Optimized Stochastic Optical Reconstruction Microscopy for Imaging Chromatin Structure in Pathological Tissue

Abstract

Direct visualization of higher-order chromatin structure at the molecular scale is of great importance for understanding the impact of chromatin organization on gene expression in many biological processes. Understanding the changes in chromatin structure during pathological processes requires the use of in vivo models and clinical samples, and formalin-fixed, paraffin-embedded (FFPE) tissue is the most widespread form of preservation. Here we describe the details of PathSTORM, an optimized stochastic optical reconstruction microscopy (STORM) protocol for high-quality super-resolution imaging of densely packed higher-order chromatin organization in pathological tissue. We discuss detailed methods for fluorescence staining of DNA and histone proteins, as well as the key technical factors for obtaining high-quality STORM images in pathological tissue samples.

INTRODUCTION

Stochastic optical reconstruction microscopy (STORM) is a powerful super-resolution microscopy technique to visualize molecular-scale structures at nanoscale resolution. We previously published a protocol that provided a detailed description of sample preparation and technical details for super-resolution imaging of higher-order chromatin structure on cultured cells using STORM (see Current Protocols article: Xu, Ma, & Liu, 2017). Precise localization in STORM imaging requires low and uniform background and sparse emitters, and thus it is mostly used on thin transparent samples (e.g., bacteria, thin cultured cells) that meet such requirements. However, pathological tissue samples are often prepared as formalin-fixed paraffin-embedded (FFPE) tissue, which offers the best-preserved tissue morphology, but presents stronger scattering and autofluorescence background. This can introduce localization inaccuracies of up to tens of nanometers that result in a significant degradation in image resolution, along with image artifacts (Deschout et al., 2014). Here we present a detailed protocol for PathSTORM, a variant of STORM that combines optimized sample preparation and high-fidelity image reconstruction for high-quality super-resolution imaging of densely packed higher-order chromatin organization in pathological tissue. We have recently demonstrated the ability of PathSTORM to detect disrupted chromatin folding at the nanoscale level, which may potentially be used to detect early-stage carcinogenesis in clinical tissue samples (Xu et al., 2020).

Basic Protocol 1 describes fluorescence staining of chromatin structure by either immunostaining or direct staining of DNA, including tissue treatment, autofluorescence quenching, and fluorescence staining.

Basic Protocol 2 describes the steps for STORM image acquisition, including optical clearing, data acquisition, background correction, and image reconstruction.

BASIC PROTOCOL 1

IMMUNOFLUORESCENCE STAINING

In this protocol, we describe the details of tissue sample preparation optimized for STORM imaging. The FFPE tissue is cut into 3-μm sections and mounted onto imaging coverslips. After deparaffinization and rehydration, an antigen retrieval procedure is applied to open the epitopes, which are masked by the cross-linking during formalin fixation. The tissue sample is then fluorescence labeled using the standard immunostaining protocol.

Materials

• Paraffin-embedded tissue blocks

• Poly-d-lysine (PDL, Sigma-Aldrich)

• Xylene (ThermoFisher, X3P-1GAL)

• Ethanol (Decon, 200 Proof)

• Antigen retrieval buffer (see recipe or purchase from ThermoFisher, 00-4956-58)

• Dulbecco’s phosphate-buffered saline (PBS; Lonza, BW17–512F)

• Sodium borohydride (NaBH₄; Sigma-Aldrich)

• Blocking buffer: 3% (w/v) BSA (Sigma, A9647) plus 0.1% (v/v) Triton X-100 (Sigma-Aldrich) in PBS

• Primary antibody:

  • Rabbit anti-H3K9me3 antibody (Abcam, ab8898; dilution; 1:600)
  • Rabbit anti H3K27me3 antibody (EMD Millipore, 07–449, dilution; 1:600)
  • Mouse anti RNAP II antibody (Abcam, ab5408; dilution; 1:600)
• Secondary antibodies:

  • Donkey anti-rabbit antibody (Jackson ImmunoResearch) conjugated with Alexa Fluor 647 (ThermoFisher A37573); dilution; 1:200
  • Donkey anti-mouse antibody (Jackson ImmunoResearch) conjugated with Alexa Fluor 647 (ThermoFisher A37573); dilution; 1:200

• 4% (w/v) paraformaldehyde (PFA) in PBS (ThermoFisher, R37814)

• DAPI (ThermoFisher, D1306)

• TOTO^™-3 iodide (ThermoFisher, T3604)

• DNase-free RNase (Sigma-Aldrich, 11119915001)

• Tissue Float Bath (Lab-Line)

• Microtome (Leica)

• #1.5 imaging coverslips (ThermoFisher)

• Tape (Scotch invisible tape, or any other commonly used tape)

• 60°C oven

• Coverslip staining jars (Electron Microscopy Sciences, 72242–21)

• Retriever machine (Electron Microscopy Sciences, 62700–10)

• Glass slides (ThermoFisher)

• Hydrophobic barrier pen (Vector Laboratories)

• Humidified chamber: e.g., plastic box containing a moist paper towel

• Shaking platform (Bellco Biotechnology)

Tissue treatment

Sectioning

• 1Put the paraffin-embedded tissue blocks on ice for ~10 min before sectioning.

