Volumetric super-resolution imaging by serial ultrasectioning and STochastic Optical Reconstruction Microscopy (STORM) in neural tissue

Tarlan Vatan University of Maryland College Park https://orcid.org/0000-0001-8750-1382 Jacqueline A. Minehart University of Maryland College Park Chenghang Zhang University of Maryland College Park Vatsal Agarwal University of Maryland College Park Jerry Yang University of Maryland College Park Colenso M. Speer (  cspeer@umd.edu ) University of Maryland College Park https://orcid.org/0000-0002-3076-7072

Introduction for the immunohistochemical assay. Furthermore, this protocol requires technical skills in histology and ultrasectioning. Additionally, users should familiarize themselves with principles of single-molecule localization microscopy (SMLM) and operation of available commercial instrumentation.
Prepare uorescent dye conjugates for IHC (Timing: 2h) The following steps describe the amine-reactive conjugation of uorescent dyes to secondary antibodies and wheat germ agglutinin (WGA). The conjugates are used in this protocol to label primary antibodies in indirect IHC and endogenous glycoproteins in neuropil, respectively. Some dyes are reagents section (see also Dempsey et al., 2011;Li & Vaughan, 2018).
Additional considerations when selecting dyes for multicolor imaging can be found in Figure 10.
2. Add 1.8 μl reporter dye and 1.2 μl activator dye to the solution.
Note: Dual labeled antibodies are used to increase reporter dye photoswitching (Dy 749/Alexa 647) through UV illumination of the activator dye (Alexa 405). Cy3B and Atto 488 reporter photoswitching is not increased in the presence of activator dyes therefore no Alexa 405 is added to those conjugation reactions. 3. Vortex the solution brie y. 4. Incubate the solution at room temperature for 20 min with constant agitation and protect from light. Meanwhile, allow a NAP-5 column to drain its storage buffer into a waste receptacle. Fill the empty NAP-5 column with 1x PBS and allow the waste to drain.
Repeat washing two more times.
Critical: Rinse the NAP-5 column at least three times (Troubleshooting 1). 5. Pipette the conjugate solution onto the surface of the NAP-5 column and allow solution to sink into the column surface for ~30 seconds. 6. Add 600 μl 1x PBS to the column. Allow the clear PBS fraction to drain from the column into a waste receptacle.
Note: Free dyes will separate from conjugates creating two bands that may be visible depending on uorophore wavelength (Cy3B is especially visible, 488 and 647 are somewhat visible and 749 is very di cult to see). 7. Add 300 μl 1x PBS and collect the owthrough into the nal storage tube. Discard the column. 8. Measure conjugate concentration and dye ratios per antibody/WGA molecule: a. Blank the spectrophotometer using the same 1x PBS as above.
b. Measure the absorbance pro le of the conjugate and record the peak absorbance values of the antibody (~280 nm) and the dye molecules (~405-750 nm depending on dyes used). Save absorbance pro le for future reference.
c. Using the concentration values, determine the number of dye molecules per antibody molecule. A calculator used to determine the molecular ratios is provided in an Excel spreadsheet along with this manual.
Note: Ideally, each labeling protein molecule (IgG, IgY or WGA) will be conjugated to 3-4 reporter dye molecules and 1-2 activator dye molecules. If no activator is used, 3-4 reporter dye molecules should still be conjugated.
Critical: If ratios are not ideal, repeat the procedure and adjust dye volumes added to solution in step 2 to achieve molecular ratios within the appropriate bounds. If more than 4 μl of dye is needed in step 2 in order to achieve correct ratios, then dye stocks may need to be remade (Troubleshooting 1). d. Recollect conjugate from cuvette and store in the dark at 4°C for up to 6 months. e. To con rm conjugate quality after a period of storage, remeasure and compare absorbance values to originals.
