Making Multiplexed Imaging Flexible: Combining Essential Markers With Established Antibody Panels

Abstract

Multiplexed immunofluorescence (IF) can be achieved using different commercially available platforms, often making use of conjugated antibodies detected in iterative cycles. A growing portfolio of pre-conjugated antibodies is offered by the providers, as well as the possibility for in-house conjugation. For many conjugation methods and kits, there are limitations in which antibodies can be used, and conjugation results are sometimes irreproducible. The conjugation process can limit or slow down the progress of studies requiring conjugation of essential markers needed for a given project. Here, we demonstrate a protocol combining manual indirect immunofluorescence (IF) of primary antibodies, followed by antibody elution and staining with multiplexed panels of commercially pre-conjugated antibodies on the PhenoCycler platform. We present detailed protocols for applying the workflow on fresh frozen and formalin fixed paraffin embedded tissue sections. We also provide a ready to use workflow for coregistration of the images and demonstrate this for two examples.

Methods

Cell Culture

U2OS cells were grown in McCoy’s 5a complete growth medium (Merck) supplemented with 10% fetal bovine serum (FBS) (Merck) and 1% l-glutamine (Merck). The cells were subcultured twice a week with 0.25% Trypsin-1 mM EDTA solution (Merck), in 1:6 ratio. To grow U2OS cells on glass slides, the slides were pre-incubated with 0.1% poly-L-lysine (Merck) at room temperature (RT) overnight. The coated slides were washed with autoclaved distilled water and sterilized with a UV lamp under the cell culture hood for 1 hr. U2OS cells were then grown on the sterilized glass slides inside a 10-cm culture dish. Sometimes the cells were also grown on optical 96-well plates for multi-well imaging experiments.

Human Tissue Samples

For tonsil, fresh frozen (FF) anonymized residual material was collected after tonsillectomy performed at Sophiahemmet, Stockholm Sweden (contact stated in Appendix 1). For bladder cancer formalin-fixed paraffin embedded (FFPE) tissue, anonymized residual material was collected from the pathology archive (contact stated in Appendix 1). For prostate tissue, sections were cut from archival tissue blocks from a radical prostatectomy and lymph node dissection performed on a patient recruited to the PROMOTE Study ¹ (contact stated in Appendix 1). Specimens were processed for routine diagnostic reporting by fixation in neutral buffered formalin (Merck) for approximately 24 hr and transferred to 70% ethanol (VWR, Darmstadt, Germany) before embedding in wax. For the purpose of this study, 5-µM sections were cut onto SuperFrost plus gold slides, baked for 30 min at 60C, and transported at room temperature.

A tissue microarray (TMA) containing nine normal and 10 cancer samples was constructed by Atlas Antibodies (Stockholm, Sweden) from FFPE human tissues. FFPE tissue blocks were purchased from commercial biobanks (Asterand, BioIVT, Burgess Hill, UK; Tissue Solutions Ltd, Glasgow, UK; and Indivumed GmbH, Hamburg, Germany) with anonymized patient material only. Before TMA construction, tissue morphology and antigenicity were quality controlled by Atlas Antibodies. TMA cores of the following tissue types are shown in Fig. 2 and A3: rectum, lymph node, colorectal cancer. FF samples were stored at −80C and FFPE at RT, respectively.

Figure 2.

Optimizing elution conditions for integration of manual indirect immunofluorescence with multiplexed immunofluorescence in cells, FF and FFPE tissue samples. (A) Indirect immunofluorescence of three protein targets: PDS5A (green), alpha tubulin (red) and calreticulin (yellow) in U2OS cells before and after elution using in-house prepared 4i buffer and Lunaphore’s commercial elution buffer. (B) Indirect immunofluorescence of calreticulin (yellow) and alpha tubulin (red) in FF tonsil sections before and after elution using in-house prepared 4i buffer and Lunaphore’s commercial elution buffer. (C) Indirect immunofluorescence of SATB2 (green) and pan-cadherin (red) in an FFPE TMA (rectum core) before and after elution using in-house prepared 4i buffer and Lunaphore’s commercial elution buffer. Scale bar = 20 µm.

Microscopy and Image Acquisition

A Leica DMi8 inverted fluorescent microscope (Wetzlar, Germany) was used for preliminary assessment of fluorescent staining in cells (ATCC, Manassas, United States). Using a 20× or 40× objective, the fluorescent images were taken with highest resolution. Sometimes a 10× objective was also used to take images from 96-well plates. Akoya Biosciences’ Phenoimager (Marlborough, MA) was primarily used to acquire the fluorescent images from tissue samples, using a 20× objective.

Manual Immunofluorescence Steps in U2OS Cells

U2OS cells grown on glass slides were fixed with 4% paraformaldehyde (PFA) (Thermo Scientific, Waltham, MA) in 1× phosphate buffered saline (PBS) (Thermo Scientific) at RT for 30 min, permeabilized with 0.1% TritonX-100 (Merck) in PBS at RT for 10 min and blocked with 4% FBS in PBS at RT for 30 min. Primary antibodies (Table 1) (Atlas Antibodies; Abcam, Cambridge, UK) in the blocking solution were incubated at 4C overnight and washed off twice with PBS. Alexa Fluor -488, -555, and -647 conjugated secondary antibodies (See Appendix 1 for details) (Thermo Scientific) were used in 1:800 dilution in blocking solution containing DAPI (Merck) at RT for 1.5 hr, and washed off thrice with PBS. Stained cells were then imaged using 10×, 20×, and 40× objective.

Table 1.

Protein Markers and Antibodies for Evaluation of Elution Buffers in U2OS Cells.

No.	Antibody ID	Protein	Subcellular Location	Antibody Dilution	Host	Exposure Time (ms)
1	HPA036662, Atlas Antibodies	PDS5A cohesin-associated factor	Nucleus	1:200	Rabbit	20
2	Ab7291, Abcam	Alpha Tubulin	Cytosol, cytoskeleton	1:1000	Mouse	10
3	Ab2908, Abcam	Calreticulin	Endoplasmic reticulum	1:100	Chicken	10

Manual Immunofluorescence Steps in FF Tissue Samples

FF tonsil samples were used in this study. In the protocol, the FF tissue samples were removed from the −80C freezer and left on ice to thaw for 15–20 min. During this time, a hydrophobic circle was drawn around the tissue using a PAP pen (VWR) and the marking was dried for 10 min. The sample was then incubated with acetone (Merck) for 10 min in a humidity chamber, air dried further for 2 min and washed twice with 1× hydration buffer (Akoya Biosciences). The FF sections were then fixed with 4% PFA diluted in PBS for 15 min in a humidity chamber, washed thrice with PBS and incubated in PBS with 4% FBS for 30 min at RT using the humidity chamber. Primary antibodies (Table 2) were diluted in PBS with 4% FBS and incubated overnight at 4C in a humidity chamber. After washing thrice with PBS, the sample was stained with Alexa Fluor-conjugated secondary antibody (1:800) dissolved in PBS with 4% FBS buffer for 1.5 hr at RT using the humidity chamber. The sample was then washed thrice with PBS, stained with DAPI (Merck) for 10 min at RT and washed again with PBS in a humidity chamber. The stained tissue section was then mounted with 1× CODEX buffer (Akoya Biosciences) and a glass coverslip, and imaged using Akoya Biosciences’ Phenoimager with a 20× objective.

Table 2.

Primary Antibodies Used for Indirect Immunofluorescence in FF Tissue Samples.

No.	Antibody ID/Barcode	Protein	Subcellular Location	Antibody Dilution	Host of Antibody Production	Exposure Time (ms)
1	EPR21950-241Abcam a , BX002 b	Eomes	Nuclear	1:25	Rabbit	500
3	Ab7291, (Abcam)	Alpha Tubulin	Cytosol, cytoskeleton	1:1000	Mouse	10
4	Ab2908, (Abcam)	Calreticulin	Endoplasmic reticulum	1:400	Chicken	10

^aIn-house conjugated Eomes antibody with the Akoya barcode^b used.

