Preparation of Primary Neurons for Visualizing Neurites in a Frozen-hydrated State Using Cryo-Electron Tomography

We show that rat embryonic neurons (dorsal root ganglion and hippocampal) can be grown on gold electron microscopy (EM) grids and frozen in vitreous ice thin enough for their neurites to be imaged using 2-D cryo-EM and 3-D cryo-ET. While hippocampal neurites have previously been imaged using cryo-ET^8,10,18,23, a protocol detailed enough for successful replication using commercially available devices has been lacking. Furthermore, while their research has pioneered the use of cryo-ET for visualizing hippocampal neurites, our studies explore the applicability of cryo-ET to more than one type of neuronal specimen.

Here, we describe a detailed protocol preparing EM grids with either hippocampal and dorsal root ganglion (DRG) neurons, blotting them to achieve optimally thin ice, and plunge-freezing them for imaging by cryo-ET. Optimal blotting time is one which does not result in ice too thick for the electron beam to penetrate, nor too thin that the vitreous ice appears with notably different gradient from edge of the holey carbon to the innermost of the holey carbon. Furthermore, achieving optimally thin ice through using a vitrification machine¹⁷ may be slightly different for each project, and robust freezing using the vitrification machine requires a learning curve.

While previous research has focused on hippocampal neurons, also explored here in our studies, we have gone further to show images at the nanometer scale of intact, frozen-hydrated whole axons of the dorsal root ganglion (DRG), never imaged before using cryo-EM or cryo-ET. DRG cells were chosen since they (in contrast to hippocampal neurons) exclusively give rise to axons³⁴. A protocol for imaging them by cryo-EM/cryo-ET could be useful for those interested to deduce axonal ultrastructure i.e. in sensory neurons. We also present and discuss both optimal and suboptimal results, as well as the potential artifacts that one could encounter using cryo-ET for such specimens.

At the proper cell concentration (50,000 cells/ml per dish), spacing of neurons is such that neurites allow for excellent access using light and electron microscopy (Figures 3A and 3B) with one or two neurons appearing per grid-square (Figure 3B). Allowing enough separation between the neurons is necessary for clear visualization of the neurites via cryo-EM (Figures 3C and 3D). The neurons and their neurites are supported on a holey carbon film of the gold grid, as seen clearly in the cryo-EM images (Figures 3C and 3D). The neuron bodies, which can vary from ~15-50 µm for hippocampal neurons and dorsal root ganglion (DRG) neurons^30,33 are too thick for imaging using cryo-EM and appear completely black, as is typical for very electron-dense objects and thick specimens, as compared to the relatively thinner projections of neurites (not exceeding 1 µm in diameter) radiating from them (Figures 3C and 5).

Resin-embedded 3-D brain tissue does not reveal neurites that are completely circular in cross-section¹⁴; in fact, their morphology is variable. It is thus not known if the noncircular shape of neurites is natural or a specimen preparation artifact. The measured thickness of the DRG axon (Figure 4) grown on the TEM grid, blotted and imaged as shown in this study was 0.32 µm in the z-direction (in the depth of the vitreous ice) and an average width of 0.53 µm in the x-y plane. The measured thickness of the hippocampal neurite (Figure 8) grown on the TEM grid, blotted and imaged as shown in this study was 0.50 µm in the z-direction (in the depth of the vitreous ice) and an average of 0.72 µm in the x-y plane. The blotting process as described herein could flatten the neurites although they remained hydrated in their original buffer during the entire process of blotting and plunge-freezing process.

The slight flattening of neurites on the grid as observed by cryoET does not appear to affect the structure of its internal constituents such as vesicles as shown herein, which are almost perfectly circular and show no signs of dehydration. By contrast, the chemical processes commonly used in 'traditional' EM preparations, particularly chemical fixation with glutaraldehyde and paraformaldehyde followed by dehydration in ethanol, result in uncontrolled tissue shrinkage. The potential of encountering such a damaging cellular effect is eliminated in the method we present for cryoET in our manuscript.

Furthermore, our neurites samples were immediately plunge-frozen in liquid ethane, resulting in instant 'embedding' in vitreous ice. Tissues that are frozen in a vitreous state appear much more similar to the physiological state than other techniques. The level of ultrastructural detail is not the only aspect that makes the technique described in our protocol to be valuable to neuroscience researchers. The technique of growing neurons on EM grids, immediately blotting/plunge-freezing them to preserve in a vitreous state for screening/imaging by cryoET is an attractive technique because it opens up the possibility of examining ultrastructure of neurons under different genetic, chemical or environmental stresses without the concern of potential dehydration artifacts.

