Cyclic Multiplex Fluorescent Immunohistochemistry Protocol to Phenotype Glial Cells in Human Formalin-Fixed Paraffin-Embedded Brain Sections

Abstract/Summary:

There is a growing interest in expanding the multiplexing capability of immunohistochemistry to achieve a deeper phenotyping of various cell types in health and disease. Here we describe a protocol of cyclic multiplex fluorescent immunohistochemistry that enables the labeling of up to 16 antigens on the same formalin-fixed paraffin-embedded section using “off-the-shelf,” commercially available, primary antibodies as well as fluorescently-conjugated secondary antibodies. Key steps include the denaturing/stripping of the antibodies by microwaving and the quenching of any remaining fluorescent signal between the cycles of otherwise traditional multiplexed fluorescent immunohistochemistry. We have successfully applied this protocol to characterize astrocytic and microglial responses to Aβ plaques and neurofibrillary tangles in Alzheimer’s disease brains, but it could be easily adapted to other user’s needs regarding cell types, disease, and organ.

1. Introduction

Fluorescent immunohistochemistry (IHC) has been traditionally used to study the simultaneous localization of a limited number—three or four—of protein markers on a tissue section. However, there is growing interest in expanding the multiplexing capability of fluorescent IHC to capture the heterogeneity of cellular changes associated with normal aging, injury, or disease while preserving their spatial context [1–5]. Single-nuclei and single-cell RNA sequencing are revealing the wide diversity of brain cell subpopulations and states under normal and pathological conditions [6–10], but are limited by the lack of spatial information. Spatial transcriptomic and proteomic methods are rapidly improving and reaching single-cell resolution [11–14], but also require expensive, highly specialized equipment, which is not accessible to many laboratories.

Here we report a novel cyclic multiplex fluorescent IHC protocol that allows the labeling and quantitation of up to 16 different markers in the same formalin-fixed paraffin-embedded (FFPE) human postmortem brain section with “off-the-shelf,” commercially available, primary and fluorescently-conjugated secondary antibodies. We have successfully applied this methodology to identify various phenotypes of astrocytes and microglia in Alzheimer’s disease (AD) and healthy control brains and examine the effect of the distance between these cells and the AD neuropathological hallmarks—amyloid-β (Aβ) plaques and neurofibrillary tangles [15]—but this method could be easily adapted by the user to characterize other cell types, diseases, and tissues.

To develop this protocol, we combined the Opal [16] and the tissue-based cyclic immunofluorescence (t-CyCIF) [17–20] methods. Briefly, the protocol first involves a traditional fluorescent IHC with primary and species-appropriate fluorescently-labeled secondary antibodies. After imaging the whole section, the antibodies are denatured by microwave heating (as described in the Opal method) [16] and any fluorescent signal from remaining secondary antibodies is quenched with an oxidizing alkaline solution (as in the t-CyCIF method) [17–20]. Next, a new round of traditional fluorescent IHC and imaging follows, and this cycle is repeated up to 8 times (Fig. 1). The protocol is versatile so that the user can design a different sequence of primary antibodies and add more markers and cycles. Finally, we have also developed an image analysis pipeline to segment glial cell profiles and quantify the signal intensity of all markers at the single-cell level.

Figure 1. Schematic of the cyclic multiplex fluorescent immunohistochemistry protocol.

(a) Each round consists of a traditional multiplex fluorescent IHC, followed by whole-slide imaging, and an antibody denaturing/quenching step (Note: elements of this figure were obtained from Sevier Medical Art by Sevier [smart.sevier.com], which is licensed under a Creative Commons Attribution 3.0 Unported License). (b) Sequence of antibodies used to phenotype astrocytes and microglia in healthy control and Alzheimer’s disease brains. The host species of the primary antibodies (Ms = mouse, Rb = rabbit, and Gt = goat) and the fluorophores conjugated to the secondary antibodies (red = Cy3, green = AF488, magenta = Cy5) are indicated.

In summary, this protocol is a useful methodology to comprehensively phenotype individual cell profiles while preserving the tissue architecture and the spatial relationships between the cell type(s) of interest and the disease neuropathological features.