Manual immunofluorescence steps in formalin fixed and paraffin embedded (FFPE) tissue samples—FFPE tissue sections were baked at 60C for 30–60 min, and the excess wax was cleaned with a kimwipe paper. For TMA samples, freshly cut sections were incubated overnight at 60C to avoid loss of sample during incubation and washes. Dewaxing was done by sequentially washing the sample for 5 min in each of the following solutions: twice in histochoice clearing solution (Merck), twice in 100% ethanol (VWR), 90% ethanol, 70% ethanol, 50% ethanol, 30% ethanol and twice in distilled water. Antigen retrieval was done by immersing the sample in 1× Tris-EDTA buffer, pH 9.0 (VWR) and using a pressure cooker (BioSB, Goleta, CA) for 20 min at high temperature setting. The sample was then cooled down to RT and washed with distilled water. Autofluorescence was quenched using a 1× quenching buffer: 4.5% (w/v) hydrogen peroxide (Merck) and 20 mM sodium hydroxide solution (Merck) in PBS. The sample was immersed inside the quenching buffer and placed between two LED lamps at the highest setting for 45 min. This step was repeated with a freshly prepared quenching buffer. After this step, the sample was cooled down to RT for 15 min by washing thrice with PBS. Blocking was done with 4% FBS-PBS solution and the sample washed with PBS. Primary antibodies were prepared as described in Table 3, in 4% FBS-PBS and incubated overnight at +4C in a humidity chamber. After washing thrice with PBS, secondary antibodies with Alexa Fluor -555 or -647 conjugates were added in 1:800 dilution in 4% FBS-PBS solution at RT for 1.5 hr in a humidity chamber. The stained section was incubated with DAPI for 10 min at RT and washed with PBS. The sample was then mounted and imaged as described above.

Table 3.

Primary Antibodies Used for Indirect Immunofluorescence in FFPE Tissue Samples.

No.	Antibody Catalog Number	Protein	Subcellular Location	Antibody Dilution	Host of Antibody Production	Exposure Time (ms)
1	Ab194979, (Abcam)	HER2	PM	1:50	Rabbit	500
2	AMAb90682, (Atlas Antibodies)	SATB2	Nucleus	1:200	Mouse	450
3	Ab239839, (Abcam)	Pan-Cadherin	PM	1:200	Rabbit	450
4	Ab248451 (Abcam) a , BX006b	MSMB	Secreted	1:100	Rabbit	150

^aIn-house conjugated MSMB antibody with the Akoya barcode^b used.

4i buffer elution of manually stained samples—1× 4i elution buffer was prepared by dissolving 0.5 M L-Glycine (Merck), 3M Urea (Merck), 3M Guanidinium chloride (Merck), and 70 mM TCEP (Merck) in a total volume of 5 ml and a pH of 2.5 adjusted with 37% HCl (Merck). The solution was used fresh and stored in −20C until use. For elution in U2OS cells, 200 µl/wash was used on the glass slide at 37C for 10 min and the cells were washed for a total of three times. The cells were then washed thrice with PBS and later mounted with 1× CODEX buffer and a glass coverslip. Images were taken with a Leica DMI8 fluorescent microscope or Akoya Biosciences’ Phenoimager, as mentioned above.

For elution in FF tissue samples, depending on the size of the tissue section, 200–400 µl/wash of 4i buffer was used on the tissue section in a humid chamber for 10 min at RT. The 4i buffer wash was repeated one more time and then washed thrice with PBS. The eluted sample was then mounted with 1× CODEX buffer and a glass coverslip. After imaging the sample using Akoya Biosciences’ Phenoimager with a 20× objective, the coverslip was removed by washing thrice in PBS. The FF tissue sample was then used for the routine multiplex staining protocol.

For elution in formalin fixed and paraffin embedded (FFPE) tissue samples, depending on the size of the tissue section, 200-400 µl/wash of 4i buffer was used at 37C for 10 min and for a total of three washes. The sample was then washed thrice with PBS and mounted with 1× CODEX buffer using a glass coverslip. Imaging was done using Akoya Biosciences’ Phenoimager with a 20× objective. The coverslip was then removed by washing thrice with PBS. The FFPE tissue sample was then used for routine multiplex staining protocol. All the steps were similar for TMA FFPE samples.

Lunaphore buffer elution of manually stained samples—1× Lunaphore elution buffer was prepared by mixing solution I (1×) and II (20×) (Lunaphore, Lausanne, Switzerland). The elution buffer was prepared fresh and stored in +4C after use. It can be reused several times, if stored properly. For elution in U2OS cells, 200 µl/wash was used on the glass slide at 50C for 10 min and the cells were washed for a total of three times. After washing thrice in PBS, the eluted samples were mounted and imaged as described previously.

For elution in FF tissue samples, 200–400 µl/wash of Lunaphore elution buffer was used at 37C for 10 min and for a total of three washes. The samples were then washed thrice with PBS, mounted and imaged as described previously. The samples were also used downstream for the multiplex staining protocol.

For elution in FFPE and TMA FFPE samples, 200–400 µl/wash of Lunaphore elution buffer was used at 50C for 10 min and for a total of three washes. The samples were then washed thrice with PBS, mounted and imaged as described previously. The samples were then used for the downstream multiplex staining protocol.

Multiplexed Staining of FF Tissue Samples With PhenoCycler

Tissue sections were prepared following the recommended standard protocol for PhenoCycler run on FF sections according to the Akoya PhenoCycler-Fusion user guide (PD-000011 REV M, Akoya Biosciences). Shortly, post-elution, FF tissue samples were washed twice with 1× hydration buffer, incubated in 1× staining buffer (Akoya Biosciences) for 20 min at RT and incubated overnight at +4C in a humidity chamber with pre-conjugated Akoya Biosciences primary antibodies (Table 4), diluted in blocking buffer (sample kit, Akoya Biosciences) according to manufacturer’s instructions. Later, the samples were washed twice with staining buffer and incubated with a post-staining fixation solution, containing 1.6% PFA in 1× hydration buffer at RT for 10 min. After washing thrice with PBS, the samples were incubated with ice cold methanol (Merck) for 5 min in a humidity chamber. The samples were then washed thrice with PBS and incubated with 1× CODEX fixative solution (Akoya Biosciences) for 20 min at RT in a humidity chamber, according to manufacturer’s instructions. The samples were then washed thrice with PBS and kept in PBS until use. Next, multiplex staining and imaging experimental setup procedures were followed in the PhenoCycler Fusion instrument (Akoya Biosciences), as described below.

Table 4.

Akoya Pre-Conjugated Antibodies as Used in the Control and Post Elution Phenocycler Runs FF Tissue Samples.

Cycle in PC Run	Catalog No. Pre-Conjugated Antibodies	Protein	Subcellular Location	Antibody Dilution	Reporter Catalog Number and Fluorophore	Exposure Time for Control Run (ms)	Exposure Time Post Elution (ms)
C1	4550103	CD3	PM	1:100	Alexa Fluor 647-RX015 for PhenoCycler	200	200
C1	4250019	Ki67	Nucleus	1:100	Atto 550-RX047 for PhenoCycler	200	200
C1	4150020	Pan-Cytokeratin	Cytosol	1:100	Alexa Fluor 488-RX019 for PhenoCycler	75	75
C2	4350010	CD4	PM	1:100	Alexa Fluor 647-RX021 for PhenoCycler	200	200
C2	4250009	CD31	PM	1:100	Atto 550-RX032 for PhenoCycler	200	200
C3	4150003	CD45	PM	1:100	Alexa Fluor 488-RX001 for PhenoCycler	200	200
C3	4250023	CD45-RO	PM	1:100	Atto 550-RX017 for PhenoCycler	200	200
C4	4250006	HLA-DR	PM	1:100	Atto 550-RX026 for PhenoCycler	200	200

AF = Alexa Fluor. PM = plasma membrane.

Multiplexed Staining of FFPE Tissue Samples Using PhenoCycler

Tissue sections were prepared following the recommended standard protocol for PhenoCycler run on FFPE sections according to the Akoya PhenoCycler-Fusion user guide (version PD-000011 REV M, Akoya). Shortly, after elution, the FFPE samples were washed thrice with PBS, twice with 1× hydration buffer and incubated in 1× staining buffer for 30 min at RT and incubated overnight at 4C in a humidity chamber with pre-conjugated Akoya primary antibodies (Tables 5 to 7), diluted in blocking buffer according to manufacturer’s instructions (Akoya Biosciences). Other downstream steps were followed as in FF tissue samples, mentioned above.

Table 5.

Akoya Pre-Conjugated Antibodies as Used in the Control and Post Elution Phenocycler Runs for FFPE Bladder Cancer Samples.