  Cold wax allows thinner sections to be obtained.

• 2Fill the Tissue Float Bath (Lab-Line) with ddH₂O and heat it to 40°–45°C.
• 3Place the tissue block on to the microtome and cut sections at a thickness of about 3 μm (discard the first few sections, as they are likely to contain holes caused by trimming).
• 4Pick up the ribbons of tissue sections that are in good shape (unbroken and flat) using tweezers and float them on the surface of the warm water.

  Tissue sections will flatten out in the water; separate the sections using tweezers if multiple sections were cut.

• 5Pick up the tissue sections out of the water bath using poly-d-lysine coated coverslips and store upright in the slide rack.

  Coverslips are coated by incubating with 0.1 mg/mL poly-d-lysine for 20 min.

  Coverslips can be taped using Scotch tape or equivalent on to a glass slide for easy handling of tissue sections, as shown in Fig. 1. Tissue sections mounted on coverslips can be stored long-term at 4°C.

  The choice of 3-μm thickness is based on the ease of routine sectioning at this thickness in a standard clinical laboratory and consistency for minimized background in PathSTORM imaging. In theory, and in our experience, the thinner sections result in a lower background for STORM imaging. The 3-μm FFPE section thickness can be regularly achieved in a clinical laboratory, and sections thinner than 3 μm can be difficult to achieve consistently. Some clinical laboratories prefer 4- or 5-μm sections. Although we have obtained reasonable performance with 4- or 5-μm-thick sections, we did observe higher background in some regions. Therefore, considering the ease of tissue sectioning and reproducibility for high-quality PathSTORM images, we recommend 3-μm section thickness for PathSTORM imaging.

Figure 1.

Example of an imaging coverslip taped on a glass slide to assist in easy handling of the tissue section.

Deparaffinization and rehydration

• 6Peel the tape on the coverslip carefully, remove the coverslip from the slide, and heat the tissue section in a 60°C oven for 30 min.
• 7Put the coverslip into a staining jar (see Fig. 2) and deparaffinize sections with three changes of xylene, each time for 10 min.
• 8Re-hydrate with two changes of 100% ethanol, each time for 10 min.
• 9Place in 95% ethanol for 5 min.
• 10Place in 70% ethanol for 5 min.
• 11Place in 50% ethanol for 5 min.
• 12Wash using a flow of ddH₂O for 5 min.

Figure 2.

Staining jar for coverslip.

Antigen retrieval

• 13Prepare antigen retrieval buffer, or use a commercially available buffer.
• 14Fill the coverslip chamber of the Retriever machine with antigen retrieval buffer and put the coverslip in the chamber. Mark or remember which side of the coverslip contains the tissue.
• 15Fill the Retriever with 750 ml ddH₂O, put the chamber into the Retriever, and press the start button.
• 16The tissues will be processed automatically. After 2 hr, when the Retriever has cooled down, open the lid and proceed with immunostaining.

  To save time, one can also carefully open the lid after 30 min, take out the chamber, and let it cool down at room temperature. Be careful of the hot surface.

  The Retriever helps to maintain consistency in the antigen-retrieval process. If you do not have the antigen retriever machine, antigen retrieval can be conducted by using a microwave oven, high-pressure cooker, or hot plate. Generally, heat tissues in the antigen retrieval buffer and maintain at a sub-boiling temperature (90°C–100°C) for 10 min. Antigen retrieval conditions should be determined for different samples. For details, see “Antigen retrieval” in Critical Parameters and Troubleshooting.

  If a microwave oven or hot plate is used for antigen retrieval, a second container is recommended in case the buffer overflows when heating. Carefully determine the proper power level to maintain a sub-boiling temperature for 8–10 min and avoid vigorous boiling. When using these methods, the antigen retrieval buffer can boil over, and a large amount of the buffer will overflow or evaporate. Carefully watch the buffer level in the container and add more buffer if necessary. It is important not to allow the slides to dry out during this process.