2. Materials

Unless otherwise indicated, prepare and store all reagents at room temperature (RT). Follow waste disposal regulations when disposing waste materials.

1. FFPE tissue sections: postmortem human temporal lobe specimens are fixed by immersion in 10% formalin for 2-3 weeks and, after fixation, a small 2×2 cm block is dissected and embedded in paraffin (see Note 1). Eight-μm-thick tissue sections are cut with a microtome and mounted on microscope glass slides. Postmortem human brain tissue samples should always be obtained following informed written consent by the next-of-kin and under an approved Institutional Review Board protocol.

2. Primary antibodies: see Table 1 for a list of primary antibodies tested and their working conditions (see Note 2). Fluorescently-labeled primary antibodies, either commercially available or home made using a labeling kit, can also be used in this protocol; in that case, just skip steps #11 and #12 in the Methods section 3.1.

3. Secondary antibodies: use your regular secondary antibodies. This protocol was developed with Cyanine3 (Cy3)-conjugated anti-mouse and anti-rabbit, AlexaFluor488 (AF488)-conjugated anti-mouse and anti-rabbit, and Cyanine5 (Cy5)-conjugated anti-goat secondary antibodies (Jackson ImmunoResearch Labs, West Grove, PA), all of them made in donkey and used at a 1:200 concentration (see Note 3).

4. Xylenes (see Note 4).

5. Ethanol: prepare 100%, 95%, 80%, and 70% ethanol solutions in distilled water (see Note 5).

6. Citrate buffer: citric acid anhydrous 10 mM, 0.05% Tween20, pH 6.0. Dissolve 1.92 g of citric acid anhydrous in 900 mL of distilled water. Adjust pH to 6.0 with 2N sodium hydroxide and add 0.5 mL of Tween20. Add distilled water up to 1 L. Store at RT for 3 months or at 4°C for 1 year.

7. Tris-buffered saline (TBS): tris base 50 mM, sodium chloride 150 mM, pH 7.4. Weigh 6.05 g Tris base and 8.76 g sodium chloride. Dissolve them in 1 L of distilled water and adjust pH to 7.4 with 2N hydrochloric acid.

8. Blocking buffer: 10% normal donkey serum (NDS) in TBS. Prepare fresh before use.

9. Antibody buffer: 5% NDS in TBS. Prepare fresh before use.

10. Sudan Black B solution or alike (e.g., Millipore Autofluorescence Eliminator Reagent): 1% Sudan Black B in ethanol 70%. For 1 L, dissolve 10 g in 1 L of 70% ethanol. Filter through a 0.22 μm syringe filter before applying it onto the sections to remove any precipitated crystals.

11. DAPI solution: DAPI 0.001 mg/mL in TBS. Add 20 μL of DAPI 5 mg/mL to 9.980 mL of TBS to prepare a 0.01 mg/mL stock. From this stock, make a 1:10 dilution to obtain the 0.001 mg/mL working solution. Store at RT protected from light for 6 months.

12. Aqueous mounting medium with DAPI (e.g., DAPI Fluoromount-G^® SouthernBiotech).

13. Coverslip removal solution: Tween20 0.05% in TBS. Dilute 500 μL of Tween20 in 1 L of TBS (see Note 6). Store at RT.

14. NaHCO₃ stock solution: 0.5M NaHCO₃, pH 11.2. Dissolve 42 g of NaHCO₃ in 1 L of distilled water. Mix thoroughly and add 2N sodium hydroxide to adjust the pH around 11.2. Store at 4°C for up to 6 months. Check the pH prior to use to confirm that it remains between 10.9 and 11.3. If the pH is outside this range, discard the buffer and prepare fresh.

15. 30% H₂O₂ stock solution.

16. Fluorescence quenching solution: 0.1M NaHCO₃, 3% H₂O₂, pH 11.2. For a final volume of 100 mL, mix 20 mL of NaHCO₃ stock solution with 70 mL of distilled water. Immediately before use, add 10 mL of 30% H₂O₂ and mix properly. Do not reuse this solution.

