Ultrastructurally smooth thick partitioning and volume stitching for large-scale connectomics

Journal name:
Nature Methods
Volume:
12,
Pages:
319–322
Year published:
DOI:
doi:10.1038/nmeth.3292
Received
Accepted
Published online

Focused-ion-beam scanning electron microscopy (FIB-SEM) has become an essential tool for studying neural tissue at resolutions below 10 nm × 10 nm × 10 nm, producing data sets optimized for automatic connectome tracing. We present a technical advance, ultrathick sectioning, which reliably subdivides embedded tissue samples into chunks (20 μm thick) optimally sized and mounted for efficient, parallel FIB-SEM imaging. These chunks are imaged separately and then 'volume stitched' back together, producing a final three-dimensional data set suitable for connectome tracing.

At a glance

Figures

  1. Overview of ultrathick sectioning.
    Figure 1: Overview of ultrathick sectioning.

    (a) Tissue volumes with dimensions too large for FIB-SEM, such as an entire adult Drosophila brain, can be sectioned into optimally sized 'ultrathick' chunks and imaged across multiple FIB-SEM microscopes. (bf) Steps involved in preparing ultrathick sections for FIB-SEM imaging. Oil-covered sections are retrieved using vacuum tweezers (b) and flattened with a hot plate (c). Oil is removed by dipping each section in a succession of Durcupan drops (d). Sections are flat embedded in Durcupan against a poly(ethylene terephthalate) (PET) laminate (green layers) (e). Sections are cut out and adhered to metal studs. Each tab is trimmed with a UV laser and diamond knife and then conductively coated to ready it for FIB-SEM imaging (f). (g) SEM image of the top surface of a sample tab. (h) FIB-SEM image of a 20-μm ultrathick section of mouse cortex cut from an original 100-μm vibratome section. Scale bars, 10 μm (center) and 1 μm (top and bottom). Magnified images of a smooth hot-knife cut edge (blue) and original vibratome cut (red) are shown. (i) Two-dimensional ultrathick sectioning. A 20-μm ultrathick section of mouse cortex was re-embedded and sliced orthogonally to the original cut plane, creating a square cross-section. Scale bar, 10 μm.

  2. Volume stitching on C-PLT-prepared Drosophila brain.
    Figure 2: Volume stitching on C-PLT–prepared Drosophila brain.

    (a) FIB-SEM images of two sequential 20-μm ultrathick sections. (b) Volume stitching over an 8 μm × 12 μm cropped region. The dotted red line on the 3D volume shows the stitch plane. Images with colored borders correspond to cut planes through this 3D stitched volume (Supplementary Video 10). (c,d) Aligned hot-knife cut edges before (c) and after (d) algorithmic flattening and stitching. (e,f) Volume tracing of the 4 μm × 4 μm × 4 μm volume from which c,d were obtained. Supplementary Video 12 contains side-by-side movies of all slices of this same volume before and after stitching as well as after volume tracing. (g) 9 μm × 10 μm cropped region that has been split open along the stitch plane. Processes in the densely packed central region were traced across the gap (Supplementary Video 11). Scale bars, 10 μm (a) and 1 μm (be).

  3. Early thick sectioning tests showing surface damage.
    Supplementary Fig. 1: Early thick sectioning tests showing surface damage.

    (a) SEM cross section of a vibrotome cut surface showing extensive ultrastructure damage. Scale bar = 5 μm. (b, c) Diamond knife boat heater clamp used for hot knife thick sectioning. (d) Schematic showing how hot knife cut sections were mounted for SEM imaging of their sectioned surfaces. (e) SEM BSE (backscatter electron) image (taken at 6kV) of a sequential pair of 20 μm thick hot knife sections mounted so that their matching faces are facing up. Scale bar = 100 μm. (f) BSE image zooming in on the tip region of one of the sections shown in (e). This BSE image may appear to be damage free, but when the same region is imaged using SE (secondary electron) imaging (g) considerable surface damage is evident. Scale bars = 1 μm. (h,i) BSE and SE images respectively of a separate hot knife cut section. Chatter damage is visible at bottom. Surface damage is clearly seen in the SE image (i) across the entire section with the exception of a triangular region at the very center of the tissue. This same triangular region is seen to contain less stain contrast in (h) supporting the hypothesis that some surface damage was due to the overly brittle nature of this ROTO prepared tissue in those regions of the tissue where this ROTO procedure fully penetrated. (see Online Methods for more details.) Scale bars = 10 μm.

  4. Improved hot knife-sectioning technique results.
    Supplementary Fig. 2: Improved hot knife–sectioning technique results.

    (a) Heating jig designed for use with the Diatome ultrasonic diamond knife to allow ultrathick sectioning with a commercial ultramicrotome. The ultramicrotome's knife mount is used to clamp the heating jig and knife base in metal-to-metal contact. Three small insulators made from rigid circuit board material help insulate the knife and heating jig. A PID controller (Omega) reads the thermocouple to regulate temperature. (b) BSE image (taken at 6kV) of a sequential pair of 20 μm thick hot knife sections mounted so that their matching faces are facing up. (c) Same sections imaged with SE. Scale bars = 100 μm, (d,e) Zoomed regions of previous imaged in SE at higher magnification. Scale bars = 50 μm, (f,g) Inset regions of previous imaged in SE at higher magnification. Scale bars = 10 μm. Hot knife cut surfaces cut with this improved technique look exceptionally free of damage over a wide area and a wide range of imaging scales.

  5. Matching ultrastructure profiles across hot-knife cut.
    Supplementary Fig. 3: Matching ultrastructure profiles across hot-knife cut.

    (a) High magnification low voltage (1.5kV) SE image of the surface of a 20 μm thick hot knife section showing small neuronal processes cut by the heated knife. (b) Mirror reversed SE image of the matching region of the sequentially cut hot knife section. (c,d) Same images colored to show matching neuronal profiles identified across the hot knife cut. Scale bar = 1 μm

  6. Ultrathick sectioning apparatus and procedure.
    Supplementary Fig. 4: Ultrathick sectioning apparatus and procedure.