Immunofluorescent staining

• 17After the samples have cooled down, transfer the coverslips from the chamber back to the staining jar. Wash with PBS three times, each time for 5 min.
• 18Treat with freshly prepared 0.1% (w/v) NaBH₄ in PBS three times, each time for 10 min. Prepare and keep the NaBH₄ solution on ice to reduce the rate of decomposition during the experiment.
• 19After washing with PBS, block the sample with blocking buffer for 1 hr.
• 20Pick up the coverslip and put it on a glass slide, making sure that the tissue is on the upper side.
• 21Draw a circle on the coverslip around the tissue with a hydrophobic barrier pen.
• 22Dilute the primary antibody with blocking buffer at the desired concentration and apply 150 μl to the section. Place the sample in a humidified chamber and incubate it with the primary antibody overnight at 4°C.
• 23The next day, wash the samples with PBS three times, each time for 5 min.
• 24Dilute the fluorophore-conjugated secondary antibody with blocking buffer at the desired concentration and apply 150 μl to the section. Incubate at room temperature for 1–2 hr.

  To minimize fluorophore exposure to light, it is best to keep samples covered.

• 25Wash three times with PBS, each time for 5 min.
• 26Post-fix the stained tissue with 4% PFA for 10 min, then wash it once with PBS.
• 27Proceed to imaging.

Fluorescence staining of DNA

Alternatively, if users want to directly visualize DNA structure, the most commonly used nucleic acid binding dye is DAPI. But the excitation of DAPI is in the ultraviolet or blue regions of visible light, which are not well-suited for STORM imaging. We have used an alternative, TOTO^™-3 iodide, which can be excited at the wavelength of ~640 nm in this protocol. This protocol is used independently of the above immunofluorescence staining protocol.

The staining concentration of TOTO^™-3 for STORM is generally lower than that for conventional imaging, and needs to be optimized to obtain the best photo-switching performance. In our experience, the photostability of TOTO^™-3 is higher than that of the conventional photo-switchable dyes (e.g., Alexa Fluor 647), and needs higher laser power (3–4 kW·cm⁻²) to convert it to the dark state. If necessary, an appropriate bleaching period (2–5 min) can be used to bleach the fluorophore to the dark state.

• 28Treat the tissue section as described above (deparaffinization, rehydration, and antigen retrieval) up to the blocking step (steps 1 to 19).
• 29Treat sections with 50 μg/ml DNase-free RNase at 37°C for 30 min.
• 30Incubate with 100 nM TOTO-3 for 30 min at room temperature.
• 31Wash three times with PBS, each time for 5 min.
• 32Post-fix the stained tissue with 4% PFA for 10 min and wash once with PBS.
• 33Store the tissue in PBS.

BASIC PROTOCOL 2

STORM IMAGE ACQUISITION

In this protocol, we will describe STORM image acquisition for the fluorophore-labeled tissue sample prepared in the previous section. Before image acquisition, we perform optical clearing, add the optimized STORM imaging buffer to the sample chamber, and mount the sample on the stage of the microscope. Higher laser power (~2.5kW·cm⁻²) is used to trigger the photo-switching of the fluorophore. Generally, 10,000 to 30,000 frames are recorded as raw data, which are subsequently used for STORM image reconstruction.

Materials

• Tissue samples from Basic Protocol 1

• 2,2′-thiodiethanol (TDE; Sigma, 166782)

• 1 M Tris·Cl, pH 8.0

• NaCl

• Glucose

• Glucose oxidase from Aspergillus niger, type VII, lyophilized powder, ≥100,000 U/g solid (Sigma-Aldrich)

• Catalase from bovine liver, lyophilized powder, ≥10,000 U/mg protein (Sigma-Aldrich)

• Cyclooctatetraene (COT; Sigma-Aldrich)

• Dimethylsulfoxide (DMSO)

• 2-mercaptoethanol (2-ME)

• Immersion oil for microscopy FluoSpheres^™ Carboxylate-Modified Microspheres, 0.1 μm, yellow-green fluorescent (505/515), 2% solids (ThermoFisher, F8803)

• Customized imaging chamber (CAD design files can be found at https://github.com/YangLiuLab/Design_ImagingChamber) or commercially available imaging chamber (Attofluor^™ Cell Chamber; ThermoFisher, A7816)

• Commercial STORM microscope (N-STORM) or custom-built STORM microscope

Optical clearing

FFPE tissue sections (even thin sections of 3 μm) are generally non-transparent (as shown in Fig. 3A), which introduces higher scattering. Optical clearing is an effective way to reduce scattering. We use a simple optical clearing approach.

Figure 3.

Mouse intestinal tissue section (3 μm) before and after optical clearing.