17. Histology supplies:

    • 17.1.
      Histological staining dish with slide holder.
    • 17.2.
      Coplin jars.
    • 17.3.
      Slide staining moisture chamber (see Note 7).
    • 17.4.
      Hydrophobic barrier pen (PAP pen or similar).
    • 17.5.
      Coverslips No.1, 50×22 mm.
18. Equipment:

    • 18.1.
      Microwave oven.
    • 18.2.
      Fluorescence microscope: a microscope capable of scanning whole slides at high resolution is recommended to facilitate image alignment and downstream analyses. For example, this protocol was developed with the Olympus VS120 Virtual Slide Microscope, which features a 40x objective, LEDs for DAPI, TRITC (spectrally similar to Cy3), FITC (spectrally similar to AF488), and Cy5, automatic focus and stitching, and a batch loader for up to 100 slides.
19. Other materials and equipment:

    • 19.1.
      Micropipettes and tips for various volumes (P1000, P200, P20, P10, and P2 µL).
    • 19.2.
      Aluminum foil.
    • 19.3.
      Vacuum aspirator (optional). If not available, a P1000 pipette can be used instead.
    • 19.4.
      4°C refrigerator or camera.
    • 19.5.
      Vortex mixer.
    • 19.6.
      Plastic disposable Pasteur pipettes.

Table 1.

Sequence of antibodies used to phenotype astrocytes and microglia across 8 cycles of multiplex fluorescent immunohistochemistry.

Cycle	Primary antibody	Target antigen	Host	Clone	RRID	Vendor (Catalog No.)	Dilution	Secondary antibody
1	MHC2	Major histocompatibility complex class II	Ms	CR3/43	AB_2313661	Dako (M0775)	1:50	Dk-Cy3 anti-Ms
	TSPO	18 kDa translocator protein	Rb	EPR5384	AB_10862345	Abcam (ab109497)	1:100	Dk-AF488 anti-Rb
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt
2	EAAT2	Excitatory amino acid transporter 2	Ms	1H8	AB_11008461	Thermo Scientific (MA5-12360)	1:50	Dk-Cy3 anti-Ms
	TMEM119	Transmembrane protein 119	Rb	Polyclonal	AB_2681645	Sigma-Aldrich (HPA051870)	1:100	Dk-AF488 anti-Rb
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt
3	CD68	CD68 molecule	Ms	KP1	AB_2314148	Dako (M0814)	1:50	Dk-Cy3 anti-Ms
	EAAT1	Excitatory amino acid transporter 1	Rb	Polyclonal	AB_304334	Abcam (ab416)	1:100	Dk-AF488 anti-Rb
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt
4	ALDH1L1	Aldehyde dehydrogenase 1 L1	Ms	N103/39	AB_2687399	Sigma-Aldrich (MABN495)	1:500	Dk-Cy3 anti-Ms
	IBA1	Ionized calcium-binding adapter molecule 1	Rb	Polyclonal	AB_839504	Wako (019-19741)	1:250	Dk-AF488 anti-Rb
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt
5	VIM	Vimentin	Ms	V9	AB_477627	Sigma-Aldrich (V6630)	1:1,000	Dk-Cy3 anti-Ms
	FTL	Ferritin	Rb	Polyclonal	AB_259684	Sigma-Aldrich (F6136)	1:500	Dk-AF488 anti-Rb
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt
6	HuC/D	Hu-antigen C/D	Ms	15A7.1	Not available	Sigma-Aldrich (MABN153)	1:500	Dk-Cy3 anti-Ms
	YKL-40	Chitinase 3-like protein 1	Rb	Polyclonal	AB_2608881	Thermo Fisher (PA5-43746)	1:100	Dk-AF488 anti-Rb
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt
7	GS	Glutamine synthetase	Ms	GS-6	AB_2110656	Sigma-Aldrich (MAB302)	1:1,000	Dk-Cy3 anti-Ms
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt
8	PHF1	Paired helical filament tau phosphorylated at Ser396 and Ser404	Ms	PHF1	AB_2315150	Provided by Dr. Peter Davis	1:200	Dk-AF488 anti-Ms
	Aβ	Aβ	Rb	D54D2	AB_2797642	Cell Signaling Technologies (#8243S)	1:500	Dk-Cy3 anti-Rb
	GFAP	Glial fibrillary acidic protein	Gt	Polyclonal	AB_880202	Abcam (ab53554)	1:100	Dk-Cy5 anti-Gt

Abbreviations: AF488 = AlexaFluor 488; Cy = cyanine; Dk = donkey; Gt = goat; Ms = mouse; Rb = rabbit. All secondary antibodies were used at a 1:200 dilution and purchased from Jackson ImmunoResearch Labs, West Grove, PA.