    Current ultrathick sectioning apparatus and procedure. (a) Custom “Ultrathick Sectioning Testbed” utilizing nanometer precision linear stages (XMS50 and GTS30V, Newport) to move the tissue block in the horizontal plane across a stationary knife which is heated via the same jig as in Supplementary Figure 2a. The piezo of the ultrasonic knife is replaced with a high-load piezo (P-010.00H, Physik Instrumente) driven by a 200V amplifier (EPA-104, Piezo Systems). A laser vibrometer (CLV-2534, Polytec) is used to monitor the peak-to-peak vibrational amplitude of the knife (typically ~0.5 μm). Cutting is recorded via two video microscopes providing side and top views of the knife. Four force sensors (208C01, PCB Piezotronics) in the tissue block mount measure normal and transverse cutting forces during sectioning (typically 1000mN transverse peak). Force measurements are helpful in monitoring for chatter. (b) Example tissue block containing fly brain trimmed for ultrathick sectioning. Block is trimmed to have a rectangular face approximately 1.5 mm wide and 3 mm long, providing a large enough section with ample blank regions for manipulation with vacuum tweezers and forceps. The side of the block which will be the trailing edge during hot knife sectioning is strengthened by giving it a ~45° slope. (c) Side view of knife cutting a 20 μm thick section. (See Supplementary Video 1 for video of sectioning process.) (d) Same view but with block's profile outlined in yellow, ultrathick section outlined in blue, and knife angles shown. (e) Side view of knife showing how sections remain adhered to top surface of knife after sectioning. (Scale bar ~1 mm)

  7. Overview of volume stitching.
    Supplementary Fig. 5: Overview of volume stitching.

    (a) FIB-SEM stacks were taken of corresponding regions of the cut surfaces of two sequentially-cut ultrathick sections (here two 20 μm ultrathick sections of a fly larva). These separate stacks are then “volume stitched” into a single volume (b). The dashed red line in (b) is the 'stitch' plane between the two algorithmically flattened FIB-SEM stacks. Purple, blue, and green bordered images show cut planes through this stitched volume which in this case was centered on a bundle of neuronal processes (see Supplementary Figures 6 and 7 for additional larva results). (c) This diagram shows the key algorithmic steps in the volume stitching procedure (see numbered text for description of each step).

  8. Algorithmic surface-flattening example (fly larva).
    Supplementary Fig. 6: Algorithmic surface-flattening example (fly larva).

    (a) Schematic (left) of how two sequential ultrathick sections of a Drosophila larva (wandering 3rd instar phase) were re-embedded, trimmed, and mounted for FIB-SEM imaging. Low energy BSE surface images (right) of the two ultrathick sections FIB cross sectioned at near the same tissue level. (b) Inset images zoomed in on the matched surfaces. Scale bar = 10 μm. (c) One image of a FIB-SEM stack acquired at the surface of the bottom ultrathick section. Scale bar = 1 μm. (d) Surface extraction algorithm applied to this image. Red curve shows boundary determined by pixel column threshold technique. This red curve overshoots the real surface in many places because re-embedding plastic has same electron signal as blank cytoplasm. Green curve is smoothed estimate of actual hot knife cut surface. (e) Same image after being algorithmically flattened by shifting each pixel column up to make estimated surface curve flat. (f) 3D rendering of whole stack after flattening. Hot knife cut surface is rendered in front outlined by dashed red lines.

  9. Volume stitching example (fly larva).
    Supplementary Fig. 7: Volume stitching example (fly larva).

    (a) Localized FIB-SEM stacks were acquired (locations outlined in blue and yellow) centered on a bundle of neuronal processes crossing the hot knife boundary between sequentially-cut sections #7 and #8 of same larva tissue shown in Supplementary Figure 6. (b) Volume stitched dataset. The dashed red line on the 3D volume shows the stitch plane between the two algorithmically flattened FIB-SEM stacks. Colored borders correspond to cut planes through this 3D stitched volume which are shown in (c). (d) Graphical depiction of the volume which has been split open along the stitch plane. All processes in the densely packed central region were traced across the gap (denoted by colored outlines).

  10. Volume-stitching results on HPF-FS fly brain tissue.
    Supplementary Fig. 8: Volume-stitching results on HPF-FS fly brain tissue.

    (a) Graphical depiction (based on reference light micrographs) of two sequential 20 μm ultrathick sections (#17 and #18) through the optic lobe of an adult Drosophila brain prepared via HPF-FS. (b) FIB-SEM images of the two sections showing quality of cut surfaces. These were cut in oil using the heated ultrasonic jig shown in Supplementary Fig. 2a, and prepared for FIB-SEM imaging as in Figure 1b-f. (Scale bar 10 μm) (see Supplementary Video 2-3) (c) Volume stitching results of a 32x32x32 nm voxel size collapsed stack spanning the entire lobula region (outlined in red in (a)). The dashed red line on the 3D volume in (c) shows the 'stitch' plane between the two algorithmically flattened FIB-SEM stacks. Images with colored borders correspond to cut planes through this 3D stitched volume (see Supplementary Video 4-5). (d-f) Data from a smaller dataset which was volume stitched at native resolution (8x8x8 nm). (d) Graphical depiction of the stitched volume, split open along the stitch plane. (e) Graphical depiction of the stitched volume with dashed red line showing the location of the stitch plane between the two algorithmically flattened FIB-SEM stacks. (f) Images with colored borders show cut planes through correspondingly-colored planes in (e) along with zoomed insets of each (see Supplementary Video 6).

  11. Volume-stitching results on mouse cortical tissue.
    Supplementary Fig. 9: Volume-stitching results on mouse cortical tissue.