Coat glass coverslips (Timing: 2h) The following steps are used to coat glass coverslips with gelatin, which will cause ultrathin sections to adhere when they are collected after ultrasectioning (Micheva & Smith, 2007). While performing all steps, avoid allowing dust to collect onto the surface of the coverslips. Work in an area free from heavy air ow and lter all solutions. Alternative: Place racks in a 50°C oven until dry.
Critical: Examine coverslips at this stage to check for clarity. If coverslips remain smudged or dusty, re lter 1M KOH and 95% EtOH, then start procedure again. 5. Completely submerge clean, dry coverslips in gelatin solution and swirl for 1 min. 6. Remove coverslip racks from gelatin solution and allow to air dry completely. Wick away excess solution from troughs of the coverslip rack.
Note: Moving coverslips to a clean, dry rack can help eliminate gelatin pooling at the lower edges of the coverslips. 7. Allow coverslips to dry completely before use or storage. Store dry coverslips in a closed container. Coverslips should be kept free from dust.
Collect and x the tissue sample (Timing: 4-5h) During the following steps, tissue will be dissected and prepared for IHC. Importantly, the region of interest is punched out creating a small disk of tissue which is easy to maneuver while minimizing mechanical damage. The small disk improves antibody penetration and allows the tissue to lay at for embedding. Fixative solution and blocking buffer should be prepared before you begin.
Note: These steps describe processing retinal tissue, but brain sections can also be used.
2. Fix tissue in room temperature 4% PFA for 10 min to 1 h. The xation conditions should be adjusted to optimize IHC labeling quality.
3. Wash retinae in 1x PBS for 2x 20 min, then place it in a clean petri dish lled with room temperature 1x PBS. 4. Punch out a 500 μm diameter region of interest using a 25G blunt ended needle under stereoscope guidance. 5. Transfer punch to a 24-well plate and incubate in blocking buffer for 3-4 hours at room temperature with constant agitation. Continue on to IHC. Immunohistochemistry (Timing: 4-5 days) The following steps are performed to label antigen targets with primary and secondary antibodies. Target-speci c staining is followed by a global neuropil stain using a WGA conjugate, which functions as an image alignment tool in later STORM data processing steps. Lastly, an aldehyde xation establishes crosslinking to stabilize all molecular labels. The following steps plasticize the tissue punches to make them suitable for ultrasectioning.
14. Perform a graded alcohol dehydration: a. Transfer tissue into a clean Eppendorf tube and immerse in a 50% EtOH solution. Incubate at room temperature for 20 min with constant agitation and protect sample from light. b. Repeat with 70%, 90%, 100% and 100% EtOH.
Critical: If tissue is not thoroughly dehydrated, the resin will not polymerize properly and crack or ake when taken out of the mold (Troubleshooting 2).

Perform a graded resin infusion:
a. Transfer tissue into a clean Eppendorf tube and immerse in a solution containing 2 parts 100% EtOH to 1 part resin. Incubate at room temperature for 2 h with constant agitation and protect sample from light.
2. Secure gelatin coated coverslip in plastic forceps and guide it into the histo Jumbo diamond knife water bath (as pictured in Figure 6).
3. Using an eyelash tool, guide sections onto the middle of the coverslip and slowly retract the coverslip from the water bath by turning the knob on the micromanipulator.
Caution: Leave su cient space in the margins of the coverslip to accommodate the imaging objective.
4. Warm coverslips with sections facing upward on a hot plate set to 60°C for 20 min.
Pause Point: Coverslips can be stored long-term prior to imaging.
Etch away resin and assemble ow chamber (Timing: 30 min) The following steps prepare tissue sections for optimal photoswitching during STORM imaging. Here, the resin surrounding individual sections is removed allowing uorophore exposure to the thiol-containing STORM imaging buffer. Fluorescent beads are spotted on the coverslips which will serve later as ducial markers. The ow chamber is assembled, lled with STORM imaging buffer, and sealed with epoxy. Prepare 10 % sodium ethoxide before you begin; sodium ethoxide solution can be reused for multiple experiments but should be remade after 1 month.