Cycle in PC Run	Catalog No. Pre-Conjugated Antibodies	Protein	Subcellular Location	Antibody Dilution	Reporter Catalog Number and Fluorophore	Exposure Time for Control Run (ms)	Exposure Time Post Elution (ms)
C1	4250019	Ki67	Nucleus	1:200	Atto 550-RX047 for PhenoCycler	100	100
C1	4550113	CD68	Cytosol	1:200	Alexa Fluor 647-RX015 for PhenoCycler	250	250
C2	4450036	Beta-Catenin	PM	1:200	Atto 550-RX020 for PhenoCycler	150	150
C2	4450054 (4550124) a	PCNA	Nucleus	1:200	Alexa Fluor 647-RX036 for PhenoCycler	200	200
C3	4550119	CD3e	PM	1:150	Alexa Fluor 647-RX045 for PhenoCycler	200	200

^aCatalog numbers within brackets are updated product numbers currently valid for the same products, previously purchased with the upper catalog number.

Table 7.

Akoya Pre-Conjugated Antibodies as Used in the Control and Post Elution Phenocycler Runs for the Tissue Microarray of FFPE Tissue Samples.

Cycle in PC Run	Catalog No.	Protein	Subcellular Location	Dilution	Fluorophore	Exposure Time for Control Run (ms)	Exposure Time Post Elution (ms)
C1	4250019	Ki67	Nucleus	1:200	Atto 550-RX047 for PhenoCycler	200	200
C1	4150020	Pan-Cytokeratin	Cytosol	1:200	Alexa Fluor 488-RX019 for PhenoCycler	75	75
C2	4550119	CD3e	PM	1:200	Alexa Fluor 647-RX045 for PhenoCycler	200	200
C2	4450036	B-catenin	PM	1:100	Atto 550-RX020 for PhenoCycler	150	150
C2	4450018	CD20	PM	1:200	Alexa Fluor 750-RX007 for PhenoCycler	200	200
C3	4250012	CD8	PM	1:200	Atto 550-RX026 for PhenoCycler	200	200
C4	4550113	CD68	Cytosol	1:300	Alexa Fluor 647-RX015 for PhenoCycler	250	250
C5	4450054 (4550124) a	PCNA	Nucleus	1:300	Alexa Fluor 647-RX036 for PhenoCycler	200	200
C7	4450018	CD4	PM	1:200	Alexa Fluor 647-RX003 for PhenoCycler	200	200

AF = Alexa Fluor.

^aCatalog numbers within brackets are updated product numbers currently valid for the same products, previously purchased with the upper catalog number.

PhenoCycler Fusion and Imaging

A reporter plate with 1× master mix solution was prepared for each experimental run according to the Akoya Biosciences’ manual instructions. The master mix contains nuclease-free water, 1× CODEX buffer, 1× buffer additive, 1× assay reagent, and 1× nuclear stain, and this was prepared according to the manufacturer’s instructions (Akoya Biosciences). Corresponding fluorescent reporter oligos were added as 5 µl/oligo, according to the barcode present on the antibodies and the number of reporters per reaction depends on the number of antibodies in the corresponding cycle. Once the reporter plate with master mix was prepared, the plate was sealed with an adhesive foil cover (Thermo Scientific) and kept on ice until use. In the PhenoCycler Fusion instrument, the experimental design was generated according to the setup for each imaging cycle. In the meantime, FF or FFPE samples were equilibrated with 1× CODEX buffer for 10 min and, after cleaning the slides, the samples were fixed with a flowcell according to the manufacturer’s instructions. The flowcell-fixed samples were again equilibrated in 1× CODEX buffer to check for any leaks and left in the same buffer until use. After preparing all the necessary buffers required for the experimental run, the PhenoCycler Fusion instrument was initiated. After following the instructions prompted by the instrument, the sample slide was carefully cleaned with distilled water and 70% ethanol, using a kimwipe (VWR), to remove any remnant buffers. Using a sharpie marker pen, a border was drawn around the tissue section and the sample slide was fixed in the slide holder. Imaging was done using a 20× objective present in the PhenoCycler Fusion and the final processed images were saved as a qptiff file, along with the metadata information.

Measurement of nuclei distortion, relative fluorescence intensity and signal-to-noise ratio—Nuclear cell segmentation with cell expansion was performed on selected representative regions for tissue samples and entire images for all U2OS cell experiments using a pre-trained StarDist model (dsb2018_heavy_augment.pb) ² as an extension in qupath. ³ The settings used for segmentation can be found in Table 8. Cell intensity measurements and circularity measurements obtained along with cell segmentation parameters were exported to Python (version 3.10.9). For assessment of nuclear distortion nucleus circularity values for all cells, as given as an output parameter from nuclear segmentation were compared before and after elution. Mean values of the pixel fluorescence intensities of each cell and respective marker were used for the calculation of relative fluorescence intensity (RFI) since marker expression levels might vary in a non-uniform manner. ⁴ Pixel mean intensities from the nuclear segmentation mask were used for nuclear markers whilst for cytoplasmic or membranous markers pixel mean intensities from the whole cell mask was used. In short, the calculation was performed by first performing ratio normalization, setting the pre-elution mean of segmented cells to 100%. The RFI value was then given as the normalized mean value from the post-elution distribution of segmented cells. A one-sided Mann-Whitney U test was performed to evaluate if the decrease in mean intensities was significant for each marker. RFI and statistical tests were performed on all segmented cells for each marker except for SATB2 in Fig. 2C. Since SATB2 expression is limited to a subset of the segmented cells, only the top 20% brightest cells from both pre- and post-elution distributions were used for RFI calculation and statistical tests to ensure comparison of true SATB2 positive cells.

Table 8.

Settings in StarDist Fluorescence Cell Detection Script Used for Segmentation of Each Figure.

Settings	Fig. 2a	Fig. 2b	Fig. 2c	Fig. 3	Fig. 4	Fig. A3c	Fig. A3d	Fig. A4
NormalizePercentiles	30-99	15-99	20-99	15-99	25-99	15-99	20-99	25-99
threshold	0.7	0.2	0.6	0.2	0.7	0.7	0.7	0.7
pixelSize	1.1	0.3	0.4	0.3	0.4	0.2	0.2	0.2
cellExpansion	30	2	6	2	3	2.5	4	2

For calculation of signal-to-noise ratio (SNR), the top20bottom10 method was used. ⁵ In short, the top20bottom10 value is calculated by dividing the average of the top 20 brightest cells with the average intensity of the weakest 10% of segmented cells, as suggested by a publication from Akoya Biosciences, the provider of the PhenoCycler technology used in this study. ⁵

Image Coregistration

Image coregistration was done using, “The Image Combiner Warpy Extension v02.6” in QuPath v04.3. To start the coregistration process, “New project” was created in QuPath software, and raw files in OME-TIFF format were added to the project. Individual images were color adjusted, if needed, using the “Brightness and contrast” setting. A reference image was selected for coregistration and using the menu option, “Analyze → Interactive image combiner warpy,” the settings for coregistration were adjusted. Using “Choose images from the project” option in “Image and overlays” panel, other images to be coregistered were selected. Opacity could be adjusted using the slider to help with the manual alignment of images. To help the automatic registration function, it is important to provide coarse alignment by manually overlaying the image to be coregistered on top of the reference image. And, the auto-alignment was done using “Affine transform” as the registration type and “Image intensity” as the alignment type in the settings. Corrections for scale and rotation could be done at this stage of manual overlay. As a next step, using “Estimate transform,” the final transformation coordinates were calculated. At this stage, if the coarse alignment was satisfactory, the coordinates were copy pasted from one image to another to enable automatic overlay. The final step was to generate the coregistered image by clicking “Create” tool. Using, “File → Export images → OME-TIFF,” the coregistered image could be exported and saved. A detailed protocol to carry out image coregistration is given in the following GitHub resource page: https://github.com/BIIFSweden/CoregistrationMultiplexed/tree/main/coregistration_qupath.

Introduction

Multiplexed immunofluorescence (mIF) and immunohistochemistry has greatly contributed to more in depth understanding of the tissue architecture and the cellular and biological processes within. While the development in the field has been driven by immuno-oncology to explore immune cell presence and distribution within tumors, the ability to simultaneously target and visualize 10-folds of protein markers within the same sample is relevant for many research areas such as in neurobiology, developmental biology, infection and inflammation to mention a few.