    (a) FIB-SEM images of two sequential 20 μm ultrathick sections (#21, #22) cut from a 100 μm thick vibratome slice (see Supplementary Video 7) (scale bar 10 μm). Also shown are light micrographs of each corresponding sample tab. (b) An 8x8 μm cropped region which has been split open along the stitch plane. Processes in the densely packed central region (total of 200) were traced across the gap (denoted by colored outlines, Supplementary Video 9). (c) Volume stitching over a 9x16 μm cropped region. The dashed red line on the 3D volume shows the stitch plane. Images with colored borders correspond to cut planes through this 3D stitched volume (Supplementary Video 8) (scale bar 1 μm).

  12. Staining quality of synaptic features in C-PLT fly brain tissue.
    Supplementary Fig. 10: Staining quality of synaptic features in C-PLT fly brain tissue.

    (a) FIB-SEM image of 20 μm ultrathick section through the medulla region of the same C-PLT fly brain which produced sections #34 and #35 shown in Figure 2. (b) Light micrograph of this section tab. (c) High resolution FIB-SEM images of this section (acquired at 8x8x8 nm under same imaging conditions as in Figure 2, see Online Methods) showing clearly discernable synaptic features. Red arrows denote T-bars, yellow arrows denote PSDs (post synaptic densities). Note that not all features of these synapses are visible in this single image plane but instead require inspection of the full 3D FIB-SEM stack encompassing the synapse.

  13. Comparison of image 'jump' across the hot-knife cut to faux gaps of various sizes.
    Supplementary Fig. 11: Comparison of image 'jump' across the hot-knife cut to faux gaps of various sizes.

    Red-green overlay images comparing image 'jump' across the hot knife cut to faux gaps of various sizes. (see Online Methods for details) (a) Red-green overlay images corresponding to mouse cortex volume stitch shown in Supplementary Figure 9. (b) Red-green overlay images corresponding to C-PLT fly brain volume stitch shown in Figure 2. (c) Quantification of tissue loss at hot knife cut for C-PLT fly brain volume stitch shown in Figure 2. (see Online Methods for details).

  14. Examples of neurite tracing across volume stitch in mouse brain tissue.
    Supplementary Fig. 12: Examples of neurite tracing across volume stitch in mouse brain tissue.

    Insets (a-d) zoom in on selected regions of the volume stitched hot knife cut face of the mouse brain tissue shown in (e). Next to each inset is a cross sectional image centered on the same region showing the line of the volume stitch. Only three images on either side of this volume stitch line are colored. (a) Inset and cross section showing a synapse split in two by the hot knife cut plane. (b-d) Inset and cross section images showing various small neural processes crossing the stitch plane at a variety of angles. To examine these and other regions in 3D one can load Supplementary Video 9 into the Fiji software.

  15. Examples of neurite tracing across volume stitch in C-PLT fly brain.
    Supplementary Fig. 13: Examples of neurite tracing across volume stitch in C-PLT fly brain.

    Insets (a-d) zoom in on selected regions of the volume stitched hot knife cut face of the C-PLT fly brain shown in (e). Next to each inset is a cross sectional image centered on the same region showing the line of the volume stitch. Only three images on either side of this volume stitch line are colored. (a) Inset and cross section showing an “easy” tracing case where a relatively large (green) process (~200 nm) crosses perpendicular to the stitch plane. (b) Inset and cross section showing more difficult case where a small (cyan) process (~70nm) crosses almost parallel to the stitch plane. (c) Inset and cross section showing another difficult case where a group of small dendritic processes squeeze together at a synapse which happens to be very close to the stitch plane. (d) Inset and cross section showing intermediate-sized processes traveling parallel to the stitch plane. To examine these and other regions in 3D one can load Supplementary Video 11 into the Fiji software.

  16. Ultrathick sectioning for reducing large blocks of embedded tissue.
    Supplementary Fig. 14: Ultrathick sectioning for reducing large blocks of embedded tissue.

    Ultrathick sectioning can in principle be used to reduce even very large blocks of embedded tissue (perhaps even an entire mouse brain) to sub-blocks optimally-sized for re-sectioning and imaging by any of today's connectomics technologies. Ultrathick sections cut along one plane through the tissue are re-embedded with blank spacers and cut in an orthogonal plane (see Figure 1i for example). Repeating this reduces the original volume to slabs filled with cubes of tissue separated from each other by blank resin. Individual cubes can be extracted for imaging in any order desired, creating a random-access “library” of tissue samples. Besides optimizing sample size and allowing parallel imaging across multiple technologies, such a library would also allow more precise targeting of imaging resources within a larger tissue volume. This would allow for follow-up studies to extend a connectome imaged and traced in one region into neighboring regions.

  17. Post-staining of ultrathick sections of mouse cortex with hot ethanolic uranyl acetate (UA).
    Supplementary Fig. 15: Post-staining of ultrathick sections of mouse cortex with hot ethanolic uranyl acetate (UA).

    Note that this mouse cortex tissue was not prepared for block face imaging and therefore showed insufficient contrast in the FIB-SEM (as opposed to the mouse cortex tissue shown in Fig. 1g-i and Supplementary Fig. 9 which used an en bloc staining protocol sufficient for direct FIB-SEM imaging, see Online Methods). Here we tested whether post staining the thick section could boost the FIB-SEM contrast similar to how post staining is used to boost contrast of ultrathin sections for TEM. (a) Full images of the FIB-milled surfaces of two 20 μm ultrathick sections (un-stained vs. post-stained) taken under the same SEM brightness and contrast settings to allow for direct comparison of staining levels. The post-stained section was post-stained with hot ethanolic UA (2%) at 60°C for 2 hours. Note the increase in overall signal level and contrast relative to the unstained control section. Also note that this increase in signal and contrast appears to be uniformly present across the full depth of the 20 μm thick section. (b) Histograms of pixel grayscale values of beam blanked, unstained, and post-stained high resolution images. Note that the histogram of the post-stained hot knife section shows increased signal as well as an overall contrast broadening. (c) Contrast inverted images (6nm pixel size, same contrast and brightness) centered on large synaptic contacts with clearly demarcated PSDs. (1) is from the unstained section (4) is from the post-stained section. (2,5) Same images thresholded to highlight membranes (i.e. threshold set at the grayscale value midway between the average cytoplasmic grayscale value and the average membrane grayscale value.) (3,6) Same images thresholded to highlight PSDs (i.e. threshold set at the grayscale value midway between the average membrane grayscale value and the average PSD grayscale value.) Comparing (3) vs. (6) one can see that PSDs in unstained tissue can only be distinguished by their morphological thickening whereas in the post-stained tissue they are distinguishable by their enhanced signal level as well. (d) FIB-SEM images of post-stained ultrathick section. Note how hot ethanol treatment has caused large cellular processes to bulge out from cut plane.