• 1Immerse the stained tissue section from the last step of Basic Protocol 1 in 60% (v/v)2,2′-thiodiethanol (TDE) for 10 min, as shown in Figure 3B.

  The STORM system requires a high-numerical-aperture (NA) objective to maximize the efficiency of the collected photons. An oil-immersion objective is often used to maximize the NA at >1.4. However, the imaging buffer used in STORM is aqueous, which introduces a refractive index mismatch between water (refractive index of ~1.334) and immersion oil (refractive index of 1.51), resulting in spherical aberration and elongated focus. The elongated focus together with the reduced laser power density lead to high and non-uniform background. Therefore, the addition of 60% TDE to the STORM imaging buffer (see step 2b) increases the refractive index of the imaging buffer to approximately 1.47–1.48, which significantly reduces the spherical aberration and thus the background under the high laser power density for STORM imaging. For details, see “Imaging buffer optimization” in Critical Parameters and Troubleshooting.

Preparation of imaging buffers for PathSTORM

• 2
  Prepare PathSTORM imaging buffers:

  1. Buffer A: 0.5 ml of 1 M Tris·Cl (pH 8.0), 0.146 g NaCl, 50 ml H₂O; vortex to mix.
  2. Buffer B (in 60% TDE): 5 g glucose, 0.029 g NaCl, 2.5 ml 1 M Tris·Cl (pH 8.0), 30 ml TDE, 17.5 ml H₂O.
  3. Glucose oxidase with catalase (GLOX): Combine 28 mg glucose oxidase, 2.5 μl of catalase, and 500 μl buffer A. Vortex to dissolve, then centrifuge 5 min at 13,400 × g, room temperature, and keep the supernatant.
  4. It is recommended to store the GLOX solution no longer than 1 week at 4°C. For longer storage, aliquot and store at −20°C.
  5. 200 mM COT stock solution: 22.5 μl COT dissolved in 1 ml DMSO.
  6. Imaging buffer (1 ml): Add 10 μl GLOX (from step 2c), 10 μl of 2-ME, and 10 μl 200 mM COT stock solution to 1 ml buffer B, and gently mix.

Data acquisition

• 3Carefully pick up the coverslip with the section (see step 1) and place onto the customized imaging chamber (as shown in Fig. 4). Seal properly.
• 4Add 1 ml imaging buffer (from step 2e) to the imaging chamber, add immersion oil to the objective, and place the imaging chamber on to the sample stage of the STORM microscope.
• 5Use a lower laser power (~1 mW) to identify the imaging objects and focal plane, and acquire a conventional wide-field fluorescence image.
• 6Increase the laser power to maximum (2 to 10 kW·cm⁻² power density) to turn “off” fluorescent molecules and trigger photo-switching.
• 7Set the exposure time (e.g., 20 to 50 ms) and the total acquisition frame numbers (10,000 to 40,000 frames).
• 8If the emitters are significantly reduced after a few thousand frames, the activation laser (405-nm) power (~1 μW) can be used at an optimized frame, and the power of the 405-nm laser is gradually increased at a rate of 0.2% per 1000 frames. Start the acquisition.
• 9If two-color dSTORM imaging is needed, a second-color dSTORM imaging can be conducted sequentially, where the first 10,000 to 30,000 frames are acquired on one fluorophore (e.g., Alexa Fluor 647), followed by 10,000 to 30,000 frames on a second fluorophore (e.g., CF568).

Figure 4.

Example of a customized imaging chamber with a magnetic base (CAD design files in https://github.com/YangLiuLab/Design_ImagingChamber).

SUPPORT PROTOCOL 1

DRIFT CORRECTION

Because of the long data acquisition time (10–20 min) for STORM imaging, real-time drift correction is needed to maintain the axial focal position within 50 nm or less. The lateral drift correction can be corrected either online or through post-processing (Wang et al., 2014).

In this protocol, we use fluorescent beads as fiduciary markers for 3D drift correction based on our previously published method (Ma, Xu, Jin, Huang, & Liu, 2017).

Additional Materials (also see Basic Protocol 2)

• FluoSpheres^™ Carboxylate-Modified Microspheres, 0.1 μm, yellow-green fluorescent (505/515), 2% solids (ThermoFisher, F8803)

• Dulbecco’s phosphate-buffered saline (PBS; Lonza, BW17–512F)

Perform Basic Protocol 2 with the following variations

1. Deposit 1 ml of 100-nm fluorescent bead suspension diluted 1:1,000,000 with PBS onto the section on the coverslip before STORM imaging (at Basic Protocol 2, step 4).

2. Incubate with the tissue sample for 20 min at room temperature.

   There should ideally be at least three beads within the70 × 70−μm² field of view.