Sequential immunofluorescence (seqIF) and imaging by barcoding are two central antibody-based approaches for visualization of multiple target proteins within the same tissue section using fluorescence as a readout. SeqIF involves a step-by-step process where one or a few primary antibodies are incubated and detected, followed by elution of the antibody in repetitive cycles to detect all target proteins in a sequential manner. The proteins can be detected using either fluorescently labeled primary antibodies with the benefit of allowing for several antibodies of the same species to be used simultaneously, or by fluorescently labeled secondary antibodies that bind to primary antibodies of a specific species (indirect immunofluorescence). As the detection is done sequentially and the fluorescence signal is removed within every cycle, the same fluorophores can be used to detect many targets, thereby overcoming the limitation of spectral overlap associated with using multiple fluorophores simultaneously.^6,7

One example of this strategy is the COMET platform from Lunaphore Technologies (Tolochenaz, Switzerland) that allows for fully automated staining, imaging, and elution, currently of up to 20 markers on four samples in parallel.^8–10

The other multiplexing approach, imaging by barcoding, involves attaching a short oligo sequence—or barcode—to an antibody to give it a unique identification tag that is used for detection of the specific antibody and the corresponding target protein. After antibody incubation on the tissue slide, these barcodes can either be cleaved by UV for downstream detection and identification or by adding complementary fluorescently labeled reporters that hybridize to the antibody barcode and allows for detection directly by the fluorescence microscope, for example, the PhenoCycler technology, formerly known as CODEX.^11–13 Using the PhenoCycler, up to three fluorescently labeled complementary reporters are added, imaged, and removed in an iterative manner until all antibodies have been detected. This is similar to the process of adding a few antibodies at a time in seqIF, while here all the antibodies are in place from the start, only the detection oligos are added and removed. A great advantage with barcoding is that all antibodies, regardless of species, can be incubated in a single staining step, ensuring proper tissue integrity and shortening the time of the protocol while allowing for a high multiplexing capacity. Several commercial platforms use the approach of conjugated antibodies—mostly for fluorescence detection with imaging^11,14,15 but also for a sequence-based readout ¹⁶ or using metal ions and mass spectrometers for detection of their specific mass spectra.^17,18 All of these platforms offer ready-to-use conjugated antibodies for a growing list of markers; however, many projects require custom conjugation to capture all essential markers for a given study.

While methods involving conjugated antibodies have several advantages in multiplexing, the antibody conjugation process presents challenges that have to be faced when including markers lacking commercially available conjugated antibodies for a particular platform or method. Conjugation can affect an antibody’s binding specificity and affinity, leading to reduced target recognition and potential cross-reactivity. Also, depending on protocol, the conjugation may alter antibody stability, causing aggregation, denaturation, or reduced shelf life.^19,20 Finally, depending on the detection chemistry and the local abundance of target protein, signals from conjugated antibodies may appear weak compared with indirect immunofluorescence where secondary antibodies carrying several fluorophores are used for detection.

For several years, our lab has contributed to the creation of the Human Protein Atlas Program,^21–23 mapping the subcellular localization of proteins using an optimized immunofluorescence protocol for proteome wide localization studies. ²⁴ A few years ago, we transitioned to multiplexed imaging, using the PhenoCycler platform based on oligo conjugated antibodies from Akoya Biosciences^11–13 and the COMET platform from Lunaphore,^8–10 allowing for automated sequential indirect immunofluorescence. Driven by the challenge of antibody conjugation and our experience from using the two complementary multiplexed approaches offered by COMET and PhenoCycler, we have combined one cycle of indirect immunofluorescence followed by elution and a panel of barcoded antibodies compatible with the PhenoCycler platform, to demonstrate the creation of a multiplexed image data set of a few custom protein markers on top of standard markers available for PhenoCycler. In summary, our workflow presents an alternative to conjugation of antibodies and allows for their inclusion into a PhenoCycler workflow.

Results

A Workflow for Integration of Indirect Immunofluorescence With Multiplexed Immunofluorescence Using PhenoCycler

Taking advantage of the use of native primary off the shelf antibodies that are used in indirect IF, the knowhow of antibody elution that is part of the COMET workflow for seq IF, and the high multiplexing capacity with the PhenoCycler platform using pre-conjugated antibodies, we developed a workflow combining the three modalities into an integrated workflow (Fig. 1). While the PhenoCycler technology works flawlessly for successfully conjugated antibodies, our experience is that several antibodies have failed after conjugation while working well in standard indirect immunofluorescence. This has resulted in many projects being delayed or progressing slowly, and for some projects we have failed to identify good antibody clones that fulfill the recommended criteria from the kit for successful conjugation. Many antibody providers offer antibodies in a formulation that fulfills the recommended criteria for successful conjugation using conjugation kits from Akoya Biosciences. However, the nature of the conjugation protocol itself, based on partial reduction of the cysteine bridges in the antibodies, might lead to disruption of the antibody binding capacity and epitope recognition. ¹³ In these cases, reagents for conjugation as well as the purchase of the antibody itself has been wasted, along with time and other resources. Our workflow for selecting antibodies for conjugation, validating them before conjugation with standard IF, followed by conjugation and validation of the post conjugated antibody is illustrated in Fig. A1A. As illustrated, a failed antibody in the pre-validation step using indirect immunofluorescence, or in staining post conjugation leads us to the first step of antibody clone selection. Two examples of image data before and after conjugation are shown in Fig. A1B and C (see antibody details in Tables 1 and 3). Figure A1B shows data for Eomes, a critical transcription factor frequently expressed in certain subsets of both T cells and NK cells. As such, it was requested as an essential marker in one of our user projects studying immune cells in the liver. However, as there were no available pre-conjugated antibodies for this target from Akoya Biosciences, in-house conjugation using the Akoya conjugation kit was needed. Staining of the primary antibody was done in FF tonsil as well as liver sections and while this antibody before conjugation stained nuclei in the majority of cells of a tonsil and liver section, no staining was observed after the conjugation process. The conjugation was also repeated with another Akoya barcode, giving the same result (data not shown). Another example of a marker requiring in-house conjugation is shown in prostate tissue (Fig. A1C). This protein, microseminoprotein beta (MSMB), staining is clearly detected as expected in glandular cells before conjugation, but only very weak staining is detected in tissue stained with the conjugated antibody. The absence of staining post conjugation can be due to weak signals from the reporter oligo used for detection of the conjugated antibody, which is in general lower than signals from fluorescently labeled secondary antibodies as used in the first regular IF step. It can also be the result of a partially fragmented antibody due to the conjugation process, leading to inability of the antibody to bind its epitope. This can be tested using the secondary antibody for detection of the conjugated antibody, to see whether staining of the target can still be achieved. In cases such conjugated antibodies give efficient and specific staining with a secondary antibody, it can still be used for indirect immunofluorescence, or qualify for our workflow presented in this work.

Figure 1.

Workflow for integration of indirect immunofluorescence with multiplexed immunofluorescence. Cells or tissue are preprocessed for immunofluorescence staining and stained with up to two antibodies. The sample is imaged followed by elution of the primary and secondary antibodies. The sample is then incubated with antibodies for a PhenoCycler run and iterative rounds of reporter addition, imaging and reporter elution are carried out on the PhenoCycler Fusion. The data from the PhenoCycler run is then coregistered with manual immunofluorescence image(s) to create a final qptiff which displays all markers in one collated image. Figure generated using Biorender.

For many of our projects requiring in-house conjugation, there has been one or two protein markers that are essential for the respective project, meaning it would be relatively easy to include them with a PhenoCycler panel using other detection strategies and thus circumvent the conjugation process. This further motivated the creation of the integrated workflow as described in Fig. 1A.

With this protocol, we perform standard antigen retrieval of the tissue sample (step 1) and make use of our standard protocol for indirect IF for cells, FF or FFPE tissue, respectively (step 2). Using the Phenoimager instrument, the proteins in indirect IF are detected, using the same or equivalent filters as for the detection of oligos in the downstream PhenoCycler experiment. In this protocol and setup, detection of up to three markers can be detected in this step (step 3). After image acquisition, the slide is removed from the instrument and primary and secondary antibodies are eluted with optimized protocols for the respective sample type (step 4). The sample is then prepared following our standard PhenoCycler protocol for FF and FFPE samples, respectively, starting with the antibody incubation (step 5) and then the image data are captured within a standard PhenoCycler run (step 6). Finally, images acquired from the indirect IF in step 3 are coregistered with the PhenoCycler image data set (step 7), allowing for downstream analysis of a merged image data set.