Videos

  1. Video of ultrathick sectioning
    Video 1: Video of ultrathick sectioning
    The video shows side and top video microscope views of the operation of the Ultrathick Sectioning Testbed (Supplementary Figure 4a) cutting a series of 20 μm ultrathick sections off a block of mouse cortex tissue. This video is speedup relative to actual sectioning (actual sectioning speed is 0.1mm/s).
  2. Unregistered FIB-SEM stack covering full area of ultrathick section 17 (HPF-FS fly brain)
    Video 2: Unregistered FIB-SEM stack covering full area of ultrathick section #17 (HPF-FS fly brain)
    The video shows unregistered FIB-SEM overview images covering a 123 × 32 × 82 μm volume. See Supplementary Figure 8a-c for volume stitching results for this tissue section.
  3. Unregistered FIB-SEM stack covering full area of ultrathick section 18 (HPF-FS fly brain)
    Video 3: Unregistered FIB-SEM stack covering full area of ultrathick section #18 (HPF-FS fly brain)
    The video shows unregistered FIB-SEM overview images covering a 115 × 32 × 86 μm volume. See Supplementary Figure 8a-c for volume stitching results for this tissue section.
  4. Low resolution volume stitch of sections 17 and 18 (HPF-FS fly brain)
    Video 4: Low resolution volume stitch of sections #17 and #18 (HPF-FS fly brain)
    The video shows a low resolution (64 × 64 × 128 nm voxel size) volume stitch of two 20 μm ultrathick sections (#17 and #18) of HPF-FS fly brain covering the full area of the optic lobula region. See Supplementary Figure 8a-c.
  5. Low resolution volume stitch of sections 17 and 18 re-sliced parallel to hot knife cut (HPF-FS fly brain)
    Video 5: Low resolution volume stitch of sections #17 and #18 re-sliced parallel to hot knife cut (HPF-FS fly brain)
    The video shows a low resolution (64 × 64 × 64 nm voxel size) volume stitch of two 20 μm ultrathick sections (#17 and #18, same as in Supplementary Video #4) of HPF-FS fly brain covering the full area of the optic lobula region. Volume has been digitally re-sliced to be parallel to the hot knife cut plane. See Supplementary Figure 8a-c.
  6. Volume stitch of sections 17 and 18 (HPF-FS fly brain)
    Video 6: Volume stitch of sections #17 and #18 (HPF-FS fly brain)
    The video shows a full resolution (8 × 8 × 8 nm voxel size) volume stitch across the hot knife cut between sections #17 and #18 in the optical medulla region. See Supplementary Figure 8d-f.
  7. FIB-SEM stack of hot knife cut mouse cortex tissue zoomed on cut edge
    Video 7: FIB-SEM stack of hot knife cut mouse cortex tissue zoomed on cut edge
    The video shows a (10 × 10 × 10 nm voxel size) FIB-SEM stack zoomed in on the hot knife cut edge of a 20 μm ultrathick section of mouse cortex tissue.
  8. Volume stitch of sections 21 and 22 (mouse cortex tissue)
    Video 8: Volume stitch of sections #21 and #22 (mouse cortex tissue)
    The video shows a full resolution (10 × 10 × 10 nm voxel size) volume stitch across the hot knife cut between ultrathick sections #21 and #22. See Supplementary Figure 9.
  9. Colored tracings across volume stitch of sections 21 and 22 (mouse cortex tissue)
    Video 9: Colored tracings across volume stitch of sections #21 and #22 (mouse cortex tissue)
    The video shows the same FIB-SEM volume stitch as in Supplementary Video #8 (10 × 10 × 10 nm voxel size) but includes colored overlays denoting the neural processes traced across the hot knife cut boundary. These are the same colored tracings as are shown in Supplementary Figure 9b.
  10. Volume stitch of sections 34 and 35 (C-PLT fly brain)
    Video 10: Volume stitch of sections #34 and #35 (C-PLT fly brain)
    The video shows a full resolution (8 × 8 × 8 nm voxel size) volume stitch across the hot knife cut between ultrathick sections #34 and #35. See Figure 2.
  11. Colored tracings across volume stitch of sections 34 and 35 (C-PLT fly brain)
    Video 11: Colored tracings across volume stitch of sections #34 and #35 (C-PLT fly brain)
    The video shows same FIB-SEM volume stitch as in Supplementary Video #10 (8 × 8 × 8 nm voxel size) but includes colored overlays denoting the neural processes traced across the hot knife cut boundary. These are the same colored tracings as are shown in Figure 2g.
  12. Side-by-side movies before and after stitching (C-PLT fly brain)
    Video 12: Side-by-side movies before and after stitching (C-PLT fly brain)
    The video shows side-by-side movies of all slices of a 4x4x4 μm volume (8 × 8 × 8 nm voxel size) of the C-PLT volume stitch (ultrathick sections #34 and #35) both before and after algorithmic flattening and stitching as well as after volume tracing. This is provided to allow direct comparison of the hot knife edge quality (before flattening and stitching) to the stitch quality (after flattening and stitching). (See Figure 2c-f)