Critical: Following step 1, samples will be di cult to see. Note sample orientation on the coverslip.
2. Meanwhile, clean a glass slide with EtOH and allow to dry completely, then apply two layers of double-sided tape to the long sides of the slide. Only cover ~5 mm of the slide edge and shave off the overhanging tape using a razor blade.
3. Remove the coverslip using forceps and immediately wash with 95% EtOH.
Critical: Dulbecco's PBS must be used to prevent bead aggregation. b. Make low density bead solution by adding 750 μl DPBS to dense bead working stock. Vortex. Apply a small droplet (~1 μL) to coverslip.
Critical: Allow water to ow in the direction away from sample so that beads do not wash into the sample. 8. Dry coverslip using forced air. 9. Place the dry coverslip tissue-side-down onto the slide with double sided tape. Avoid crooked placement. Use a pipette tip to apply pressure on the tape sides to seal the coverslip to the adhesive.
10. Prepare STORM imaging buffer and mix by gentle pipetting. Avoid forcing air into the solution.
11. Draw up 160 μl of the STORM buffer into a pipette and slowly release it in the space between the coverslip and slide. Avoid creating bubbles by applying gentle pressure to the coverslip with a blunt object. Avoid cracking the coverslip when doing this.
12. Lay a piece of foil on the benchtop and push some epoxy out of the double-barrel syringe onto the foil. Mix with a pipette tip.
13. Apply a strip of epoxy to all sides of the coverslip (short sides rst) to seal the ow chamber. Place in a dark drawer to dry for 5 mins.
14. Test that the epoxy created a seal around the ow chamber by applying gentle pressure with a pipette tip to the surface of the coverslip. If no buffer leaks, then the seal is good. If some buffer leaks, then patch the hole using more, freshly mixed epoxy.
Critical: Be careful to not overapply epoxy and encroach on the sample area.
15. Apply a single strip of double-sided tape onto the short sides of the slide and trim off overhangs with razor blade. This is useful when securing slide to the STORM microscope. The preparation is ready for imaging.
Acquire STORM and conventional imaging data (Timing: 1-3 hours to set up, 1-3 minutes per STORM movie (automated acquisition)) The following steps describe our approach for collecting serial section conventional, wide-eld (diffraction-limited) and STORM data, together with ducial marker images for correcting chromatic aberration and uneven illumination of the sample. The key steps are to 1) collect conventional images of all color channels in rapid succession for the ROI(s) in each physical section of the sample. This limits sample drift between color channels for conventional images in each ROI. During subsequent image postprocessing/alignment, STORM images will be correlated to their conventional image counterparts to correct for sample drift.
2) Collect STORM movies for ROI(s) across all physical sections. For each ROI, the red emitters (Dy749P1/Alexa647) are imaged rst (sequentially within each ROI across the entire sample) as these dyes are susceptible to STORM buffer acidi cation (Dempsey et al., 2011). Then, Cy3B and Atto488 emitters are imaged sequentially for each ROI in a second image pass across the sample. Sample drift that occurs during each STORM movie acquisition is corrected as part of the single-molecule tting analysis process by back correlating binned movie frames to the start of the movie. 3) Collect ducial marker images that will later be used to correct chromatic aberration in conventional and STORM images.
Alternative: Commercial instrumentation and software may be used in lieu of custom instrumentation and the speci c codebases we reference here.