Evaluation of Two Elution Buffers for Cells, FF and FFPE Sections

To create a robust, standardized workflow for integration of manual indirect IF followed by multiplexed IF with PhenoCycler in different sample types, our first step was to test elution conditions on cell cultures, FF and FFPE sections (Fig. 1A, step 4). To allow for flexibility and alternatives, we used two different elution buffers: (1) an in-house prepared 4i elution buffer, initially published in the 4i multiplexing method by Gut et al. ⁷ and (2) the commercial Lunaphore elution buffer ⁸ that we already have in our lab and use for elution with the COMET platform. As the 4i protocol originally published by Gut et al. ⁷ primarily was used on cells, and the elution buffer from Lunaphore is intended to be used in the COMET instrument with very short incubation times, the respective elution protocols used here were slightly adapted in terms of temperature, time and number of incubations to ensure compatibility for tissue sections with the 4i buffer, and for off-COMET elution with the Lunaphore elution buffer. The exact elution protocols for the respective sample type are described in the material section.

In Fig. 2A, U2OS cells were stained with primary antibodies targeting the nuclear protein PDS5 cohesin-associated factor A (PDS5A), alpha tubulin targeting microtubules and calreticulin targeting the endoplasmic reticulum (Table 2). These markers were chosen as representing different cellular structures and all expressed in the U2OS cell line. Alfa tubulin and calreticulin further serves as fixative- and morphology controls and are used as reference markers within all IF staining s in the HPA workflow. ²⁴ The impact of the elution buffers on the cell morphology was evaluated, from comparing the DAPI staining before and after elution (Fig. A2A). By measuring the nuclear circularity of all cells before and after elution, where the value of 1 is a perfect circle and a lower value indicates less circularity, a lower value after elution would indicate a negative impact on the sample integrity. From these measurements of the U2OS cells, the two elution buffers yielded similar results as before elution and indicate very little or no impact on the nuclear morphology. As shown by high-resolution images (Fig. 2B), both 4i and Lunaphore elution buffers showed a near-complete removal of the stained markers PDS5A, alpha tubulin, and calreticulin. Quantification of the relative fluorescence intensity (RFI) of each marker, defined as the normalized mean of the cell or nuclear mean intensities for all segmented cells after elution as compared with the staining intensity before elution, revealed that both the elution conditions efficiently and significantly removed the staining of PDS5A, alpha tubulin, and calreticulin (Fig. A2B). While the signal of all markers was completely removed with the 4i buffer, a few percentage of intensity remained with the Lunaphore buffer in this experimental condition. However, this is very low and could likely be reduced further by extra elution washes if required.

To evaluate the two elution buffers in FF tissue, tonsil sections were used (Fig. 2B). Again, antibodies toward alpha tubulin and calreticulin were used as they are key components of the cytoskeleton and endoplasmic reticulum, respectively, and expected to be expressed also in the tonsil tissue sections. The sections were counterstained with DAPI and imaged on the Phenoimager. The antibody dilutions are shown in Table 3. Similarly as for cells, the staining of the respective marker was efficiently removed with both the elution buffers. Also the RFI was calculated for the markers after elution and demonstrated significant decrease in cell mean intensities for both elution protocols in thousands of cells (Fig. A2D). RFI values are higher for both markers compared with results from cell lines, which can be partially explained by the greater endogenous autofluorescence expected from tissue samples compared with cell lines, meaning fluorescence intensity after elution cannot be expected to be gone completely. Again, RFI values could likely be improved with further optimization of the elution times and cycles. However, as the aim of our experiments were to compare performance of the elution buffers and these markers are not of particular interest for any project, this was not done. However, it can be concluded that the Lunaphore protocol developed specifically for tissues performed slightly better compared with the 4i protocol for the markers tested. As for the U2OS cells, the tissue integrity of the FF sections after elution was evaluated by measuring the nuclear circularity of all segmented cells before and after elution. While the nuclei of every cell in a tissue is not expected to be a perfect circle due to its position in the section, a lower value after the elution treatment would indicate that the morphology of the nuclei has been compromised and that subsequent staining of other proteins in a PhenoCycler run or other staining could be less successful. However, nucleus circularity appears to be maintained after use of either 4i elution protocol or Lunaphore elution protocol for FF samples (Fig. A2C). Regarding the staining pattern of the two markers, it is worth commenting that the filaments of the microtubules and the network like structure of the endoplasmic reticulum that are clearly visualized in cell cultures (Fig. 2A), was not observed in the tissue sections but rather appear as a cytosolic staining pattern. This is still acceptable given the resolution and nature of the sample and as the main objective was to evaluate the elution efficiency.

For evaluation of the two elution buffers in FFPE sections, a TMA consisting of several different tissue types of normal and cancer origin was used. In Fig. 2C, the core of a rectum tissue stained for the proteins SATB homeobox 2 (SATB2) and cadherin using a pan-cadherin antibody were used to represent the results from the elution (antibody details in Table 1). The alfa tubulin and calreticulin markers as used for cells and FF sections did stain well in the FFPE samples with our standard antigen retrieval protocol used for IF and PhenoCycler. Other protein markers were therefore used instead. Similar as for cells and FF tissue sections, both elution buffers yielded efficient and significant elution of the stained protein markers as shown by the images in Fig. 2C and from quantitative measurements of the RFI after elution (Fig. A2F). As for FF samples, RFI values for the markers in FFPE samples are slightly higher than in U2OS cells. Similarly to FF samples this can partially be explained by higher endogenous autofluorescence in tissue samples. Furthermore, the staining intensity of the two markers were low with the image acquisition settings being used, meaning the impact of autofluorescence to the complete signal detected was high. However, the two elution buffers yielded very similar results.

Again, evaluating the effect of the elution protocols on the tissue integrity by measuring the nuclear circularity, showed preserved tissue integrity after elution also for FFPE tissue (Fig. A2E). While the rectum core was used to illustrate elution efficiency of FFPE samples in Fig. 2C, the TMA consisted of several FFPE tissue types (including kidney, pancreas, breast, brain, lymph node, prostate, liver, stomach, small intestine and colon). While the staining of SATB2 and pan-cadherin was not perfect in all cores due to differential expression of SATB2 and the epitope accessibility variation between tissue types, the protein markers were efficiently eluted across the tissue types (data not shown).

To summarize the results from the evaluation of the two elution buffers across cells, FF and FFPE sections, both elution protocols gave satisfactory results with regards to efficient and significant reduction of staining intensity. The impact of the elution buffers on the tissue integrity was neglectable for all sample types evaluated with both of the elution protocols. Therefore, depending on preference, either an in-house prepared elution buffer or a commercial ready to use buffer can be used to efficiently remove stainings from indirect IF in cells, FF or FFPE sections. Given these equally successful results, we from here on used the Lunaphore elution buffer protocol to evaluate compatibility with subsequent PhenoCycler runs.

Integration of Indirect Immunofluorescence With Multiplexed Immunofluorescence in FF Tissue Samples