References

  1. Knott, G., Marchman, H., Wall, D. & Lich, B. J. Neurosci. 28, 29592964 (2008).
  2. Xu, C.S. & Hess, H. Microsc. Microanal. 17, 664665 (2011).
  3. Briggman, K.L. & Bock, D.D. Curr. Opin. Neurobiol. 22, 154161 (2012).
  4. Harris, K.M. et al. J. Neurosci. 26, 1210112103 (2006).
  5. Bock, D.D. et al. Nature 471, 177182 (2011).
  6. Hayworth, K.J. et al. Front. Neural Circuits 8, 68 (2014).
  7. Denk, W. & Horstmann, H. PLoS Biol. 2, e329 (2004).
  8. Denk, W., Briggman, K.L. & Helmstaedter, M. Nat. Rev. Neurosci. 13, 351358 (2012).
  9. Morgan, J.L. & Lichtman, J.W. Nat. Methods 10, 494500 (2013).
  10. Hayat, M.A. Principles and Techniques of Electron Microscopy 4th edn. (Cambridge University Press, 2000).
  11. McGee-Russell, S.M. & Gosztonyi, G. Nature 214, 12041206 (1967).
  12. McGee-Russell, S.M., De Bruijn, W.C. & Gosztonyi, G. J. Neurocytol. 19, 655661 (1990).
  13. West, R.W. Stain Technol. 47, 201204 (1972).
  14. Studer, D. & Gnaegi, H. J. Microsc. 197, 94100 (2000).
  15. Meinertzhagen, I.A. & O'Neil, S.D. J. Comp. Neurol. 305, 232263 (1991).
  16. Walther, P. & Ziegler, A. J. Microsc. 208, 310 (2002).
  17. Seligman, A.M., Wasserkrug, H.L. & Hanker, J.S. J. Cell Biol. 30, 424432 (1966).
  18. Tapia, J.C. et al. Nat. Protoc. 7, 193206 (2012).
  19. Locke, M. & Krishnan, N. J. Cell Biol. 50, 550557 (1971).
  20. Schindelin, J. et al. Nat. Methods 9, 676682 (2012).
  21. Cardona, A. et al. PLoS ONE 7, e38011 (2012).
  22. Sommer, C., Strähle, C., Köthe, U. & Hamprecht, F.A. in IEEE Int. Symp. Biomed. Imaging 230233 (2011).

Download references

Author information

Affiliations

  1. Janelia Research Campus, Howard Hughes Medical Institute, Ashburn, Virginia, USA.

    • Kenneth J Hayworth,
    • C Shan Xu,
    • Richard D Fetter &
    • Harald F Hess
  2. Department of Psychology, Dalhousie University, Halifax, Nova Scotia, Canada.

    • Zhiyuan Lu
  3. Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland.

    • Graham W Knott
  4. Department of Neuroscience, Columbia University Medical Center, New York, New York, USA.

    • Juan Carlos Tapia
  5. Department of Molecular and Cell Biology, Harvard University, Cambridge, Massachusetts, USA.

    • Jeff W Lichtman

Contributions

K.J.H. conceived of, designed and performed the experiments and wrote the manuscript. C.S.X., H.F.H. and K.J.H. constructed the custom FIB-SEM systems. C.S.X. designed and implemented the custom FIB-SEM control hardware and software. Z.L. developed the C-PLT technique and prepared the C-PLT tissue. G.W.K. prepared mouse cortex tissue for FIB-SEM (Fig. 1 and Supplementary Figs. 9 and 12). J.C.T. prepared mouse tissue for earlier studies (Supplementary Figs. 1,2,3 and 15). R.D.F. prepared larva tissue. R.D.F. and Z.L. prepared HPF-FS tissue. Early research (Supplementary Figs. 1,2,3) was performed in the laboratory of J.W.L. All FIB-SEM research was performed in the laboratory of H.F.H.

Competing financial interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to:

Author details

Supplementary information

Supplementary Figures

  1. Supplementary Figure 1: Early thick sectioning tests showing surface damage. (655 KB)

    (a) SEM cross section of a vibrotome cut surface showing extensive ultrastructure damage. Scale bar = 5 μm. (b, c) Diamond knife boat heater clamp used for hot knife thick sectioning. (d) Schematic showing how hot knife cut sections were mounted for SEM imaging of their sectioned surfaces. (e) SEM BSE (backscatter electron) image (taken at 6kV) of a sequential pair of 20 μm thick hot knife sections mounted so that their matching faces are facing up. Scale bar = 100 μm. (f) BSE image zooming in on the tip region of one of the sections shown in (e). This BSE image may appear to be damage free, but when the same region is imaged using SE (secondary electron) imaging (g) considerable surface damage is evident. Scale bars = 1 μm. (h,i) BSE and SE images respectively of a separate hot knife cut section. Chatter damage is visible at bottom. Surface damage is clearly seen in the SE image (i) across the entire section with the exception of a triangular region at the very center of the tissue. This same triangular region is seen to contain less stain contrast in (h) supporting the hypothesis that some surface damage was due to the overly brittle nature of this ROTO prepared tissue in those regions of the tissue where this ROTO procedure fully penetrated. (see Online Methods for more details.) Scale bars = 10 μm.

  2. Supplementary Figure 2: Improved hot knife–sectioning technique results. (739 KB)

    (a) Heating jig designed for use with the Diatome ultrasonic diamond knife to allow ultrathick sectioning with a commercial ultramicrotome. The ultramicrotome's knife mount is used to clamp the heating jig and knife base in metal-to-metal contact. Three small insulators made from rigid circuit board material help insulate the knife and heating jig. A PID controller (Omega) reads the thermocouple to regulate temperature. (b) BSE image (taken at 6kV) of a sequential pair of 20 μm thick hot knife sections mounted so that their matching faces are facing up. (c) Same sections imaged with SE. Scale bars = 100 μm, (d,e) Zoomed regions of previous imaged in SE at higher magnification. Scale bars = 50 μm, (f,g) Inset regions of previous imaged in SE at higher magnification. Scale bars = 10 μm. Hot knife cut surfaces cut with this improved technique look exceptionally free of damage over a wide area and a wide range of imaging scales.