3. Methods

Figure 1a depicts a schematic of the protocol. Briefly, the first cycle is a traditional fluorescent IHC on the FFPE tissue sections followed by whole-slide imaging. Next, the coverslips are removed and the sections are treated to both denature and strip the primary and secondary antibodies and to quench any remaining fluorescent signal from the fluorophore-conjugated secondary antibodies. In FFPE sections, microwaving is also helpful for retrieval of many antigens whose epitopes are hidden by the extensive cross-linking caused by the formaldehyde fixative [21]. After this treatment, the sections are ready for a new cycle of IHC with the next set of primary and secondary antibodies. This process of IHC-stripping/quenching-imaging is repeated up to 8 cycles. Lastly, the resulting images are aligned and subjected to the quantitative analysis pipeline.

3.1. First cycle of immunohistochemistry

1. Dewaxing: place the FFPE tissue sections in the slide rack and immerse twice for 10 min in successive glass dishes containing xylenes to remove the paraffin from the tissue (see Note 8).

2. Rehydration: immediately immerse the sections in solutions with decreasing concentrations of ethanol as follows: two different dishes with 100% ethanol for 5 min each, 95% ethanol for 5 min, and 70% ethanol for another 5 min (see Note 9).

3. Wash: immerse the slides in distilled water for 5 min or until being ready for step #4.

4. Antigen retrieval: microwave at high power a microwave-compatible dish filled with citrate buffer until boiling (~3-4 min). Immediately, switch the sections from the dish with distilled water to the dish containing boiling citrate buffer. Heat for 20 min at 95°C (i.e., typically a low power program) (see Note 10).

5. Cooling: cool the slides for ~45 min at 4°C (see Note 11).

6. Wash: immerse the slides in a dish containing TBS for 5 min. Repeat this step twice.

7. Hydrophobic barrier: blot the glass slides with a wipe to remove the excess TBS. Place the samples horizontally in a slide staining tray and outline the tissue using a hydrophobic barrier pen (see Note 12). Wait ~30 s to let the barrier dry.

8. Blocking: to avoid non-specific antibody binding, gently pipette 300 μL of blocking buffer on top of each tissue section and incubate with low speed shaking for 1 hour.

9. Primary antibody incubation: dilute the primary antibodies in antibody buffer at the appropriate concentrations (see Note 13). Remove the blocking buffer with vacuum aspiration (recommended) or a pipette (avoiding damage to the tissue) and add 300 μL of the primary antibody solution onto each section. Incubate overnight at 4°C without shaking in a slide moisture chamber (see Note 14).

10. Wash: remove the primary antibody solution and wash each section with 300 μL of TBS for 5 min with gentle shaking. Repeat this step twice.

11. Secondary antibody incubation: prepare a 1:200 dilution of secondary antibodies in antibody buffer. Once prepared, the secondary antibody solution should be protected from the light until its use (i.e., cover the conical tube with aluminum foil). Aspirate the TBS from the sections and apply 300 μL of the secondary antibody solution onto each section. Incubate for 2 hours at RT with gentle shaking in the dark (see Note 15).

12. Wash: remove the secondary antibody solution and wash each section with 300 μL of TBS for 5 min with gentle shaking in the dark. Repeat this step twice.

13. Autofluorescence elimination (optional, see Note 16): add 300 μL of 70% ethanol onto each section and incubate for 5 min protected from light. Add a few drops of autofluorescence eliminator reagent (or 1% Sudan Black B in 70% ethanol) to cover each section and incubate for another 5 min. After 5 min, remove the excess autofluorescence eliminator reagent by immersing the sections in three serial 1-min washes in fresh 70% ethanol.

