This protocol describes in detail how immune cell characterization of the tumor microenvironment using multiplex immunohistochemistry is carried out.
The immune cell landscape of the tumor microenvironment potentially contains information for the discovery of prognostic and predictive biomarkers. Multiplex immunohistochemistry is a valuable tool to visualize and identify different types of immune cells in tumor tissues while retaining its spatial information. Here we provide detailed protocols to analyze lymphocyte, myeloid, and dendritic cell populations in tissue sections. Starting from cutting formalin-fixed paraffin-embedded sections, automatic multiplex staining procedures on an automated platform, scanning of the slides on a multispectral imaging microscope, to the analysis of images using an in-house-developed machine learning algorithm ImmuNet. These protocols can be applied to a variety of tumor specimens by simply switching tumor markers to analyze immune cells in different compartments of the sample (tumor versus invasive margin) and apply nearest-neighbor analysis. This analysis is not limited to tumor samples but can also be applied to other (non-)pathogenic tissues. Improvements to the equipment and workflow over the past few years have significantly shortened throughput times, which facilitates the future application of this procedure in the diagnostic setting.
Immune cells play a crucial role in the protection against pathogens such as viruses and bacteria, but also against cancerous cells1. Therefore, the immune system within the tumor microenvironment (TME) holds a lot of promise for discovering prognostic and predictive biomarkers2. Immune cell infiltrates have been correlated to prognosis in various types of cancer, although this has not been implemented in clinical care yet3,4. In most tumor types, high numbers of cytotoxic T cells and T helper 1 cells and/or low numbers of regulatory T cells are linked to good prognoses. Efforts are ongoing to incorporate a so-called "Immunoscore" into the TNM staging of colorectal cancer, turning it into TNM-I staging5,6. The Immunoscore is derived from the total number of T cells (detected with CD3) and cytotoxic T cells (detected with CD8) in two different tumor regions: the tumor core versus the invasive margin (IM) of tumors. The Immunoscore has also been proposed to be of prognostic value in other cancer types, such as melanoma, lung cancer, and breast cancer6,7,8,9. Furthermore, immune cell infiltrates may also correlate to response to checkpoint blockade immunotherapy10. However, these predictive biomarkers must be validated in prospective studies before they can be routinely implemented in clinical practice. Moreover, it has also been proposed that a single biomarker will be insufficient for meaningful prediction11. Therefore, creating a complete map of a patient sample by combining different biomarkers has been proposed as a more comprehensive predictive biomarker in a so-called "cancer immunogram"12.
Among the methods for studying immune cells within the TME, the oldest and most well-known technique is immunohistochemistry (IHC), routinely used for diagnostic testing in several diseases, especially cancer13. This technique was limited to the use of one or only a few markers14 for a long time and therefore, was outcompeted in research settings by other techniques such as flow cytometry and gene expression profiling (GEP). However, the formalin-fixed and paraffin-embedded (FFPE) tumor tissues typically used in routine diagnostics and research are not (optimally) suitable for flow cytometry and GEP. Furthermore, although GEP and flow cytometry provide a lot of insight into cell phenotype and function, the lack of spatial information is a major disadvantage. Therefore, heterogeneity within a sample, such as differences in immune cell-infiltrated versus immune cell-excluded areas of a tumor, could go undetected15. Novel platforms have been developed for multiplex analysis of FFPE tissues, such as multiplex IHC, imaging mass cytometry, and CO-Detection by indEXing (CODEX) that can be used to detect multiple markers simultaneously within a tissue section16. Immune cells in the TME are being widely studied to find the best biomarkers for immunotherapy. However, multiplex techniques and automated image analysis pose hurdles of their own.