  3. Supplementary Figure 3: Matching ultrastructure profiles across hot-knife cut. (637 KB)

    (a) High magnification low voltage (1.5kV) SE image of the surface of a 20 μm thick hot knife section showing small neuronal processes cut by the heated knife. (b) Mirror reversed SE image of the matching region of the sequentially cut hot knife section. (c,d) Same images colored to show matching neuronal profiles identified across the hot knife cut. Scale bar = 1 μm

  4. Supplementary Figure 4: Ultrathick sectioning apparatus and procedure. (693 KB)

    Current ultrathick sectioning apparatus and procedure. (a) Custom “Ultrathick Sectioning Testbed” utilizing nanometer precision linear stages (XMS50 and GTS30V, Newport) to move the tissue block in the horizontal plane across a stationary knife which is heated via the same jig as in Supplementary Figure 2a. The piezo of the ultrasonic knife is replaced with a high-load piezo (P-010.00H, Physik Instrumente) driven by a 200V amplifier (EPA-104, Piezo Systems). A laser vibrometer (CLV-2534, Polytec) is used to monitor the peak-to-peak vibrational amplitude of the knife (typically ~0.5 μm). Cutting is recorded via two video microscopes providing side and top views of the knife. Four force sensors (208C01, PCB Piezotronics) in the tissue block mount measure normal and transverse cutting forces during sectioning (typically 1000mN transverse peak). Force measurements are helpful in monitoring for chatter. (b) Example tissue block containing fly brain trimmed for ultrathick sectioning. Block is trimmed to have a rectangular face approximately 1.5 mm wide and 3 mm long, providing a large enough section with ample blank regions for manipulation with vacuum tweezers and forceps. The side of the block which will be the trailing edge during hot knife sectioning is strengthened by giving it a ~45° slope. (c) Side view of knife cutting a 20 μm thick section. (See Supplementary Video 1 for video of sectioning process.) (d) Same view but with block's profile outlined in yellow, ultrathick section outlined in blue, and knife angles shown. (e) Side view of knife showing how sections remain adhered to top surface of knife after sectioning. (Scale bar ~1 mm)

  5. Supplementary Figure 5: Overview of volume stitching. (312 KB)

    (a) FIB-SEM stacks were taken of corresponding regions of the cut surfaces of two sequentially-cut ultrathick sections (here two 20 μm ultrathick sections of a fly larva). These separate stacks are then “volume stitched” into a single volume (b). The dashed red line in (b) is the 'stitch' plane between the two algorithmically flattened FIB-SEM stacks. Purple, blue, and green bordered images show cut planes through this stitched volume which in this case was centered on a bundle of neuronal processes (see Supplementary Figures 6 and 7 for additional larva results). (c) This diagram shows the key algorithmic steps in the volume stitching procedure (see numbered text for description of each step).

  6. Supplementary Figure 6: Algorithmic surface-flattening example (fly larva). (730 KB)

    (a) Schematic (left) of how two sequential ultrathick sections of a Drosophila larva (wandering 3rd instar phase) were re-embedded, trimmed, and mounted for FIB-SEM imaging. Low energy BSE surface images (right) of the two ultrathick sections FIB cross sectioned at near the same tissue level. (b) Inset images zoomed in on the matched surfaces. Scale bar = 10 μm. (c) One image of a FIB-SEM stack acquired at the surface of the bottom ultrathick section. Scale bar = 1 μm. (d) Surface extraction algorithm applied to this image. Red curve shows boundary determined by pixel column threshold technique. This red curve overshoots the real surface in many places because re-embedding plastic has same electron signal as blank cytoplasm. Green curve is smoothed estimate of actual hot knife cut surface. (e) Same image after being algorithmically flattened by shifting each pixel column up to make estimated surface curve flat. (f) 3D rendering of whole stack after flattening. Hot knife cut surface is rendered in front outlined by dashed red lines.

  7. Supplementary Figure 7: Volume stitching example (fly larva). (636 KB)

    (a) Localized FIB-SEM stacks were acquired (locations outlined in blue and yellow) centered on a bundle of neuronal processes crossing the hot knife boundary between sequentially-cut sections #7 and #8 of same larva tissue shown in Supplementary Figure 6. (b) Volume stitched dataset. The dashed red line on the 3D volume shows the stitch plane between the two algorithmically flattened FIB-SEM stacks. Colored borders correspond to cut planes through this 3D stitched volume which are shown in (c). (d) Graphical depiction of the volume which has been split open along the stitch plane. All processes in the densely packed central region were traced across the gap (denoted by colored outlines).

  8. Supplementary Figure 8: Volume-stitching results on HPF-FS fly brain tissue. (523 KB)

    (a) Graphical depiction (based on reference light micrographs) of two sequential 20 μm ultrathick sections (#17 and #18) through the optic lobe of an adult Drosophila brain prepared via HPF-FS. (b) FIB-SEM images of the two sections showing quality of cut surfaces. These were cut in oil using the heated ultrasonic jig shown in Supplementary Fig. 2a, and prepared for FIB-SEM imaging as in Figure 1b-f. (Scale bar 10 μm) (see Supplementary Video 2-3) (c) Volume stitching results of a 32x32x32 nm voxel size collapsed stack spanning the entire lobula region (outlined in red in (a)). The dashed red line on the 3D volume in (c) shows the 'stitch' plane between the two algorithmically flattened FIB-SEM stacks. Images with colored borders correspond to cut planes through this 3D stitched volume (see Supplementary Video 4-5). (d-f) Data from a smaller dataset which was volume stitched at native resolution (8x8x8 nm). (d) Graphical depiction of the stitched volume, split open along the stitch plane. (e) Graphical depiction of the stitched volume with dashed red line showing the location of the stitch plane between the two algorithmically flattened FIB-SEM stacks. (f) Images with colored borders show cut planes through correspondingly-colored planes in (e) along with zoomed insets of each (see Supplementary Video 6).