14. Wash: rinse sections with 300 μL of TBS for 5 min in the dark.

15. DAPI staining: aspirate the TBS and add 300 μL of DAPI solution onto each section. Incubate 10 min with gentle shaking at RT in the dark.

16. Wash: rinse the sections with 300 μL of TBS for 10 min with gentle shaking at RT in the dark. Repeat this step twice.

17. Cover the slides: aspirate the TBS and add a few drops of aqueous mounting medium with DAPI (see Note 17) using a plastic Pasteur pipette. For a 2-3 cm² tissue section, 3-4 drops should suffice to cover the entire tissue area. Carefully, place the coverslip on the slide and, using your fingers or the bottom of a tip to apply a light pressure, move any air bubbles away from the tissue ensuring that the tissue section is completely covered by mounting medium without deforming (flattening) the tissue (see Note 18). Blot the excess mounting medium with a wipe. Keep the slides at 4°C in the dark until imaging.

3.2. Imaging

1. Cleaning: gently wipe the coverslip and the bottom of the slide with 70% ethanol. Care should be taken to avoid displacing the coverslip and damaging the underlying tissue.

2. Determination of image acquisition parameters: place the slide in the fluorescence microscope and select the optimal acquisition parameters, mainly exposure time. To keep consistency across different slide batches, each marker should be scanned with the same exposure time for all sections (see Note 19). An exception to this rule is the markers to be used for alignment (i.e., DAPI and GFAP in this protocol), for which the parameters providing the best signal-to-noise ratio in the second and later cycles should prevail (see Note 20).

3. Scanning: if possible, scan the entire section using a fluorescence microscope (see Note 21).

4. Review output images for quality control: make sure that all sections are properly focused and have the same acquisition parameters, and that these parameters are valid across all sections (i.e., no saturation) before proceeding with the next step. Rescan sections that have out-of-focus areas if needed.

3.3. Antibody denaturalization and fluorescence quenching

1. Coverslip removal: immediately after imaging, place the slides in a rack and immerse them in a dish with coverslip removal solution. Wait until the coverslips fall off completely. Time necessary will vary from minutes to hours depending on whether air bubbles were purposefully left between the coverslips and the glass slides and the time past between coverslip placement and removal. If some slides are still covered after a few hours, gently pull up and immerse the rack ~3-5 times to facilitate this process.

2. Wash: after each coverslip drops, move the slides to another dish filled with TBS. Wait until all coverslips have dropped and wash the sections with TBS for at least 5 min before starting the next step.

3. Antibody denaturing: use the conditions previously described in steps #4-6 of section 3.1.

4. Fluorescence quenching: immediately after mixing the H₂O₂, add the fluorescence quenching solution to a Coplin jar and immerse the slides into it. Incubate for 30 min at RT (see Note 22).

5. Wash: immerse the slides in TBS for 5 min three times (see Note 23).

3.4. Second and successive cycles of immunohistochemistry

1. New IHC cycle: repeat steps #8-12 and #15-17 from section 3.1., using the recommended combination of antibodies for each IHC cycle (Fig. 1b and Table 1). Figure 2 shows an example ROI from the same tissue section labeled with each of 16 antibodies, plus the nuclear dye DAPI. The last four images illustrate the merge of all 17 channels, the 6 microglial markers, the 7 astrocytic markers, and the 2 AD pathological hallmarks (Aβ plaques and phospho-tau neurofibrillary tangles) together with DAPI and the neuronal marker HuC/D, respectively. See also Supplemental Videos for an animated version. This protocol could be customized with as many IHC cycles and antibodies as desired as long as antibodies can be stripped efficiently, and the tissue sections withstand the procedures (see Note 24).

2. Proceed to imaging (step #3.2.).

Figure 2. Example of cyclic multiplex fluorescent immunohistochemistry from a cortical region of interest (ROI).

The individual images of an ROI from layer III of the same temporal cortex FFPE section of an AD donor, which was immunolabeled with 16 different primary antibodies across 8 IHC cycles and stained with the nuclear dye DAPI, are shown. Markers are distributed by cell type/feature (astrocytes, microglia, and AD lesions/neurons/DAPI). The last 4 merge images result from the alignment of all 17 channels, the 6 microglial markers, the 7 astrocytic markers, and the 2 AD pathological hallmarks (Aβ plaques and p-tau neurofibrillary tangles) together with DAPI and the neuronal marker HuC/D, respectively. Scale bar: 50 μm.