The blotting conditions prior to freezing, as detailed in this protocol, are important to generate ice thin enough for visualization by cryo-ET of the ultrastructural features, including internal components such as mitochondria, vesicles and microtubules (Figures 4, 6, and 7A). Mitochondria, for example, could be identified in our preparations since they appear structurally similar to what has been seen in cells (the edges of mouse embryonic fibroblasts) via cryo-electron tomography²⁰. They also display cristae, which can be seen in the 3-D color-annotated images present in this manuscript. Suboptimal images can result from thick ice in which the cryo-EM image appears generally opaque, preventing successful identification of such features as mitochondria and microtubules within the neurite. A lack of sufficient details for preparing such specimens, particularly regarding how to produce optimally thin ice, almost certainly contributed to the limited success in defining the ultrastructure of neurites^8,10,17,23.

Suboptimal images can also result from plating too many neurons on the grid (>50,000 cells/ml per dish), which results in overcrowding of neurites (Figures 7B and 7C). Blurry images may be a result of incorrect defocus and must be adjusted; the defocus in the images shown here was taken with a targeted underfocus of 7 µm. Images were collected at that defocus to ensure higher contrast and visibility of ultrastructural features; however, a lower defocus (i.e. 2-3 µm) could also be used to obtain higher resolution details of such features.

Other factors can contribute to suboptimal images. For example, although it may be tempting to use other features of the vitrification machine, particularly the semi-automated blotting feature, rather than manual blotting described herein, this gave undesirable results. Double-sided blotting was initially attempted using this semi-automated feature, in which the vitrification machine (rather than the user) blots the sample directly using such parameters as blotting time and number of blots as specified by the user. However, after imaging such samples, it was found that all neurons had lysed open, spilling out their contents (multivesicular bodies, etc.). Hence, the vitrification machine is best used in this case for its temperature-controllable humidity chamber in which the sample is blotted prior to plunge-freezing. Maintaining the sample in such a state (with humidity set close to 100%) minimizes water loss during preparation and helps to ensure optimal vitreous ice post-freezing^3,7.

Ideally, to minimize exposure of the neurons to external perturbations, they should be grown in a CO₂ incubator (temperature set to 37 °C) in close proximity to the room with the vitrification machine, and the temperature for this apparatus should be set to 37 °C prior to flash-freezing the neurons on the EM grids. One should avoid exposing neurons to large fluctuations in temperature prior to blotting. In this report, temperature was set to 32 °C for blotting purposes; neurons were grown in a different building and a time lapse of ~10 min occurred between carrying the plates of neurons from one room to another. Care was taken to protect the samples in a water-resistant, semi-insulated container during the transport process. A similar change in temperature and CO₂ concentration (as experienced by the neurons prior to blotting/plunge-freezing) occur to the neurons every ~2 days when the neurons are taken out of the incubation chamber for their media to be changed under the biosafety hood, as is typically done for neuronal cultures. This is not seen to result in toxic effects to the cell.

Regarding the presence of what appear to be electron-dense granules within the neurites' mitochondria, several conflicting reports exist. Such granules are found in a variety of different tissues, not specifically neuronal cells. They are perceived as a sink of cations that regulate the internal ionic environment of the mitochondrion. They also appear to create contact sites between inner and outer mitochondrial membranes in which enzymes can function efficiently¹⁶.

Experimental studies indicate that calcium and other divalent cations accumulate in mitochondria when these ions are present in the fluid bathing either isolated mitochondria or intact cells. It appears likely that the ions are organically bound to the pre-existing intra-mitochondrial granules. It is suggested, therefore, that these granules are linked to the regulation of the internal ionic environment of the mitochondrion³¹.

Furthermore, the neurons in our preparation were cultured on EM grids for 14 days, the timing of which is typical for purposes of neurite outgrowth^15,36, prior to imaging by cryoEM/ET. It has been reported that there is a dramatic increase in intracellular calcium concentration in aged neurons as compared to young neurons³². Therefore, it is plausible that the mitochondria in the neurites of these aged neurons would display mitochondrial granules not necessarily as a sign of stress, but rather because they possess a higher presence of intracellular calcium.

Electron-dense mitochondrial granules have also been reported in mouse embryonic fibroblasts grown in-culture and imaged by the same technology (cryo-electron tomography) as used in our manuscript²⁰. These cells did not show typical patterns of ultrastructural change associated with cellular injury, such as cytoskeletal disruption and vacuolization of the cytoplasm. Likewise, within our neurites as visualized by cryoET, we did not observe a disrupted cytoskeleton that would otherwise be associated with cellular toxicity. Given the above considerations, we conclude that the neurons in our preparation did not display mitochondrial granules as a result of stress or toxicity.