   Some commercial STORM microscopy systems (e.g., N-STORM) use the reflection of infrared light at the interface between the coverslip and the mounting medium to “lock” the focus. However, the TDE-based imaging buffer in this protocol was designed to match the refractive index between the immersion oil and the mounting medium. Therefore, the reflection-based drift correction does not work well for such sample.

SUPPORT PROTOCOL 2

IMAGE RECONSTRUCTION

Background correction

The optical clearing process significantly reduces the background in the FFPE tissue section being imaged, but some residual non-uniform background remains. To achieve consistent and reproducible image quality, we have further developed an algorithm-based drift-correction method utilizing extreme-value-based emitter recovery (EVER) to remove residual heterogeneous background (Ma, Xu, & Liu, 2019).

1. Given a raw image stack, transform the intensity value from digital counts to physical photons.

2. Segment it into a series of 100-frame sub-stacks.

3. Calculate the minimum value of each pixel along the temporal axis for each sub-stack.

4. Calculate the expected background value according to our derived transform function between the expected value and the minimum value of the Poisson distribution (Ma et al., 2019). The ImageJ plugin and the source code can be found on our website at https://pitt.box.com/v/EVER-BackgroundCorrection.

Given the high laser power density used in STORM imaging, the background is generally much higher compared to that of conventional wide-field fluorescence images. As shown in Figure 5A and B, the background structural features under the STORM imaging conditions (Fig. 5B) are not present in the conventional wide-field fluorescent image.

Figure 5.

Demonstration of background correction. (A) Conventional wide-field fluorescence image. (B) Single raw image obtained under PathSTORM imaging conditions. (C) The estimated background using EVER. (D) The recovered fluorescent emitters after background correction.

Super-resolution image reconstruction

The above background correction is used as a pre-processing step to remove the background. After the background is removed, the super-resolution image can be reconstructed via commonly used STORM image-reconstruction software as previously described (Xu et al., 2017). Many reconstruction software packages are freely available. For example, the ImageJ plugin ThunderSTORM is one of the most widely used software packages for STORM image reconstruction, with a friendly user interface (Ovesný, Křížek, Borkovec, Švindrych, & Hagen, 2014). For users looking for more options, there is a detailed review of the quantitative evaluation of single-molecule localization microscopy software packages (Sage et al., 2015). Of note, in some cases, if the labeled structure is highly condensed, overlapping emitters may not be avoided without significantly sacrificing the throughput. The ability to account for overlapping emitters helps to improve the localization accuracy and the emitters’ recall rate. For example, high-density localization methods such as the multi-emitter fitting options in ThunderSTORM, Falcon, and 3D-DAOSTORM all account for overlapping emitters. However, most dense emitter localization methods are significantly slower and more computationally intensive than their single emitter localization counterparts. Our recently developed STORM image reconstruction algorithm termed WindSTORM (Ma et al., 2019) accounts for both background correction and overlapping emitters and has achieved high-speed high-density image reconstruction. However, the sampling rate of the STORM imaging system needs to be adjusted to slightly above the Nyquist rate (1.2 to 2 times the Nyquist rate, or effective pixel size ≤100 nm). WindSTORM codes are freely available on our website at https://pitt.app.box.com/v/WindSTORM (password: Biophotonics).

HEMATOXYLIN AND EOSIN (H&E) STAINING

H&E staining is the most widely used staining technique in histopathology, and is considered the gold standard. H&E staining makes use of a combination of two dyes, namely hematoxylin and eosin. This combination differentially stains various tissue elements and makes them easy to observe. The principle behind H&E staining is the chemical attraction between tissue and dye. Hematoxylin, a basic dye, imparts blue-purple contrast to basophilic structures, primarily those containing nucleic acid moieties such as chromatin, ribosomes, and cytoplasmic regions rich in RNA. Acidic eosin counter-stains the basic elements such as red blood cells, cytoplasm, muscle, and collagen in varying intensities of pink, orange, and red.

H&E staining remains the gold standard for defining the pathological phenotype on the tissue section. It is used in this protocol for two purposes. First, the H&E-stained tissue from a serial section of FFPE tissue used for STORM imaging is useful to identify a specific type of lesion (e.g., low-grade dysplasia) and corresponding regions before STORM imaging. It also helps in identifying the types of mixed lesions, if present. Second, H&E staining can be performed on the same tissue after STORM imaging to ensure one-to-one correspondence between the H&E-stained image and STORM image. Although consecutive tissue sections display similar tissue architecture, they are often not identical. For precise correlation, H&E staining is required on the same tissue section used for PathSTORM imaging.