3.5. Image analysis

We provide two alternative workflows for the region of interest (ROI) selection, cropping, and alignment steps, which are graphically summarized in Figure 3. While in our recent publication [15] we used the cellSens image analysis software (Olympus, Tokyo, Japan) for alignment, here we also propose an open-source option using the Fiji distribution of the public domain software ImageJ [22]. The ImageJ/Fiji macro is available here: https://github.com/serrano-pozo-lab/glia-ihc.

Figure 3. Image analysis pipeline.

HuC/D neuronal staining (scale bar: 100 μm) is used to define the cortical layers (I-VI) vs. the white matter (WM). The GFAP (or DAPI) channel repeated in each IHC cycle is used for alignment of all images across IHC cycles using the cellSens Alignment Module or the ImageJ/Fiji MultiStackReg plugin. Single-cell segmentation of the resulting high-plex stacks is achieved using either a semiautomated approach with VGG VIA followed by a custom ImageJ/Fiji macro (available in https://github.com/serrano-pozo-lab/glia-ihc) or an automatic segmentation tool such as QuPath Cell Detection. Quantitative analyses may include disease vs. control comparisons of mean gray intensity data by marker, clustering of high-plex single-cell level intensity data to identify phenotypes or states within a cell type, and machine learning classifiers based on single-cell level intensity data or on all features of high-plex single-cell images (i.e., convolutional neural networks).

1. Image file format conversion (.vsi to .tif): opening the .vsi image files provided by the Olympus VS120 slide scanner with ImageJ/Fiji requires an additional step of conversion to .tif format using the OlympusViewer plugin.

2. ROI selection and cropping: create individual ROIs encompassing the same area across all IHC cycles of the same tissue section (see Note 25). Although the ROI location will vary depending on the research aims and the antibodies used, we chose ROIs from each of the six cortical layers of the temporal neocortex, which we identified using the HuC/D neuronal marker (Fig. 3). Thus, full-width 1 mm-long rectangular sections were cropped from each cortical layer to new images using the “Crop” tool in the ImageJ/Fiji Image menu or “Crop to new image” tool in cellSens. If using cellSens, it is recommended to rename the channels to the corresponding cell marker to facilitate their recognition in later steps. Next, the same ROI is selected and cropped to new images from each of the whole-slide images obtained in the remaining 7 IHC cycles of the same tissue section, using DAPI and GFAP as reference markers (see Note 26).

3. Alignment: the alignment process will differ depending on the software employed but see Note 27 for the possible troubleshooting needed in both cases.

   1. ImageJ/Fiji: download the MultiStackReg v1.45 plugin if not already listed in the Plugins > Registration menu. Split the channels from each IHC cycle image and save them as individual files (Image > Color > Split Channels). Then, create a channel stack (Image > Stacks > Images to Stack) for all the images of each of the channels across all IHC cycles (i.e., one stack for all DAPI images, another for all Cy3-stained markers, and so on). Change all channel stacks to 8-bit grayscale images (Image > Type > 8-bit). In the MultiStackReg dialog, select “DAPI” as “Stack 1,” “Align” as “Action 1,” and Rigid Body as “Transformation,” then click on “Save Transformation File” followed by “Ok,” and save the resulting transformation file. If another marker has been included as reference in all IHC cycles, repeat the same process with this channel stack (in our protocol, the Cy5 stack corresponding to GFAP). Choose the best aligned stack (e.g., DAPI or GFAP) and use its transformation file to align all the other channel stacks with MultiStackReg (i.e., now select “Cy3” as “Stack 1” and “Load transformation file” in the “Action 1” dropdown menu). If two reference markers have been included in the cyclic IHC protocol, the one with worse alignment should also be realigned with the same transformation file as the other stacks. After aligning all channel stacks, convert the stacks back to images (Image > Stacks > Stack to Images) and, finally, convert all images from all 4 channels and all IHC cycles to a single stack (Image > Stacks > Images to Stack) to obtain a high-plex image.
   2. cellSens: ensure that all .vsi image files corresponding to all IHC cycles of the same ROI are stored in one single folder and rename them using the same name followed by consecutive numbers corresponding to their position in the cyclic IHC sequence. Save this folder into a directory folder with the same name. In the Alignment Module, select the directory folder as “Source directory,” the folder containing the images as “Subdirectory,” and the desired output folder as “Destination directory.” Select DAPI or GFAP as reference channel. Tune the “Size of search area” and “Correlation threshold” to be the highest working parameters. Once aligned, convert the .vsi files to .tif format, using the OlympusViewer plugin of the ImageJ/Fiji software before the next step.

