondria in an anatomically defined space such as presynapse one can interpret if the transport of mitochondria may be altered ³⁹. A second disadvantage is that it is performed only in a small portion of the brain, therefore circuitry specific defects in mitochondrial distribution may be missed. Correlative approaches ^28,46 afford the opportunity to combine the confocal and multiphoton imaging with SBFSEM technology to generate both the molecular and ultrastructural data. For mitochondrial analysis, therefore, it may be useful to perform the SBFSEM after light microscopic examination of brain sections either by using the mitochondrial antigen specific antibody or by using a transgene which expresses mitochondrially-targeted fluorescent protein. This combinational strategy may provide a robust methodology to identify mitochondrial defects in mouse models of neurological and mitochondrial disorders.

There are also technical limitations common to many forms of electron microscopy due to use of harsh fixatives, heavy metal staining and uneven penetration of chemicals. Other limitations may stem from the plastic embedding of tissue samples. Empty regions of resin and sparsely stained structures retain electrons during imaging, causing beam deflection and drift (image warping) as well as artifactual charging signals (e.g. dark nuclei and blood vessel lumen), which both impinge on resolution. Beam damage to the resin also promotes uneven cutting, and for this reason, imaging typically requires slices of 50 – 100 nm to be cut from the block-face, producing non-isotrophic voxels in the datasets. Other limitations for some applications include that the sections are destroyed and cannot be reexamined subsequently, which may force users to collect high resolution images of areas that might not contain useful data.

A major advantage of SBFSEM is that it automates the process of sectioning and imaging blocks of tissue by incorporating a custom microtome into a low vacuum SEM chamber ^32,33. Since the images are obtained directly from the block-face prior to each cut, the problems of section wrinkling, compression and loss during handling are substantially avoided, although debris deposition and warping due to block-face charging do contribute to some image loss and distortion. Furthermore, the images obtained in raw datasets are already aligned and require only micrometer-scale registration to accommodate beam drift in order to be amenable to most analysis. Because of the automated sectioning process, once the system vacuum has stabilized, large volumes of tissue can be imaged without significant operator involvement. There are multiple advantages of using this technology in performing morphological and quantitative studies on organelles and intracellular structures. One advantage is getting the information on 3D structure of mitochondria within a reasonable amount of time. The availability of open source reconstruction software packages like Reconstruct ⁴⁷, TrakEM ⁴⁸ and Knossos ^49-51 have made this technology a powerful analytical tool where detailed 3D ultrastructure of neuronal network from experimental animal models can be directly compared. The semiautomated and fully automated image analysis approaches ⁵² are likely to produce highly mechanistic data in animals where mitochondrial morphology, function and/or biogenesis are expected to be affected. Earlier SBFSEM analysis was greatly hindered due to lower resolution, however newer sample preparation techniques involving enhanced staining methods ⁴⁰ have considerably improved the resolution to an extent that even a single synaptic vesicle can be easily resolved. At this resolution ultrastructural defects in mitochondria such as vacuolation, membrane disruption, and loss of cristae can be easily observed, and the distribution of defective mitochondria within cells can be determined ^38,39. The high resolution of this technique provides a major advantage over light microscopy where the resolution is limited to ~200 nm in XY axis and ~500 nm in Z-axis. Mitochondria are therefore just at the resolution limit of light microscopy ⁵³. Although mitochondria having dimensions below the resolution limit can still be visualized by light microscopy, their dimensions cannot be reliably measured. Importantly, mitochondria that are separated by a distance lower than the resolution limit of a light microscope cannot be resolved by light microscopy. Another advantage is that the mitochondrial structural analysis can be performed within the context of the tissue, without isolation of the organelle. This can allow for appropriate comparisons within the tissue based on mitochondrial localization, for e.g. axonal vs. somatic and/or dendritic mitochondria. An additional advantage is that it is usually possible to determine whether the mitochondria are neuronal or neuroglial in localization by following the processes to a defining organelle, such as glial filaments and glycogen for astrocytes or myelin for oligodendrocytes. The processes of neurons and neuroglial cells are often in close proximity, and with 2D approaches such as TEM it may be difficult to assign with confidence which processes are neuronal and which are neuroglial.

In conclusion, the SBFSEM technology provides stacks of serial images covering tissue areas 20 - 1,000 µm in size with ultrastructural resolution of 5 – 10 nm or higher, which allows entire central nervous system cells and organelles such as mitochondria to be imaged, measured and reconstructed. In this paper, we have provided some practical approaches to making use of this technology. Future applications include analyzing the mitochondrial structure and distribution in variety of animal models of neurological diseases such as neurodevelopmental and neurodegenerative disease models as well as analyzing the various subcellular ultrastructures such as endoplasmic reticulum, nucleus and/or lysosomes in whole cells or tissues under healthy and disease states.