Serial block-face scanning electron microscopy to reconstruct three-dimensional tissue nanostructure

Abstract

Three-dimensional (3D) structural information on many length scales is of central importance in biological research. Excellent methods exist to obtain structures of molecules at atomic, organelles at electron microscopic, and tissue at light-microscopic resolution. A gap exists, however, when 3D tissue structure needs to be reconstructed over hundreds of micrometers with a resolution sufficient to follow the thinnest cellular processes and to identify small organelles such as synaptic vesicles. Such 3D data are, however, essential to understand cellular networks that, particularly in the nervous system, need to be completely reconstructed throughout a substantial spatial volume. Here we demonstrate that datasets meeting these requirements can be obtained by automated block-face imaging combined with serial sectioning inside the chamber of a scanning electron microscope. Backscattering contrast is used to visualize the heavy-metal staining of tissue prepared using techniques that are routine for transmission electron microscopy. Low-vacuum (20-60 Pa H(2)O) conditions prevent charging of the uncoated block face. The resolution is sufficient to trace even the thinnest axons and to identify synapses. Stacks of several hundred sections, 50-70 nm thick, have been obtained at a lateral position jitter of typically under 10 nm. This opens the possibility of automatically obtaining the electron-microscope-level 3D datasets needed to completely reconstruct the connectivity of neuronal circuits.

Figure 1. Resolution and Contrast Using the Backscattered Electron Signal

(A and B) Presynaptic vesicles (SV) and postsynaptic folds (SF) are clearly visible (A) in a motor endplate preparation embedded in Spurr's resin. Similarly, the hexagonal array of actin filaments (AA) can be clearly resolved (B) in a different region from the same image (both images were smoothed using the ImageJ “smooth” command). Imaging conditions for (A) and (B): electron energy, 7.5 keV; spot, 3.5; chamber pressure, 30 Pa (H₂O); pixel dwell time, 30 μs. The scanning resolution was 6.7 nm/pixel. (C) The effect of beam exposure on the block surface. Note the increased brightness and the lack of chatter in the central region (inside the dashed rectangle), from which a stack was acquired at higher resolution before taking the image shown. The tissue was rat neocortex embedded in Spurr's resin. Imaging conditions for (C): electron energy, 7.5 keV; spot, 3; digital resolution for stack acquisition, 26.7 nm/pixel; dwell time, 30 μs. (D and E) Cortical tissue embedded in Epon. Synapses (SD) are clearly discernable (E). Imaging conditions for (D) and (E): electron energy, 7.5 keV beam current; spot, 3; chamber pressure, 30Pa (H₂O); pixel dwell time, 30 μs. The scanning resolution was 9.5 nm/pixel. Note that more backscattering corresponds to darker pixels in (A), (B), (D), and (E) but to brighter pixels in (C).

Figure 2. SEM Microtomy

(A) Principle of SBFSEM operation: (1) a SEM image is taken of the surface of the plastic-embedded tissue preparation (amber trapezoid). (2) Then with a diamond knife (blue) an ultrathin slice is cut off the top of the block. (3) After retraction of the knife, the next picture is taken. The pictures shown are from an actual stack (cerebellar cortex) but are not successive slices; rather, they are spaced by five images (about 315 nm) to make the changes more apparent. (B) Usually cut-off slices pile up on the top of the knife. Protruding into the picture from the right is a puffer pipette, occasionally used to remove debris from the knife. (C and D) The mechanical design for the in-chamber microtome is shown in an overview (C) and a close-up of knife and sample (D) in renderings from the computer-aided design software. Most parts are nonmagnetic stainless steel (grey). A large-motion leveraged piezo actuator (green part on the left) drives the knife holder back and forth. The custom diamond knife (light blue) is clamped in a special holder. The sample (amber) advance is driven via a lever by a direct-current-motor-driven micrometer (dark blue). The retraction during the backwards knife motion is again piezo actuated (green cylinder in the lower right of [C]). Bearing springs are brown. The BSE detector (red) is depicted schematically above the sample. Not shown is the lateral positioning mechanism.

Figure 3. 3D Datasets: Five Slices

(A) Five slices from a stack containing a total of 365 slices at 63-nm section thickness; same tissue as in Figure 1A and 1B. Note beginning of myelin sheath (MS) in the lowest slice. Imaging conditions as in Figures 1A and 1B except resolution is 13.4 nm/pixel. (B) Reslice of the same dataset along the red line show in the top image of (A). Arrows point to the presynaptic ending (PS) and the z-band (ZB). Image corners in (A) touch the reslice (B) at the depths at which they were taken. (C and D) Cerebellar tissue displayed at low (C) and high (D) resolution; note nuclear envelope (NM). Note differences in lateral and vertical resolution. Imaging conditions for (C) and (D): electron energy, 7.5 keV; spot size, 3.5; digital resolution, 12.7 nm/pixel.

Figure 4. The Alignment of Successive Images in a Stack

Shifts between images were quantified using the positions of the peaks of the cross correlation (see Materials and Methods); same dataset as in Figure 3A and 3B. (A) The peak shifts in x-direction are shown for five different subregions distributed over the field of view. Four of the regions have a size of 256 × 256 pixels, one has a size of 512 × 512 (black trace). The peaks around slices 59 and 202 are caused by slice debris on the block face (see also streaks in Figure 3B). (B) The x/y displacement for the 512 × 512 region is shown in a scatter plot. For the central cluster the standard deviations are 10.9 nm and 11.8 nm for x and y, respectively.

Figure 5. Energy Dependence of the Depth Resolution

The lateral resolution does not change very much as the electron energy is reduced from 7.5 keV (A) to 4 keV (B), but the resolution along the z-direction is dramatically different (C). The lines between (A), (B), and (C) indicate the z-positions in the stack from which (A) and (B) were taken. In the high-resolution view (lower panel in [C]) membranes that were en-face (EM) in the original slices can be clearly recognized. The nominal slice thickness was 55 nm. Tissue is rat cortex. Imaging conditions: dwell time, 25 μs; spot sizes, 3.5 and 2.8 for 4 and 7.5 keV, respectively.