Next, to evaluate the impact from elution on downstream multiplexing with the PhenoCycler platform, two FF tonsil sections were used. One sample was prepared according to standard PhenoCycler workflow, including standard tissue pre-processing as described in the “Methods” section. For comparison, another sample was preprocessed in the same way, following indirect IF and elution using the Lunaphore elution buffer and protocol as described in the “Methods” section. We used a few representative pre-conjugated primary antibodies from Akoya Biosciences, commonly used for multiplexed immunofluorescence with PhenoCycler in FF tissue samples (Table 4). An overview of the two tonsil samples with all PhenoCycler markers (n=8) are shown in Fig. 3A for the control sample and in Fig. 3B for the sample undergoing indirect IF and elution before the PhenoCycler run. It should be noted that the two sections of tissue used were from different tonsil samples, however the sample collection and pre-processing was the same and the age of the two sections were similar. No difference in performance of the two samples was observed for the overall PhenoCycler process. Both samples remained well attached on the glass slides and the PhenoCycler runs were successfully completed for both samples with signals for all markers. The selected markers for multiplexed immunofluorescence with PhenoCycler (Table 4), represents nuclear markers (Ki67), cytoplasmic markers (pan-cytokeratin or PanCK) and cell surface markers (CD3e, CD4, CD8, CD45, CD45RO, HLA-DR) and CD31, which was used to visualize vessels in the sections by staining endothelial cells. Side by side images for each of the 8 markers from the control and post elution PhenoCycler runs, are shown from representative regions in Fig. 3C. The selected regions are also displayed in the overview of the respective samples (Fig. 3A and B, respectively). The results showed similar staining patterns of all the markers in the control and post-Lunaphore elution FF tonsil samples and that the tissue integrity was well preserved also after the elution. As the tissue integrity of FF sections had already been evaluated in previous experiments after both 4i and the Lunaphore elution buffer, the nuclei circularity was not quantified in these samples. Furthermore, as this buffer is a commercial elution buffer intended for sequential IF in both FF and FFPE tissue sections, ¹⁰ the preservation of the tissue integrity is expected and thus also successful staining of all PhenoCycler antibodies. To obtain a more quantitative evaluation of the stainings obtained with PhenoCycler after the elution, the signal to noise ratios (SNR) for all markers were calculated and compared between the control sample (Fig. 3A) and post elution sample (Fig. 3B) using the top20bottom10 approach as published by Akoya Biosciences, the provider of the PhenoCycler technology. ⁵ All SNR values for the respective markers in the 2 samples are shown in Table 9. According to this SNR approach, markers with SNR values above 10 are considered acceptable as allowing for proper thresholding and downstream analysis. All markers in both the control and post elution sample displayed SNR values well above 10, with higher SNR values for some of the markers in the post elution sample. While the samples are not from consecutive sections of the same tissue block, exact comparison of signal intensities and SNR values cannot be done. However, the aim of the SNR calculations was to ensure that all markers in the control PhenoCycler run were performing equally well also in a sample undergoing the elution protocol before the PhenoCycler run. In this regard the results of all markers are satisfactory and demonstrate the feasibility of combining manual IF followed by elution and subsequent multiplexing with PhenoCycler.

Figure 3.

Comparison of a control and post-elution PhenoCycler run in FF tonsil tissue. (A) Overview of an eight-plex PhenoCycler run after indirect immunofluorescence staining and elution with Lunaphore elution buffer. Scale bar = 500 µm. White boxes indicate regions of interest (ROIs) shown in (C). (B) Overview of a control eight-plex PhenoCycler run on the same tissue type and using the same marker panel as in (A). Scale bar = 500 µm. White boxes indicate ROIs shown in (C). (C) Results of each marker in the eight-plex panel for the post-elution PhenoCycler run on the left and the control PhenoCycler run on the right. Scale bar = 20 µm. ROI1 from (A) is used for CD4 staining and ROI2 is used for all other markers for the post-elution run. ROI2 from (B) is used for pan-cytokeratin and ROI1 is used for all other markers for the control run.

Table 9.

Top20bottom10 Values for Markers in Fig. 3 (Lymphnode Samples).

Marker	Top20btm10 Control	Top20btm10 Post Lunaphore elution
Ki67	4570	inf
Pan-Cytokeratin	12,800	821
CD3e	16.5	318
CD31	1070	inf
CD45	58.1	18.1
CD4	89.5	366
CD45RO	355	15,200
HLA-DR	735	936

Successful Integration of Indirect Immunofluorescence With Multiplexed Immunofluorescence in Various FFPE Tissue Types

As a next step, we wanted to use FFPE tissues to show the adaptability of this protocol. Following the same procedure as for the tonsil FF sections, two FFPE sections from bladder cancer, from a control PhenoCycler run (Fig. 4A) and from a section undergoing elution before the PhenoCycler run (Fig. 4B) were compared. While the panel for the FFPE sample is small with only five markers (Table 5), they represent both nuclear and membrane markers, covering both broadly expressed markers (PCNA, beta catenin) as well as more cell specific markers (CD3, CD68, Ki67). As all antibodies in a PhenoCycler run are incubated in a single step, we believe smaller PhenoCycler panels to be representative for demonstrating the concept of a PhenoCycler run following elution of markers from indirect IF. The PhenoCycler runs of both sections completed successfully and selected representative regions of interest (ROI) from both samples are displayed in panel A and B and compared for each marker in Fig. 4C. All markers stained as expected in both samples and showed intact membranes as evaluated from the Beta-catenin and CD3 staining. The nuclear markers were also stained satisfactorily with stained nucleoplasm (PCNA) and the nucleoli (Ki67), although the Ki67 exposure was a bit high and thereby saturated the signal in some cells. The SNR values of both the control sample and post elution sample are shown in Table 10, and shows values well above the threshold for both control and post elution sample.

Figure 4.

Comparison of a control and post-elution PhenoCycler run in FFPE bladder cancer tissue. (A) Overview of a five-plex PhenoCycler run after indirect immunofluorescence staining and elution with Lunaphore elution buffer. Scale bar = 500 µm. (B) Overview of a control five-plex PhenoCycler run on the same tissue type and using the same marker panel as in (A). Scale bar = 500 µm. (C) Results of each marker in the five-plex panel for the post-elution PhenoCycler run on the left and the control PhenoCycler run on the right. Scale bar = 20 µm.

Table 10.

Top20bottom10 Values for Markers in Fig. 4 (Bladder Cancer Samples).

Marker	Top20btm10 Control	Top20btm10 Post Lunaphore Elution
PCNA	16.7	35.5
Ki67	88.9	inf
CD3e	170	3650
b-Catenin	30.3	1060
CD68	22.8	243

To further test the compatibility of indirect IF followed by PhenoCycler for FFPE samples, two sections of a TMA were used with the same experimental setup to evaluate staining between a control PhenoCycler run and one following elution in other tissue types. The markers used in this experiment are shown in Table 7. While some of the markers are not expected to be well represented across all tissue cores in the TMA (e.g., pan-cytokeratin and beta catenin), the results of two cores and a subset of markers are used to illustrate results in Fig. A3, with examples from lymph node and colorectal cancer. The overview of the cores from the control and post elution run are displayed in panels A and B, and the comparison of each marker between the samples in panels C and D. The immune cell markers (CD3, CD4, CD8, CD68) were performing well in both the lymph node and colorectal cancer core, although the latter is less immunogenic and displays a sparser distribution of these immune cells. The nuclear markers PCNA and Ki67 both stained specifically, however displayed higher background in the post elution sample for both cores. Yet, the SNR values for all markers were still satisfactory, except for Ki67 in the lymph node core which was just below 10 (Tables 11 and 12). Both nuclear markers have in general displayed lower SNR values also in the control samples across different tissue types. While SNR values of 10 or above are considered good, a value just below would still allow for proper downstream analysis according to the publication of the method. The background of these markers could potentially be reduced by lowering exposure time. Overall, all cores of the TMA were retained on the glass slides after the control and post elution PhenoCycler runs, and from inspection of all the cores, no negative impact on tissue integrity was observed. Taken together, from testing our protocol on various FFPE tissue types, these results suggest that manual IF followed by elution can be combined with downstream PhenoCycler runs without negative impact on the PhenoCycler results.

Table 11.

Top20bottom10 Values for Markers in Fig. A3 (Colorectal Cancer Core).

Marker	Top20btm10 Control	Top20btm10 Post Lunaphore Elution
CD3e	267	177
Ki67	1395	19.2
CD68	36.8	1770
CD4	10.2	51.4
PCNA	12.5	13.9
CD8	17,260	115

Table 12.

Top20bottom10 Values for Markers in Fig. A3 (Lymphnode Core).

Marker	Top20btm10 Control	Top20btm10 Post Lunaphore Elution
CD3e	63.7	38.9
Ki67	5900	9.69
CD68	46.4	11,100
CD4	21.9	44.1
PCNA	18.5	10.2
CD8	2860	119

Image Coregistration of Indirect IF and Subsequent PhenoCycler Runs

After successfully demonstrating the compatibility of combining one cycle of indirect IF with multiplexed IF on the PhenoCycler platform, the images of the two imaging experiments have to be coregistered to allow for downstream analysis. The workflow for using a pipeline called image coregistration in QuPath 4.3, and “The Image Combiner Warpy extension v0.2.6q” tool, is summarized in (Fig. 5A). First, individual data sets obtained from the same tissue sample with indirect IF and multiplexed immunostaining were analyzed for any lack of compatibility. This includes checking for non-overlapping colors used for representing different markers, orientation and morphology of the tissue section, and image adjustments like brightness and contrast. It is also important to keep the data format as OME-TIFF for uniformity. Using the “Run Warpy Extension” command, the data sets were imported into a new project and the image coregistration protocol was initiated. Next, manual adjustments of the alignment of the images like scaling and rotation were done. After this, the coordinates of the alignment were fixed and an automatic registration step was executed as the final step. The output was exported as the coregistered image, combining the data sets. A detailed protocol to carry out image coregistration is given in the following GitHub resource page: https://github.com/BIIFSweden/CoregistrationMultiplexed/tree/main/coregistration_qupath.