1. Power on the microscope and lasers.
2. Clean the oil immersion objective prior to imaging and position 4x or 10x objective for tile imaging.
3. Secure the sample slide in the stage holder.
Critical: Sample should be immobile once loaded in the stage. If necessary, use a top-side clamping stage insert or otherwise use double-sided tape applied to the glass slide to secure the specimen in the stage as described (previous section, step 15). Critical: Reduce laser power to low level to avoid image saturation and uorophore bleaching. a. Launch 'Steve.py' software and place cursor at or near the 0, 0 coordinates and press the spacebar to take an image.
b. Continue taking images of the eld in manual or semi-automated modes: i. manual imaging mode: click on an ROI and press the spacebar to take an image.
ii. semi-automated mode: click on an ROI and press keypad #3 or #9 for automated 3x3 or 9x9 image tiling, respectively. Press the spacebar to interrupt the sequence.
Note: User may navigate the sample view in Steve.py by right-clicking on an ROI and selecting "Go to Position." c. Finish mosaic and save. Adjust the stage position back to the sparse bead eld.

Create a high-magni cation mosaic of the bead elds and sample (serial sections):
a. Position high-magni cation oil/water objective (either 60x or 100x) for imaging.
b. Place a droplet of immersion oil on the objective (if using oil).
c. Focus on the sparse beads: i. Engage the high magni cation array setting in Hal.py.
ii. Using the 'Illumination' control panel in Hal.py, turn on a visible imaging laser (488nm or 561nm) and bring the sample into focus.
iii. Use the 'Focus lock' control panel to lock the focal plane.
Alternative: Use commercial instrument with automated focus maintenance system and image tiling software to generate a lowmag and high-mag overview of the sample eld and select ROIs for subsequent automated image acquisition. Alternative opensource software may also be used.
d. Set TIRF angle to maximize signal-to-noise at low illumination power. e. Reimage selected regions (both sample and beads) of the low magni cation mosaic at high magni cation.
Note: Steve.py allows users to correct offsets between low and high magni cation by adjusting the mosaic positions within the 'Objective Settings' panel.  c. Adjust laser powers using the 'Illumination' control panel for each channel to be imaged.
Critical: Avoid saturating dense bead images. Ensure su cient illumination intensity to use the full dynamic range of the camera.
14. Prepare illumination power settings for sparse bead eld image acquisition to ensure that images are not dim/saturated: a. Engage the 'Regbeads' setting in Hal.py.
b. Focus on a representative ROI within the sparse bead eld.
c. Adjust 750nm laser power using the 'Illumination' control panel to visualize IR beads in 750 and 647 channels. c. Adjust the 750nm laser power to maximum and the 405nm laser to minimum using the 'Illumination' control panel.
f. Click 'Record' to begin manual STORM movie acquisition.
g. Manually ramp 405nm laser power using the 'Illumination' control panel to reactivate uorophore switching to maintain optimal emitter density.
h. Pressing 'Stop' ends the movie recording and saves the power progression le.
i. Repeat steps 16 (a-h) for 647nm, 561nm, and 488nm STORM movies. Change the directory lenames for each movie according to wavelength using the format speci ed above.
Note: Cy3B (561nm) and Atto488 (488nm) probes do not require 405nm illumination for reactivation. Alternative: The above steps 16 (a-i) describe the setting of custom 'power progression' les to control illumination settings during STORM movie acquisition. Hal.py also allows users to engage linear or exponential power ramping using the 'Progressions' control panel.
17. Begin automated image acquisition of all conventional, bead, and STORM les: a. In main Hal.py interface, set working directory to local storage drive.
Critical: Ensure local storage drive has su cient capacity for image acquisition and/or set up automated transfer routine to clear drive of images following acquisition. Over lling the drive will cause the acquisition control software to crash.
b. Adjust desired imaging parameters (e.g. number of frames, frame rate, focus lock position, illumination control, etc.) for multichannel acquisition in a 'master.xml' le.
c. Launch 'Master_xml_generator.py'. d. Select the desired 'master.xml' le from Step B above (.xml format) containing imaging parameters. e. Change le name and save out nal 'master_run.xml' le.
g. Click 'File/load' and select the 'master_run.xml' le for automated imaging.
h. Click 'Validate' to con rm the proper function of the automated image acquisition.
i. Select 'Run' to begin automated imaging.