Materials

• Tissue section

• Harris hematoxylin (Anatech Ltd., Ref 842)

• Eosin-Y solution (Anatech Ltd., Ref. 834)

• Glacial acetic acid (EMD Millipore, AX0073)

• Scott’s Tap Water Substitute (Cancer Diagnostic Inc, CM4951W)

• Ethanol (Decon, 200 Proof)

• Xylene (ThermoFisher, X3P-1GAL)

• Cytoseal Mounting Medium (Thermo Scientific, REF 8312–4)

• Staining jars

• Microscope

NOTE: If H&E staining is to be performed on the STORM imaging sample (Basic Protocol 2), first wash out the TDE-containing imaging buffer with PBS three times, then perform the staining procedure described in the protocol below starting at step 4. The steps may require optimization or longer staining time.

1. Deparaffinize and rehydrate the tissue section as described in Basic Protocol 1. Wash briefly with H₂O.

2. Stain tissue with Harris hematoxylin solution for 7 min.

3. Wash tissue in running tap water for 5 min.

4. Treat tissue with 10% acetic acid for 1 min, then wash with tap water for 1 min.

5. Treat with Scott’s Tap Water Substitute for 1 min, then wash with tap water for 1 min.

6. Stain tissue with eosin-Y for 3 min.

7. Wash twice with 70% ethanol.

8. Wash twice with 100% ethanol.

9. Wash twice with xylene.

10. Mount the stained slide with Cytoseal Mounting Medium or Permount as follows. Place a drop of mounting medium on the slide. Taking care not to generate air bubbles, tilt the coverslip and let it fall gently on the glass slide. Allow the mounting medium to spread beneath the coverslip, covering the entire tissue sample. Carefully wipe up any extra mounting medium with Kimwipes.

11. Examine under bright-field microscope: nuclei are stained blue and cytoplasm is pink to red.

REAGENTS AND SOLUTIONS

Antigen retrieval buffer

• Antigen retrieval buffer can be prepared in the lab if not using a commercially available version. Two commonly used recipes are listed below.

• Sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0):

• 2.94 g trisodium citrate (dihydrate)

• 1 L distilled water

• Mix to dissolve

• Adjust pH to 6.0 with 1 N HCl

• Add 0.5 ml Tween 20 and mix well

• Store at room temperature for 3 months or at 4°C for longer storage

• Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA, 0.05% Tween 20, pH 9.0)

• 1.21 g Tris base

• 0.37 g disodium EDTA

• 1 L distilled water

• Mix to dissolve

• Adjust pH to 9.0 using 1 M

• Add 0.5 ml of Tween 20 and mix well

• Store at room temperature for 3 months or at 4°C for longer storage

• The Tris-EDTA buffer is the author’s buffer of choice for use in these protocols.

COMMENTARY

Background Information

Abnormal chromatin structure is one of the most universal and striking characteristics of cancer cells and has remained the gold standard for cancer diagnosis for two centuries (Brock, Herman, & Baylin, 2007; Zink, Fischer, & Nickerson, 2004). Due to the diffraction limit of conventional light microscopy, the observable abnormal chromatin structure of cancer cells is still limited to micrometer-scale features such as enlarged nuclear size, irregular shape, and coarse chromatin texture (Zink et al., 2004). Recently developed super-resolution techniques have revolutionized biological imaging by breaking the diffraction barrier that limits the resolution of conventional microscopy to ~200 nm. In particular, STORM-based super-resolution imaging has shown great promise for understanding in situ nanoscale chromatin structure down to the level of nucleosome clusters (or nanodomains) on a scale of 20–30 nm in cultured cells and simple organisms (e.g., Drosophila) (Beliveau et al., 2015; Prakash et al., 2015; Ricci, Manzo, García-Parajo, Lakadamyali, & Cosma, 2015; Xu et al., 2018). However, characteristic higher-order chromatin structure has been difficult to visualize in well-preserved pathological tissue samples from patients due to a number of challenges. The super-resolving imaging capability of STORM is largely based on precise localization of sparsely distributed single fluorescent emitters with nanometer precision, and it achieves the best performance in thin transparent samples (e.g., bacteria, thin cultured cells) with uniform background and sparse emitters. However, stronger scattering and autofluorescence in tissue often generate high and non-uniform background that can significantly degrade image resolution and introduce artifacts (Deschout et al., 2014).

In this article, we introduce a PathSTORM, a STORM-based super-resolution imaging method optimized for robust and high-quality super-resolution imaging of higher-order chromatin structures in pathological tissue. It allows direct visualization of higher-order chromatin structures down to a scale of 20–30 nm in the spatial context of tissue architecture.