4. Background subtraction: to minimize background differences across sections, apply a regional background subtraction to all images with the rolling ball algorithm in ImageJ/Fiji (Process menu > Subtract Background…). The rolling ball algorithm determines a local background value for every pixel by averaging over a given radius around the pixel, and this value is then subtracted from the original image to remove spatial variations in the background intensities [23]. In our protocol we used a 200-pixel radius, but users should test several radii before selecting the best option.

5. Single-cell segmentation

   • 5.1.
     ImageJ/Fiji:

     • 5.1.1.
       Manual annotation of individual profiles: for single-cell analyses, it is necessary to segment thousands of individual cell profiles. While sophisticated machine learning-based methods have been developed for automatic cell segmentation (e.g., QuPath [24]), a simple manual annotation tool such as the open-access Visual Geometry Group (VGG) Image Annotator (VIA) (https://www.robots.ox.ac.uk/~vgg/software/via/) may also suffice. Using the VIA box tool, manually annotate all ALDH1L1+ astrocytes and IBA1+ microglia as well as all Aβ plaques and p-tau+ neurofibrillary tangles in the cropped images.
     • 5.1.2.
       Profile segmentation: parse VIA profile annotations using R and convert them to ImageJ/Fiji ROIs, then segment single-cell ROIs using the ImageJ/Fiji adaptive Otsu thresholding (Image > Adjust > Threshold > Otsu).
     • 5.1.3.
       Mean gray intensity (MGI) measurement and pre-processing: measure the MGI for each astrocytic marker within each ALDH1L1+ astrocytic profile and for each microglial marker within each IBA1+ microglial profile using the Measure tool within the ImageJ/Fiji Analyze menu. Since intensity-based measurements of protein expression often follow log-normal distributions, apply a natural log transformation, and obtain the z-scores of the log-transformed MGI values [25]. While MGI is typically used as a proxy for protein expression level in fluorescent IHC, for some research objectives the area fraction (i.e., the fraction of the profile ROI occupied by each marker’s immunoreactivity) may be of interest as well. To select the parameters of interest, click on “Set Measurements” menu within the ImageJ/Fiji Analyze menu.
     • 5.1.4.
       Distance between cell profiles and AD pathological lesions: record the XY coordinates of each astrocyte, microglia, Aβ plaque, and p-tau+ neurofibrillary tangle profile within each image crop, which are provided by the Measure tool in the ImageJ/Fiji Analyze menu. With these coordinates, the distance between each glial profile and the nearest Aβ plaque or -tau+ neurofibrillary tangle can be calculated using R. For Aβ plaques, the radius can be back-calculated from the area (assuming they are round) and should be subtracted from each raw distance measurement to obtain the exact distance between each glial profile and the nearest plaque edge (rather than the center of the nearest plaque).
   • 5.2.
     QuPath