Alternative: Use commercial instrument with automated image acquisition software to set imaging parameters for selected ROIs. Alternative open-source software may also be used.
Note: Adjust the total laser output power as needed to ensure optimal image quality during image acquisition. For conventional image acquisition this may require reducing laser output while for STORM movie acquisition power output should be increased to achieve su cient density at the sample plane (1-3 kW/cm2).
Process images to generate 3D volumes (Timing: varies (largely automated) The following steps convert raw conventional, bead, and STORM imaging data to nal images. Steps include single-molecule tting, drift/aberration corrections, and elastic registration of serial sections. Links to repositories hosting previously published and custom code (this paper) can be found in the 'Software and algorithms' section.
2. Inspect raw .dax STORM movies in renderer software to determine frame range for single molecule tting.
Note: We recommend testing tting parameters on ~10 frames to view results of parameter settings before tting the nal number of desired frames. 4. Run SMLM tting analysis (fully automated): a. Open '1_batch_ tting.py' script.
b. Change working directory path to the location of the data to be analyzed (experiment folder  Poor photoswitching is often caused by the use of old STORM buffer solution. Remake fresh STORM buffer solution immediately prior to each imaging experiment.

Potential Solution 2:
The chemical etching process exposes the uorophores to the thiol-containing imaging buffer for STORM imaging. Without proper etching, the dyes will not be exposed to the imaging buffer and photoswitching could be slow or absent. Remake the sodium ethoxide solution and ensure proper etching time (5 minutes). It is crucial that the sodium ethoxide solution is not contaminated with water (see Problem 3).

Potential Solution 3:
SMLM methods require high-power excitation to drive optimal single-molecule photoswitching. Ensure the imaging system achieves su cient power density (1-3 kw/cm2) at the sample plane for each imaging wavelength.

Anticipated Results
By following this protocol, users can expect to collect volumetrically-aligned STORM and conventional (diffraction limited) image stacks of arbitrary volumes determined by the number and thickness of sections imaged. In Figure 8, we show maximum projection images of a YFP-expressing retinal ganglion cell with presynaptic protein and postsynaptic gephyrin labeling, demonstrating the volume and resolution capabilities of the approach. Entire neurons and surrounding synaptic elds can be imaged with nanoscale resolution and molecular speci city based on IHC labeling of speci c protein targets. In Figure 9, a separate example, we show a maximum projection transverse section image across the retina highlighting the subsynaptic molecular organization of photoreceptor terminals. The imaged volume captures nanoscale molecular details for thousands of individual synapses in the outer and inner plexiform layers.
In STORM imaging experiments, differences in dye properties impact the nal image resolution (Dempsey et al., 2011). Users should consider wavelength-dependent differences that in uence single-molecule localization precision and biological measurements. In Figure 10, we illustrate this point in the context of synapse imaging in mouse brain tissue. We immunolabeled synapses in the dorsal lateral geniculate nucleus (dLGN) of the mouse thalamus with primary antibodies against presynaptic (bassoon and vesicular glutamate transporter 2 [VGlut2]) and postsynaptic targets (Homer1). We tested two alternative secondary staining conditions that ipped the dyes, Dy749P1 or Alexa647, used to label the synaptic proteins Homer1 and Bassoon ( Figure   10 A). We then collected STORM images and measured the synaptic properties (density, volume, signal intensity) for thousands of synapses across ~ 40,000 cubic microns of dLGN tissue for both conditions (Figure 10 B-D). Results indicate the measured synaptic densities are independent of dye selection, while synaptic cluster volume and total within-cluster signal intensity is greater when imaging with Dy749P1 compared with Alexa647 ( Figure 10 C and D). This is consistent with an increase in the size of the point spread function and decrease in photon emission of red-shifted probes compared to Alexa647 dyes (Dempsey et al., 2011).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.