Figure 5.

Image coregistration of indirect immunofluorescence and PhenoCycler data. (A) Workflow of image coregistration of the images from indirect immunofluorescence and the multiplexed PhenoCycler data set. (B) Selected regions of FFPE bladder cancer sample stained with indirect IF for HER2 in yellow (left) followed by multiplexed immunofluorescence in PhenoCycler with 5 different markers including DAPI (middle) and the final coregistered image following the workflow in panel (A) (right). All images are counterstained with DAPI (blue). Scale bar = 50 µm. (C) Selected regions of FFPE prostate cancer sample stained with indirect IF for MSMB in white (left) followed by multiplexed immunofluorescence in PhenoCycler with 6 different markers including DAPI (middle) and the final coregistered image (right) following the workflow in panel (A). All images are counterstained with DAPI (blue). Scale bar = 50 µm.

Figure 5B shows the individual data sets from the FFPE bladder cancer section (as previously presented in Fig. 4) representing indirect IF with HER2 staining (Table 1) and multiplexed immunostaining with a five-plex panel including Ki67, CD68, PCNA, Beta-Catenin (BCAT) and CD3e markers (Table 5), as well as the merged image after coregistration. The HER2 oncoprotein was selected as a suitable test marker due to its known expression pattern in bladder cancer and relevance for several malignancies. ²⁵

In Fig. 5C, a second example for FFPE prostate tissue is presented, showing MSMB staining by indirect IF sample, followed by elution and a PhenoCycler run with a five-plex panel according to Table 6. As presented in Fig. A1C, the MSMB protein marker was first unsuccessfully conjugated with the Akoya conjugation kit and therefore unable to be included in the PhenoCycler panel. However, as the primary MSMB antibody stained as expected during validation with Indirect IF, it was later used in an indirect IF, followed by elution with the Lunaphore buffer as described in the method section, and then combined with the markers from the PhenoCycler run. A comparison of each protein marker in the PhenoCycler runs for a control and after elution sample were done also for the prostate samples and presented in Fig. A4 and with correlating SNR values presented in Table 13. This example concretely demonstrates how a well-performing primary antibody can be combined with a PhenoCycler panel in cases where conjugation fails. As indirect IF is a standard procedure for antibody validation before any in-house conjugation, we recommend this same tissue section to be directly used to test the elution efficiency of the marker. If elution is successful, the workflow of combining indirect IF with PhenoCycler is applicable and an option in case conjugation fails. Optionally, the same slide can proceed directly to the subsequent PhenoCycler run to save time and reagents.

Table 6.

Akoya Pre-Conjugated Antibodies as Used in the Control and Post Elution PhenoCycler Runs for FFPE Prostate Tissue Samples.

Cycle in PC Run	Catalog No.	Protein	Subcellular Location	Dilution	Fluorophore	Exposure Time for Control Run (ms)	Exposure Time Post Elution (ms)
C1	4450018	CD4	PM	1:200	Alexa Fluor 647-RX003 for PhenoCycler	250	350
C2	4150017	CD31	PM	1:200	Alexa Fluor 750-RX001 for PhenoCycler	300	300
C3	4450031	Keratin 14	PM	1:800	Atto550-RX002 for PhenoCycler	50	50
C3	4550113	CD68	Cytosol	1:400	Alexa Fluor 647-RX015 for PhenoCycler	150	150
C5	4550082	Keratin 8/18	PM	1:800	Alexa Fluor 647-RX081 for PhenoCycler	50	50

AF = Alexa Fluor.

Table 13.

Top20bottom10 Values for Markers in Fig. A5 (Prostate).

Marker	Top20btm10 Control	Top20btm10 Post Lunaphore Elution
CD8	inf	inf
CD31	inf	inf
Keratin 14	3230000	inf
CD68	inf	inf
Keratin 8/18	2220	651

Discussion

In this study, we demonstrate the concept of combining indirect IF with the highly multiplexing capacity from oligo conjugated antibodies, here using the PhenoCycler platform. The protocol we present is intended to make multiplexed imaging more flexible and allow for the inclusion of key protein targets that are central for a specific project. It is not intended to increase the number of total protein markers in the experiment, although this is an indirect effect. We believe our contribution to making multiplexed imaging more flexible, lies in the optimization of two efficient elution protocols, and the demonstration of the overall compatibility with the PhenoCycler platform for both FF and FFPE sections.

There is however no guarantee the exact protocol presented in this study will work for all proteins and samples. A successful and adequate staining is context dependent, relying on the target protein concentration and epitope accessibility, to allow for the best binding efficiency while limiting off target effects. Changing incubation time, fixation conditions, and antigen retrieval in case of FFPE can make a big difference in antibody performance. Therefore, some optimization for the single or low plex IF upstream of a PhenoCycler run has to be expected.^26,27 However, as little impact was observed from the different elution protocols on the variety of samples tested, we expect these sample types to generally be straightforward to work with using this integrated protocol of indirect IF and PhenoCycler.

An important parameter for successful combination of indirect IF and multiplexing is the tissue sample quality at the start of the experiment but also after rounds of elution. ²⁸ Our general experience is that the FF tissue sections are more fragile and therefore can be more challenging to use in multiplexed imaging that involves elution steps. The measurement of the nuclear distortion and inspection of membrane marker staining will help evaluate the sample quality and potential impact of the elution on the tissue integrity. Based on the results in this study, the elution protocols had little impact on the sample integrity and our experience is that the sample collection and pre-preparation is more crucial than the impact from the elution protocol itself. For tissue samples showing poor quality already from start, impact from elution is likely to be more disruptive.

The commercial elution buffer from Lunaphore used throughout most of the experiments in this work normally allows for sequential IF using multiple rounds of elution on the COMET platform. ⁸ Therefore, in theory, several rounds of indirect IF could be performed upstream of PhenoCycler, to increase the number of essential markers needed for the study. However, every single IF also requires optimizations and the total amount of work increases with the number of cycles to be added. Therefore, for studies requiring a larger number of non-conjugated antibodies to target essential markers, different options should be considered, ²⁹ taking into account the previous experience from the markers as well as access to different instrumentation and automated solutions.

One possible solution to increase the number of antibodies upstream of PhenoCycler beyond 3 as demonstrated in this study, would be to use primary antibodies directly labeled with different fluorophores. This would allow for more antibodies to be used in parallel independently of species and make selection of antibodies easier. Many directly labeled antibodies are available for flow cytometry or other imaging platforms such as MACSima. ³⁰ Our approach of staining and imaging followed by elution upstream of PhenoCycler, should work for eluting such primary labeled antibodies as well, making them compatible with PhenoCycler runs. Depending on the acquisition options for the instrument the number of antibodies used in this first step could be 3 (using the same filters as for PhenoCycler) or up to six if the instrument allows for broader filter selection and spectral unmixing. In the case of the PhenoCycler Fusion instrument as used in this study, primary conjugates with fluorophores matching the filters used for Opal dyes could be used and allows for up to six antibodies to be detected and spectrally unmixed using in Form. However, this has not been demonstrated in practice in this study as no direct need for this has been identified and would still require some work and optimization.

As a national facility providing services within highly multiplexed imaging, we strive toward broad adaptation and to offer more flexible solutions that can further expand the use of these powerful multiplexing platforms. We hope our overall workflow, the presented elution protocols and image coregistration will be applied as a starting point for others who wish to combine their own key protein markers with their available multiplexing platform.