Our laboratory has extensive experience in multiplex IHC staining using the Opal/Tyramide signal amplification (TSA) method and has automated this on a IHC platform (see the Table of Materials)17,18,19,20,21,22,23,24,25,26,27,28,29,30,31. We have optimized immune cell panels for the detection of different subsets of lymphocytes, myeloid cells, and dendritic cells (DCs). Tissues that contain dense immune cell areas – for lymphocytes or complex cell morphologies (i.e., myeloid cells and DCs) – are particularly challenging to analyze, with a risk of over- or underestimating the number of immune cells present. To overcome this problem, ImmuNet analysis software was developed by our group32, and this machine-learning pipeline improved the quality of the detection of these different types of immune cells immensely. A detailed protocol from obtaining the FFPE material to the analysis of immune cell densities in different tissue compartments and distances between immune cell types is described here.
This protocol outlines how the multiplex IHC panels are performed at the Radboud University Medical Center since the implementation of the digital pathology imager in 2022. The described multiplex IHC panels can be used for different carcinomas (e.g., lung, prostate, colorectal, bladder, breast) with the use of a pan-cytokeratin antibody as a tumor marker or for melanoma with the use of melanocyte-associated antibodies as tumor markers. These multiplex IHC protocols have been carefully optimized in terms of primary antibody concentration, fluorophore combinations, and the sequence of the staining procedure. We and others have described multiplex IHC panel optimization earlier17,33,34,35. Multiplex IHC panels can be adapted, but the described analysis pipelines need to be evaluated and potentially adjusted or retrained accordingly. The described seven-color multiplex IHC protocols make use of the Opal fluorophores Opal480, Opal520, Opal570, Opal620, Opal690, Opal780, and 4',6-diamidino-2-phenylindole (DAPI), so that easy unmixing and fast scanning on the imager is enabled with "Multispectral One Touch ImmunoFluorescence" (MOTiF). Nine-color staining and scanning is not described in this protocol as this requires even more finetuning of the experimental setup and another mode of scanning on the imager that uses the liquid crystal tunable filter.
Spatial analysis of the TME is a sought-after technique to learn more about the immune cell compartment and discover new prognostic and predictive biomarkers, particularly in the field of immuno-oncology16. Many different techniques are being developed for this purpose, involving the detection of proteins, mRNA transcripts, or a combination of the two, with estimations up to 100-1,000 targets. However, higher multiplexing comes at the cost of less high-throughput experiments, higher experimental costs, and technical challenges, and often, only a small part of the TME can be analyzed. Multiplex IHC using the TSA-based method that we describe here, detects six different markers + DAPI simultaneously, is relatively less expensive to perform, and whole tissue sections are imaged in under 20 min, ready to be analyzed fully. This technique has become less complex with the automation of the staining procedure. Improvements in the multispectral microscope, which include the addition of two extra filters, have improved spectral unmixing and scanning times tremendously. It is possible to detect up to eight different markers + DAPI simultaneously. However, by expanding the multiplexing with more markers, the aforementioned benefits disappear as spectral unmixing becomes more challenging and scanning times for whole slides increase substantially. Efforts are being undertaken to standardize multiplex IHC between different institutions to facilitate implementation in the diagnostic setting more easily. For this standardization of multiplex IHC, we advise users to adhere to the more accessible protocol with six different markers + DAPI. Nevertheless, still quite some technical know-how is necessary and downstream analysis can be challenging, for which we have developed methodologies that are described in this protocol.