Critical Parameters and Troubleshooting

Antigen retrieval

Formalin fixation immobilizes antigens and best preserves the morphology of pathological tissue through protein cross-linking. But this process masks antigens and restricts antigen-antibody binding. Antigen retrieval is widely used on FFPE tissue to unmask the antigens so that the antibody can access the target protein. The type of antigen retrieval depends on many factors such as tissue type, target protein, and antibody used. Heat-induced antigen retrieval is the most widely used method. Different antigen retrieval buffers are available: sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0) and Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA solution, 0.05% Tween 20, pH 9.0) are two commonly used buffers for heat-induced antigen retrieval. Laboratories often optimize their antigen-retrieval protocol to identify the best conditions for any particular antigen. For our experiment here, we choose the Tris-EDTA buffer.

Autofluorescence reduction

Autofluorescence is a general term describing background fluorescence unrelated to the target signal in the tissue section during fluorescence imaging. Autofluorescence frequently hampers clean visualization of the targeted protein and decreases the SNR (signal to noise ratio) and resolution. Autofluorescence in tissue has been attributed to many factors, including endogenous tissue elements such as lipofuscin, elastin, red blood cells (RBCs), and collagen (Banerjee, Miedema, & Chandrasekhar, 1999; Monici, 2005). In addition, formalin fixation introduces a significantly higher amount of autofluorescence (Baschong, Suetterlin, & Laeng, 2001) compared to frozen tissue. The autofluorescence signal in FFPE is a more serious problem in the green and yellow spectral regions. Although red and far-red fluorophores tend to present lower background autofluorescence compared to green and yellow fluorophores, autofluorescence still occurs in the 600–700 nm range in some cases. Under the high laser power density during STORM imaging, the autofluorescence intensity can also be enhanced compared to conventional wide-field fluorescence microscopy. Therefore, it is more important to reduce autofluorescence in STORM imaging.

Certain chemical reagents can reduce autofluorescence, such as sodium borohydride (NaBH₄, 0.1%; Clancy & Cauller, 1998), CuSO₄ in ammonium acetate buffer (Schnell, Staines, & Wessendorf, 1999), NH₄Cl (50 mM in PBS), or Sudan Black B in 70% ethanol (Schnell et al., 1999). In this protocol, we choose NaBH₄ treatment to reduce autofluorescence without apparent effect on the photo-switching properties of fluorophores such as Alexa Fluor 647 and CF568.

Imaging buffer optimization

Besides autofluorescence, FFPE tissue also presents stronger scattering, which results in a high and non-uniform background that can significantly degrade image resolution and introduce artifacts. As discussed above in Basic Protocol 1, the use of aqueous imaging buffer induces spherical aberration due to the refractive index mismatch, which reduces imaging resolution. Under the high laser power density of STORM imaging, it also leads to a significant background that makes accurate determination of single-molecule localization virtually impossible. The use of an index-matched mounting medium not only reduces spherical aberration, but, importantly, dramatically reduces the background. Most index-matching mounting media are designed for conventional fluorescence microscopy and are not compatible with the recipes used in the STORM imaging buffer to enable photo-switchable blinking of the fluorophores. In this protocol, we use a water-soluble medium, TDE (2,2′-thiodiethanol), which has a high index of 1.521. By adjustment of its proportion of water, STORM imaging buffer achieves a closer refractive index (~1.47–1.48) to immersion oil (1.51) without apparent effect on the photo-switching performance of the fluorophores. It should be noted that although the higher concentration of TDE (>80% TDE) affords a better-matched refractive index with immersion oil in theory, based on our observation, the number of single fluorescent emitters is significantly reduced.

Photon yield enhancement

In STORM or any other single molecule localization–based super-resolution mi croscopy, localization precision—defined as the standard deviation σ_x for the estimated and true spatial coordinates of each fluorescent emitter—determines the resolution of the reconstructed super-resolution image via full width at half maximum (FWHM) for the distribution of the measured emitter position coordinates, or $FWH{M}_x=2\sqrt{2ln2{\sigma }_x}\approx 2.355{\sigma }_x$. For fluorescent emitters approximated as 2D Gaussian distribution with a shot-noise-limited detection, the limit for localization precision is described as:${\sigma }_x\ge {\sigma }_{psf}/\sqrt{N}$, where σ_psf is the standard deviation of the point spread function (PSF) and N is the number of collected photons. Therefore, one efficient way to improve the localization precision (thus, image resolution) is to increase the photon number of fluorescent emitters. The polyunsaturated hydrocarbon cyclooctatetraene (COT) has been reported to significantly increase the photon yields and subsequently improve the localization precision in STORM imaging, with satisfactory blinking performance of the fluorophore (Olivier, Keller, Gonczy, & Manley, 2013). We also added COT in our imaging buffer to improve the resolution of STORM images in the pathological tissue. Figure 6 shows the histogram distribution of photons from the emitters.