  9. Supplementary Figure 9: Volume-stitching results on mouse cortical tissue. (429 KB)

    (a) FIB-SEM images of two sequential 20 μm ultrathick sections (#21, #22) cut from a 100 μm thick vibratome slice (see Supplementary Video 7) (scale bar 10 μm). Also shown are light micrographs of each corresponding sample tab. (b) An 8x8 μm cropped region which has been split open along the stitch plane. Processes in the densely packed central region (total of 200) were traced across the gap (denoted by colored outlines, Supplementary Video 9). (c) Volume stitching over a 9x16 μm cropped region. The dashed red line on the 3D volume shows the stitch plane. Images with colored borders correspond to cut planes through this 3D stitched volume (Supplementary Video 8) (scale bar 1 μm).

  10. Supplementary Figure 10: Staining quality of synaptic features in C-PLT fly brain tissue. (409 KB)

    (a) FIB-SEM image of 20 μm ultrathick section through the medulla region of the same C-PLT fly brain which produced sections #34 and #35 shown in Figure 2. (b) Light micrograph of this section tab. (c) High resolution FIB-SEM images of this section (acquired at 8x8x8 nm under same imaging conditions as in Figure 2, see Online Methods) showing clearly discernable synaptic features. Red arrows denote T-bars, yellow arrows denote PSDs (post synaptic densities). Note that not all features of these synapses are visible in this single image plane but instead require inspection of the full 3D FIB-SEM stack encompassing the synapse.

  11. Supplementary Figure 11: Comparison of image 'jump' across the hot-knife cut to faux gaps of various sizes. (478 KB)

    Red-green overlay images comparing image 'jump' across the hot knife cut to faux gaps of various sizes. (see Online Methods for details) (a) Red-green overlay images corresponding to mouse cortex volume stitch shown in Supplementary Figure 9. (b) Red-green overlay images corresponding to C-PLT fly brain volume stitch shown in Figure 2. (c) Quantification of tissue loss at hot knife cut for C-PLT fly brain volume stitch shown in Figure 2. (see Online Methods for details).

  12. Supplementary Figure 12: Examples of neurite tracing across volume stitch in mouse brain tissue. (506 KB)

    Insets (a-d) zoom in on selected regions of the volume stitched hot knife cut face of the mouse brain tissue shown in (e). Next to each inset is a cross sectional image centered on the same region showing the line of the volume stitch. Only three images on either side of this volume stitch line are colored. (a) Inset and cross section showing a synapse split in two by the hot knife cut plane. (b-d) Inset and cross section images showing various small neural processes crossing the stitch plane at a variety of angles. To examine these and other regions in 3D one can load Supplementary Video 9 into the Fiji software.

  13. Supplementary Figure 13: Examples of neurite tracing across volume stitch in C-PLT fly brain. (512 KB)

    Insets (a-d) zoom in on selected regions of the volume stitched hot knife cut face of the C-PLT fly brain shown in (e). Next to each inset is a cross sectional image centered on the same region showing the line of the volume stitch. Only three images on either side of this volume stitch line are colored. (a) Inset and cross section showing an “easy” tracing case where a relatively large (green) process (~200 nm) crosses perpendicular to the stitch plane. (b) Inset and cross section showing more difficult case where a small (cyan) process (~70nm) crosses almost parallel to the stitch plane. (c) Inset and cross section showing another difficult case where a group of small dendritic processes squeeze together at a synapse which happens to be very close to the stitch plane. (d) Inset and cross section showing intermediate-sized processes traveling parallel to the stitch plane. To examine these and other regions in 3D one can load Supplementary Video 11 into the Fiji software.

  14. Supplementary Figure 14: Ultrathick sectioning for reducing large blocks of embedded tissue. (312 KB)

    Ultrathick sectioning can in principle be used to reduce even very large blocks of embedded tissue (perhaps even an entire mouse brain) to sub-blocks optimally-sized for re-sectioning and imaging by any of today's connectomics technologies. Ultrathick sections cut along one plane through the tissue are re-embedded with blank spacers and cut in an orthogonal plane (see Figure 1i for example). Repeating this reduces the original volume to slabs filled with cubes of tissue separated from each other by blank resin. Individual cubes can be extracted for imaging in any order desired, creating a random-access “library” of tissue samples. Besides optimizing sample size and allowing parallel imaging across multiple technologies, such a library would also allow more precise targeting of imaging resources within a larger tissue volume. This would allow for follow-up studies to extend a connectome imaged and traced in one region into neighboring regions.

  15. Supplementary Figure 15: Post-staining of ultrathick sections of mouse cortex with hot ethanolic uranyl acetate (UA). (629 KB)

    Note that this mouse cortex tissue was not prepared for block face imaging and therefore showed insufficient contrast in the FIB-SEM (as opposed to the mouse cortex tissue shown in Fig. 1g-i and Supplementary Fig. 9 which used an en bloc staining protocol sufficient for direct FIB-SEM imaging, see Online Methods). Here we tested whether post staining the thick section could boost the FIB-SEM contrast similar to how post staining is used to boost contrast of ultrathin sections for TEM. (a) Full images of the FIB-milled surfaces of two 20 μm ultrathick sections (un-stained vs. post-stained) taken under the same SEM brightness and contrast settings to allow for direct comparison of staining levels. The post-stained section was post-stained with hot ethanolic UA (2%) at 60°C for 2 hours. Note the increase in overall signal level and contrast relative to the unstained control section. Also note that this increase in signal and contrast appears to be uniformly present across the full depth of the 20 μm thick section. (b) Histograms of pixel grayscale values of beam blanked, unstained, and post-stained high resolution images. Note that the histogram of the post-stained hot knife section shows increased signal as well as an overall contrast broadening. (c) Contrast inverted images (6nm pixel size, same contrast and brightness) centered on large synaptic contacts with clearly demarcated PSDs. (1) is from the unstained section (4) is from the post-stained section. (2,5) Same images thresholded to highlight membranes (i.e. threshold set at the grayscale value midway between the average cytoplasmic grayscale value and the average membrane grayscale value.) (3,6) Same images thresholded to highlight PSDs (i.e. threshold set at the grayscale value midway between the average membrane grayscale value and the average PSD grayscale value.) Comparing (3) vs. (6) one can see that PSDs in unstained tissue can only be distinguished by their morphological thickening whereas in the post-stained tissue they are distinguishable by their enhanced signal level as well. (d) FIB-SEM images of post-stained ultrathick section. Note how hot ethanol treatment has caused large cellular processes to bulge out from cut plane.