Limitations of the Study

The study brings together different imaging approaches for generating spatial proteomics data using antibodies. As such, overall awareness and careful evaluation of each and every antibody is important to receive accurate results. One of the underlying motivations for this study is the challenge of antibody conjugations. We do not have experience from using other conjugation methods, only the kits recommended and provided by Akoya for custom conjugation of antibodies intended for the PhenoCycler platform. As such, we want to emphasize that conjugation with other kits and with other reagents might have different outcomes. In this study, we rather focus on generating an alternative solution to conjugation, which allows hard-to-conjugate markers to still be studied using multiplexed imaging platforms. Also, here we do not present solutions or demonstrate our approach on any other multiplexed imaging platform but PhenoCycler. Yet, we believe our protocols and approach of performing conventional immunofluorescence followed by elution, can be applied upstream of other multiplexing approaches with antibodies and fluorescence as a readout. From a technical perspective, we did not demonstrate different options for the regular immunofluorescence step, as we only tested this using indirect IF and not with primary labeled antibodies. We believe however that the protocol we present would work equally well for this, and allows for easier combination of markers as antibodies can be species agnostic. To our knowledge, the exact temperature and buffer incubation time are not crucial for successful elution of most protein markers. Furthermore, we demonstrate the efficient elution with both a commercial and in-house prepared buffer on a variety of tissue types. While some protein markers and tissue types could need further optimization for best results, our protocol serves as a good starting point.

Appendix

Reagents or resource	Source	Identifier
Primary Antibodies
See Tables 1 to 7, for all antibodies used in this study	This paper	Tables 1 to 7
Secondary Antibodies
Alexa Fluor and Alexa Fluor Plus conjugated anti-mouse, anti-rabbit and anti-chicken	Thermo Scientific, Massachusetts, USA	A-11008, A-31571, A-31570, A-21449, A-31573 and A-31572
Biological samples
FFPE bladder cancer tissue samples	Provided by Dr. David Krantz (Karolinska Institute Pathology department, Stockholm, Sweden)	david.krantz@ki.se
FFPE prostate samples	Provided by Dr. Alistair Lamb (University of Oxford, Oxford, UK)	alastair.lamb@nds.ox.ac.uk
TMA FFPE samples	Provided by Atlas Antibodies (Stockholm, Sweden). Contact Dr. Carolyn Marks.	carolyn.marks@atlasantibodies.com
Fresh frozen normal tonsil samples	Provided by MD, PhD Mattias Jangard (Sophiahemmet, Stockholm, Sweden)	Mattias.jangard@sophiahemmet.se
Experimental models: Cell lines
U2OS, Osteosarcoma cells	ATCC	HTB-96, https://www.atcc.org/products/htb-96
Chemicals, peptides and recombinant proteins
0.1% poly-l-lysine	Merck	P8920
l-glutamine	Merck	G7513
0.25% Trypsin-1mM EDTA	Merck	SM-2003
l-glycine	Merck	50046
8M Guanidine hydrochloride	Merck	50937
Tris(2-carboxyethyl)phosphine hydrochloride powder (TCEP)	Merck	C4706
8M Urea	Merck	51457
37% Hydrochloric acid	Merck	258148
Elution buffer I and II	Lunaphore Technologies, Tolochenaz, Switzerland	BU07 and BU07L
1× Tris-EDTA (10mM Tris and 1 mM EDTA) pH 9.0 antigen retrieval buffer	VWR, Pennsylvania, USA	10-0037
Phosphate buffered saline (PBS) tablets	Thermo Scientific	003002
16% paraformaldehyde (PFA) solution	Thermo Scientific	28906
10X CODEX buffer	Akoya Biosciences, Massachusetts, USA	SKU 7000001
TritonX-100	Merck	X100
Methanol	Merck	34860-1L-R
10X buffer additive	Akoya Biosciences	SKU 7000019
DMSO	Merck	472301
4′,6-diamidino-2-phenylindole (DAPI)	Merck	D9542
Fluoromount-G	Thermo Scientific	00-4958-02
Nuclear counterstain solution	Akoya Biosciences	SKU 7000003
Assay reagent	Akoya Biosciences	SKU 7000002
Antibody staining kit in PhenoCycler Fusion protocol	Akoya Biosciences	SKU 7000017
Antibody conjugation kit in PhenoCycler Fusion protocol	Akoya Biosciences	SKU 7000009
PhenoCycler reporter oligos with Alexa Fluor -488, -647 and -750 conjugates and Atto -550 conjugate	Akoya Biosciences	https://www.akoyabio.com/phenocycler/assays/
Acetone	Merck	650501-1L
1× hydration buffer	Akoya Biosciences	SKU 7000017
Fetal bovine serum (FBS)	Merck	F2442
Histochoice clearing solution	Merck	H2779
Hydrogen peroxide	Merck	H1009
Sodium hydroxide	Merck	S2770
Software and algorithms
QuPath version 0.4.3		https://qupath.github.io/
ImageJ		https://fiji.sc/
GraphPad Prism version 9.5		https://www.graphpad.com/features
Adobe Photoshop 25.1.0		https://www.adobe.com/products/photoshop.html

Lead contact:

Further information and requests for resources and methods should be directed to and will be fulfilled by the corresponding author, Charlotte Stadler (charlotte.stadler@scilifelab.se).

Figure A1.

Workflow of antibody conjugations and examples of unsuccessful antibody conjugations. (A) Flowchart of the overall process of antibody conjugation with barcodes for the PhenoCycler platform. Antibody criteria apply to the commercial conjugation kit for PhenoCycler. (B) Upper panel shows indirect immunofluorescence staining of the Eomes transcription factor (red) detected by Alexa 555 secondary antibody in an FF tonsil and FF liver section. Lower panel shows the same Eomes antibody after conjugation to a PhenoCycler barcode, detected by the complementary PhenoCycler reporter. (C) Upper panel shows indirect immunofluorescence staining of MSMB (red) detected by Alexa 647 secondary antibody in an FFPE prostate section. Lower panel shows the same MSMB antibody after conjugation to a PhenoCycler barcode, detected by the complementary PhenoCycler reporter. Scale bar = 20 µm.

Figure A2.

Nuclear distortion and relative fluorescence intensity measurements. (A) Nuclear distortion before and after elution with 4i (top) respective Lunaphore (bottom) protocol in U2OS cells. (B) Table for U2OS cell measurements displaying total cell count after segmentation for each condition, RFI values for (PDS5A, alpha tubulin, and calreticulin), statistical significance (*** indicates p<0.001) for both Lunaphore and 4i protocol. (C) Nuclear distortion before and after elution with 4i (top) respective Lunaphore (bottom) protocol in FF tissue. (D) Table for FF tissue measurements displaying total cell count after segmentation for each condition, RFI values for (Alpha Tubulin and Calreticulin), statistical significance (*** indicates p<0.001) for both Lunaphore and 4i protocol. (E) Nuclear distortion before and after elution with 4i (top) respective Lunaphore (bottom) protocol in FFPE tissue. (F) Table for FFPE tissue measurements displaying total cell count after segmentation for each condition, RFI values for (SATB2 and Pan-cadherin), statistical significance (*** indicates p<0.001) for both Lunaphore and 4i protocol.

Figure A3.

Comparison of a control and post-elution PhenoCycler run in two FFPE TMA cores. (A-B) Overview of a six-plex PhenoCycler run after indirect immunofluorescence staining and elution with Lunaphore elution buffer on the left. Overview of a control six-plex PhenoCycler run on the same tissue type and using the same marker panel on the right. (A) Illustrates a lymph node core and (B) illustrates a colorectal cancer core. Scale bar = 200 µm. (C–D) Results of each marker in the six-plex panel for the post-elution PhenoCycler run on the left and the control PhenoCycler run on the right. (C) Illustrates a lymph node core and (D) illustrates a colorectal cancer core. Scale bar = 50 µm.

Figure A4.

Comparison of a control and post-elution PhenoCycler run in FFPE prostate tissue. (A) Overview of a five-plex PhenoCycler run on FFPE prostate after indirect immunofluorescence staining and elution with Lunaphore elution buffer. Scale bar = 500 µm. White boxes indicate ROIs shown in (C and D). (B) Overview of a control five-plex PhenoCycler run on FFPE prostate cancer and using the same marker panel as in (a). Scale bar = 500 µm. White boxes indicate ROIs shown in (C and D). (C and D) Results of each marker in the five-plex panel for the post-elution PhenoCycler run on the left and the control PhenoCycler run on the right. Scale bar = 20 µm. ROI1 from (A) is used for Keratin 14 and Keratin 8/18 staining, ROI2 is used for CD4 and CD68, and ROI3 is used for CD31 for the post-elution run. ROI1 from (B) is used for Keratin 14 and Keratin 8/18 staining and ROI2 is used for all other markers for the control run.