Figure 6.

Histogram of photon counts for PathSTORM on intestinal tissue (with a single frame acquisition time of 20 ms.

Optimized background correction and improved localization precision

The use of optical clearing and index matching effectively suppresses background in STORM imaging, but cannot fully remove all the residual background, which can be heterogeneous from different regions of the tissue or can vary among different samples. We have also applied an algorithm-based approach to estimate and remove the heterogeneous background and recover single emitters (Ma et al., 2017, 2019). However, due to the inherent densely packed chromatin structure, overlapping emitters are often present following background removal. To accurately localize the overlapping emitters without compromising the throughput, we adopted a computationally efficient emitter localization method, which is a variant of the deflation-localization method used in high-density localization algorithms (Holden, Uphoff, & Kapanidis, 2011; Sergé, Bertaux, Rigneault, & Maguet, 2008) to restore overlapping emitters by subtracting surrounding emitters without computationally intensive iterative methods. Lastly, we applied our previously developed single-iteration localization algorithm—gradient fitting (Ma et al., 2015) for fast and precise emitter localization. As shown in Figure 7, PathSTORM improved the localization accuracy from ~30 to 17 nm, increased the emitter recall rate from 81% to 91%, and reduced false-positive rate from 10% to 0.2%.

Figure 7.

(A1) Simulated single-frame raw image of emitters at a density of 1.5 emitters/μm² in the presence of a heterogeneous background. The emitters have an average of 3000 photons, the point spread function has a standard deviation of 1 pixel, and background fluctuates with a sinusoidal pattern at 500 ± 300 (sine cycle per image = 2). (A2-A3) Comparison of ground-truth emitters and emitters recovered using extreme-value based emitter recovery, which shows ~98% agreement (~98% image similarity). (B1) Illustration of overlapping emitter decomposition based on surrounding emitter subtraction. (B2) The localization accuracy of ThunderSTORM (single-emitter fitting algorithm) and PathSTORM, quantified by root mean square error (RMSE) between the positions of ground-truth emitters and the recovered emitters, as a function of the separation distance between emitters. At a distance smaller than 500 nm, the localization error using ThunderSTORM significantly increases, while the overlapping emitter decomposition used in PathSTORM remains low even in the case of overlapping emitters. (C1-C3) Comparison of (C1) localization accuracy (defined by RMSE), (C2) emitter recall rate, and (C3) false-positive rate for STORM images reconstructed by a sparse emitter algorithm: ThunderSTORM (blue) and PathSTORM (red).

Understanding Results

Figure 8 shows an example of reconstructed super-resolution images of distinct chromatin structures labeled by two heterochromatin marks, H3K9me3 and H3K27me3, with corresponding histology images and conventional wide-field fluorescence images of single nuclei. Although all of these images mark the condensed regions of chromatin, the super-resolution images exhibit distinct structural features. The H3K9me3-dependent heterochromatin consists of microscopically visible large foci, and each focus is composed of nano-sized clusters, while H3K27me3-dependent heterochromatin also has nano-sized clusters that spatially distributed within the entire nucleus. Figure 9 shows an example of a super-resolution image of TOTO-3 labeled DNA structure in epithelial cells where the heterogeneous DNA nanodomains are visible. Figure 10 shows an example of two-color STORM images of H3K9me3 overlaid with phosphorylated RNA polymerase II on wild-type mouse intestinal tissue. RNAPII shows distinct nanoclusters, as we previously observed in cell lines (Xu et al., 2018). As expected, the active RNAPII showed little co-localization with the condensed regions of transcriptionally repressed H3K9me3.

Figure 8.

Single-color imaging of histone modifications in pathological tissue. (A1-B1) H&E-stained pathology images and (A4-B4) STORM images of heterochromatin stained with H3K9me3 and H3K27me3 of normal tissue from wild-type mice. (A2-B2) Conventional wide-field fluorescence images and (A5-B5) corresponding STORM images of heterochromatin from a single nucleus. (A3-B3) and (A6-B6) are progressively zoomed regions of (A2-B2) and (A5-B5).

Figure 9.

STORM images of DNA stained with TOTO-3 of normal tissue from wild-type intestinal epithelial tissue.

Figure 10.

Two-color dSTORM imaging of H3K9me3 and active RNAP II in wild-type intestinal mouse tissue.

Time Considerations

The consumption of time for each step is indicated in the flowchart in Figure 11.

Figure 11.

The workflow of optimized fluorescence staining and imaging for pathological tissue.