To sterilize the EM grid surface to prevent contamination to the neurons, and also to generate a favorable, hydrophilic surface on the EM grid to which the poly-L-lysine could adhere, flaming the EM grids prior to plating neurons was found to be advantageous. UV sterilization was not chosen since it requires irradiation of the grid for a longer time (e.g. 20-30 min) in the biosafety hood. Then, the grids would need to be re-exposed again to the environment when being glow-discharged for purposes of hydrophilizing the grid's surface. Glow-discharging the grid ensures that the subsequent poly-lysine coating would effectively 'stick' to the grid. The technique of flaming the grids is advantageous since it combines both the hydrophilizing step and sterilization step in one. Primary neurons are known to be more sensitive than tumor-based cell lines and it is crucial to keep their environment (grid, Petri dish) as sterile as possible.

To visualize features that extend along the length of a neurite, it is preferable to first take a lower magnification image at a lower electron dose (Figures 3C and 5), which covers more of the grid area for purposes of selecting which neurite to image. Then, 2-D cryo-EM images can be taken at intervals and digitally stitched together into a montage (Figure 6). This technique may be useful for ultrastructural features, such as microtubule organization, over a greater area than that of a single cryo-EM image.

Taking low-dose, low-magnification images (Figures 3C and 5) is also useful in identifying areas of the neurite that are consistent in diameter across the holey carbon film (both in the holes and across the carbon) and hence more ideal for imaging and subsequent analyses. Enlarging of neurites over holes of the carbon film is typically seen in larger neurites (Figure 5) with clearly more internal material than thinner neurites (Figure 3C). This phenomenon is thought to occur due to the lack of physical support in the holes of the carbon film. When the EM grid is removed from the dish in which the neurons were growing, and subsequently blotted from the backside, neurites with more internal mass (Figure 5) than others (Figure 3C) tend to expand and/or sag into the holes. Although this appears to be an artifact, it has a benefit as to show more clearly the internal constituents of the neurite. Using a continuous carbon film overlaid on this holey carbon film may help to circumvent this issue in future studies. This setup was initially attempted but resulted in vitreous ice too thick for imaging. Subsequent trials with different thicknesses of continuous carbon overlaid on the EM grid may need further investigations.

An important benefit of the current method is that the spatial organization of cell components within the neurite can be visualized at the nanometer scale by taking several 2-D images of a single neurite at different tilt angles, and reconstructing this stack of images into a tomogram. Individual structural features can then be manually annotated for each 2-D slice of the 3-D stack in different colors using the appropriate software (see Table of Materials and Reagents) as shown for an axon of a rat E18 dorsal root ganglion (DRG) neuron (Figure 4) and a neurite of a rat E18 hippocampal neuron (Figure 8).

By choosing to take an image every 5 degrees during the tilt series, rather than 1 or 2°, fewer images are collected but a higher electron dose can be used per image. This results in higher contrast and less noise in the raw images. This is better for purposes of reconstruction (alignment by cross-correlation as one step) and subsequent tomogram annotation, which is done by-hand for each image comprising the tomogram stack in the z-direction. Choosing an increment size also depends on the size of the specimen; between different tilt angles, larger specimens will not vary as much. In this case, using a larger increment size (5°) was thought to be more suitable due to the relatively large specimen size of the neurite.

Furthermore, adding fiducial gold markers to the EM grid would assist for fine image alignment if desired. Fiducial markers were not used because the sample was relatively large compared to other cryo-EM samples (viruses or proteins in isolation) and showed high contrast with a variety of structural features that could be used for proper alignment of each of the 2-D images (comprising the 3-D stack) by cross-correlation. The resulting image stack was well aligned using this approach, and useable for subsequent 3-D color annotation.

For future studies, using a microscope with an energy filter could reduce noise caused by inelastically scattered electrons but such a feature may not be readily available at most electron microscopy facilities. Our procedure reveals what images can result from a standard 200 kV microscope with no energy filter, which may be more commonly available to the neuroscience community at large.

The protocol described here is optimized for preparing and visualizing both rat embryonic DRG axons and hippocampal neurites using commercially available equipment for cryo-EM and cryo-ET. One of the main advantages of this technique is that it yields structural information from whole neurites, not sectioned, milled or treated by various chemical reagents in order to view their ultrastructure. We have demonstrated how cryo-ET is a particularly excellent method for use in future ultrastructural analyses of neurons in a close-to-native state for examining the impact of neurological disease on neurite structure.

The Electron Microscopy Data Bank accession numbers for the 3-D tomograms as reported in this paper are as follows: For the rat hippocampal neurite as shown in Figure 8 and in the main video (from 8:45-9:01), the EMDB accession number is EMD-5887. For the rat dorsal root ganglion neurite as shown in Figure 4 and in the main video (from 9:02-9:15), the EMDB accession number is EMD-5885.