     • 5.2.1.
       Image preparation: create a new project and open the .vsi high-plex image files or drag and drop them into the QuPath window. Select Image Type as Fluorescence and double click on each image to select it and display it in the main window. To select and view each of the channels, select them in the View > Brightness and Contrast menu. Rename the channels with the corresponding marker names.
     • 5.2.2.
       Automatic profile segmentation: draw a rectangular annotation (i.e., yellow box) over the entire image, then open the cell detection menu (Analyze > Cell detection) and select the detection channel of interest (e.g., ALDH1L1 for astrocytes, IBA1 for microglia, Aβ for plaques, and p-tau for tangles). Tune the cell detection parameters to optimize the segmentation results and click Run.
     • 5.2.3.
       MGI and distance measurements: go to the Hierarchy > Annotation (Rectangle) to see a summary of the measurements for the annotation rectangle (i.e., number of detected cells). Open the drop-down list of all the cells detected within the Annotation rectangle and see the measurements for each cell across all channels in the bottom window. To tabulate all measurements, go to Measure > Show detection measurements. The “Nucleus: Marker mean” column contains the MGI of each cell, whereas the “Centroid X px” and “Centroid Y px” columns correspond to the X and Y coordinates of the cells, which will be used to calculate their distance to the nearest AD pathological hallmark. Click “Save” to export these results for downstream analyses. To crop the cells detected in QuPath to individual single-cell images in ImageJ/Fiji, select all detected cells in the Hierarchy window and click on the ImageJ tollbar button or go to Extensions > ImageJ > Send region to ImageJ. In the dialog box, keep “Resolution” at 1 pixel (i.e., no downsampling) and unselect “Include ROI,” “Include overlay,” and “Apply color transforms;” ImageJ/Fiji should automatically open with the single-cell ROIs created.

6. Quantitative analyses: the highly dimensional dataset resulting from this image analysis pipeline can inform various downstream analyses. First, in the simplest approach, single-cell MGI data for each marker can be compared between control and diseased brains to determine whether there are statistically significant changes in its expression levels associated with the disease. To this end, mixed effect models are most appropriate to control for within donor’s correlation (since profiles from the same donor are not independent observations) [26]. Second, unsupervised machine learning clustering algorithms could be applied to cluster MGI data into different cellular phenotypes or states based on the expression levels of the selected combination of markers. These include spectral clustering, hierarchical clustering, and k-means clustering, among others. Lastly, supervised classifiers can be trained to discern between health and disease diagnoses using either the MGI data or the actual high-plex single-cell images (via convolutional neural networks).

Supplementary Material

Video 1

Supplemental Video 1. High-plex images of six microglial markers and the two AD neuropathological hallmarks. Individual (left) and aligned (right) pictures of the same ROI from the temporal cortex layer III of an AD donor stained for the nuclear dye DAPI and immunolabeled for 6 microglial markers (MHC2, TSPO, TMEM119, CD68, IBA1, and ferritin [FTL]), plus Aβ plaques and PHF1+ neurofibrillary tangles (Note: individual images are also shown in Figure 2). Scale bar: 50 μm.

Video 2

Supplemental Video 2. High-plex images of seven astroglial markers and the two AD neuropathological hallmarks. Individual (left) and aligned (right) pictures of the same ROI from the temporal cortex layer III of an AD donor stained for the nuclear dye DAPI and immunolabeled for 7 astroglial markers (GFAP, EAAT2, EAAT1, ALDH1L1, vimentin [VIM], YKL-40, and glutamine synthetase [GS]), plus Aβ plaques and PHF1+ neurofibrillary tangles (Note: individual images are also shown in Figure 2). Scale bar: 50 μm.

Video 3

Supplemental Video 3. High-plex images of all 16 markers (plus DAPI) across 8 cycles of multiplex fluorescent immunohistochemistry on the same FFPE brain section. Individual (left) and aligned (right) pictures of the same ROI from the temporal cortex layer III of an AD donor stained for the nuclear dye DAPI and immunolabeled for 16 markers (in order: MHC2, TSPO, GFAP, EAAT2, TMEM119, CD68, EAAT1, ALDH1L1, IBA1, vimentin [VIM], ferritin [FTL], HuC/D, YKL-40, and glutamine synthetase [GS]), plus Aβ and PHF1 for plaques and neurofibrillary tangles, respectively. (Note: individual images are also shown in Figure 2). Scale bar: 50 μm.

Figure 4. Minimal damage to FFPE sections after eight IHC cycles.

Whole-slide scan images of the same FFPE section from an AD donor after cycle #1 (DAPI in blue, MHC2 in red, TSPO in green, and GFAP in magenta) and after cycle #8 (DAPI in blue, Aβ in red, PHF1 in green, and GFAP in magenta). White arrows point to small damaged areas. Scale bar: 2 mm.