Video

  1. Video 1: Video of ultrathick sectioning (35.35 MB, Download)
    The video shows side and top video microscope views of the operation of the Ultrathick Sectioning Testbed (Supplementary Figure 4a) cutting a series of 20 μm ultrathick sections off a block of mouse cortex tissue. This video is speedup relative to actual sectioning (actual sectioning speed is 0.1mm/s).
  2. Video 2: Unregistered FIB-SEM stack covering full area of ultrathick section #17 (HPF-FS fly brain) (40.49 MB, Download)
    The video shows unregistered FIB-SEM overview images covering a 123 × 32 × 82 μm volume. See Supplementary Figure 8a-c for volume stitching results for this tissue section.
  3. Video 3: Unregistered FIB-SEM stack covering full area of ultrathick section #18 (HPF-FS fly brain) (40.85 MB, Download)
    The video shows unregistered FIB-SEM overview images covering a 115 × 32 × 86 μm volume. See Supplementary Figure 8a-c for volume stitching results for this tissue section.
  4. Video 4: Low resolution volume stitch of sections #17 and #18 (HPF-FS fly brain) (108.83 MB, Download)
    The video shows a low resolution (64 × 64 × 128 nm voxel size) volume stitch of two 20 μm ultrathick sections (#17 and #18) of HPF-FS fly brain covering the full area of the optic lobula region. See Supplementary Figure 8a-c.
  5. Video 5: Low resolution volume stitch of sections #17 and #18 re-sliced parallel to hot knife cut (HPF-FS fly brain) (35.14 MB, Download)
    The video shows a low resolution (64 × 64 × 64 nm voxel size) volume stitch of two 20 μm ultrathick sections (#17 and #18, same as in Supplementary Video #4) of HPF-FS fly brain covering the full area of the optic lobula region. Volume has been digitally re-sliced to be parallel to the hot knife cut plane. See Supplementary Figure 8a-c.
  6. Video 6: Volume stitch of sections #17 and #18 (HPF-FS fly brain) (88.25 MB, Download)
    The video shows a full resolution (8 × 8 × 8 nm voxel size) volume stitch across the hot knife cut between sections #17 and #18 in the optical medulla region. See Supplementary Figure 8d-f.
  7. Video 7: FIB-SEM stack of hot knife cut mouse cortex tissue zoomed on cut edge (81.77 MB, Download)
    The video shows a (10 × 10 × 10 nm voxel size) FIB-SEM stack zoomed in on the hot knife cut edge of a 20 μm ultrathick section of mouse cortex tissue.
  8. Video 8: Volume stitch of sections #21 and #22 (mouse cortex tissue) (97.04 MB, Download)
    The video shows a full resolution (10 × 10 × 10 nm voxel size) volume stitch across the hot knife cut between ultrathick sections #21 and #22. See Supplementary Figure 9.
  9. Video 9: Colored tracings across volume stitch of sections #21 and #22 (mouse cortex tissue) (41.59 MB, Download)
    The video shows the same FIB-SEM volume stitch as in Supplementary Video #8 (10 × 10 × 10 nm voxel size) but includes colored overlays denoting the neural processes traced across the hot knife cut boundary. These are the same colored tracings as are shown in Supplementary Figure 9b.
  10. Video 10: Volume stitch of sections #34 and #35 (C-PLT fly brain) (92.07 MB, Download)
    The video shows a full resolution (8 × 8 × 8 nm voxel size) volume stitch across the hot knife cut between ultrathick sections #34 and #35. See Figure 2.
  11. Video 11: Colored tracings across volume stitch of sections #34 and #35 (C-PLT fly brain) (104.45 MB, Download)
    The video shows same FIB-SEM volume stitch as in Supplementary Video #10 (8 × 8 × 8 nm voxel size) but includes colored overlays denoting the neural processes traced across the hot knife cut boundary. These are the same colored tracings as are shown in Figure 2g.
  12. Video 12: Side-by-side movies before and after stitching (C-PLT fly brain) (79.29 MB, Download)
    The video shows side-by-side movies of all slices of a 4x4x4 μm volume (8 × 8 × 8 nm voxel size) of the C-PLT volume stitch (ultrathick sections #34 and #35) both before and after algorithmic flattening and stitching as well as after volume tracing. This is provided to allow direct comparison of the hot knife edge quality (before flattening and stitching) to the stitch quality (after flattening and stitching). (See Figure 2c-f)

PDF files

  1. Supplementary Text and Figures (19,726 KB)

    Supplementary Figures 1–15

Zip files

  1. Supplementary Software (22,901 KB)

    Custom software used to “volume stitch” FIB-SEM stacks of sequentially-cut hot knife sections

    The compressed folder contains Matlab source code for our “Hot Knife Edge Extractor Software” which is used to algorithmically flatten the cut surface to allow stitching to another flattened stack. Brief instructions for use and a small example dataset are also provided. The compressed folder also contains a Matlab script to assist in aligning and volume stitching flattened stacks together.

Additional data