Standardization begins with multiplex IHC panel development. The importance of the choice of primary antibodies detecting particular protein targets has been emphasized before17. Our multiplex IHC panels are mostly developed with primary antibody clones that are also used and validated for IHC at our diagnostics department. However, in the case of the dendritic cell multiplex IHC panel, most antibodies were not used in the diagnostic setting (van der Hoorn et al., manuscript in submission). To ensure specificity and minimize batch differences, we chose to use monoclonal antibodies over polyclonal antibodies and have also validated most antibodies using transfected cell lines and primary cells. Over the years, different versions of multiplex IHC panels have been used in numerous studies using the Vectra 3 system18,21,23,24,25,26,27,28,29,30,31,32. To implement these multiplex IHC panels optimally on the PhenoImager HT system, some adjustments had to be made in primary antibody and fluorophore combinations. To benefit from better spectral unmixing and faster scanning times of whole tissue sections, implementation of the latest Opal480 and Opal780 fluorophores and avoiding the use of Opal540 and Opal650 fluorophores in seven-color multiplex IHC panels is necessary. Scanning times are ~3-10 times faster depending on the size of the tissue section. Multiplex IHC panel adjustments were quite easy to achieve, but some considerations need to be kept in mind. The fluorescent spectrum of Opal480 overlaps a lot with the autofluorescence spectrum and therefore, interferes with the spectral unmixing of erythrocytes and other autofluorescent structures. Using an increased concentration of the primary antibody paired with Opal480 solved this issue in most cases. The implementation of the proprietary Sample AF filter on the PhenoImager HT facilitates the unmixing of Opal480 and autofluorescence. However, it is best to use a primary antibody that yields a clear signal when used with Opal480 so that its signal is higher than the autofluorescence.
Even though these multiplex IHC panels are established, batch-to-batch variation is something that needs to be considered. By performing monoplex IHC controls before starting the full multiplex IHC experiment, we sometimes observed that primary antibodies perform either stronger or weaker from experiment to experiment. The reasons for this could be pipetting errors, suboptimal reagent storage conditions, and shelf life. We solved this by adjusting the primary antibody solution based on our experience. Even when none of the aforementioned adjustments had to be made, with every multiplex IHC batch experiment, it is important to set exposure times based on monoplex IHC-stained control slides.
Because our research was initially focused on different types of carcinomas and melanoma, multiplex IHC panels were required to be interchangeable between tumor types with minimal adjustments. Therefore, we always included multiple (tumor) tissue types in the optimization process and observed that dilutions for primary antibodies for immune cell markers can be kept similar between different tumor types. However, tumor tissue detection between carcinomas and melanoma needs different tumor markers. Accordingly, the tumor marker was always optimized to work at the end of each multiplex IHC panel and is currently always used in conjunction with Opal780, which coincidentally also has to be at the last fluorophore in a multiplex IHC staining procedure. By using the tumor marker consequently at the end of the multiplex IHC, these multiplex IHC panels can be easily exchanged for other tumor types, such as glioblastoma (i.e., GFAP) and Hodgkin lymphoma (i.e., CD30). For angiosarcoma, we used this lymphocyte multiplex IHC panel with erythroblast transformation-specific-related gene (ERG) as the tumor marker with only two optimization experiments25. The optimization included titration of the ERG primary antibody and testing the multiplex IHC panel with ERG at the end.
Other adjustments to these multiplex IHC panels can also be made by exchanging a certain immune cell marker for another immune or functional marker. Every change requires optimization. The protocol for optimization could be followed as described previously17. Certain changes to the proposed multiplex IHC panels will interfere with the ImmuNet algorithms that we have created. Sufficient data must be generated and time has to be spent to implement these changes into the algorithm (at least 750 annotations for every new marker and/or cell phenotypes, and 150 annotations for validation of previously trained markers). The panels presented here do not contain functional markers, although the implementation of immune checkpoint markers such as PD-1 and PD-L1 into multiplex IHC panels is performed in our laboratory. However, the analysis of markers that are less binary in negative and positive signals has proven to be more difficult and is an area of active research in our group.
The number of markers that can be simultaneously assessed with multiplex IHC is limited compared to other novel techniques. While this can be circumvented by analyzing different panels on consecutive slices of an FFPE block, it will be hard to compare these slices spatially. Orientation and folded artifacts are likely not the same after slide preparation. Nevertheless, multiplex IHC is quite accessible, which makes it an attractive tool for more institutions and researchers and therefore, more suitable for future implementation in a diagnostic setting. With the standardization of multiplex IHC immune cell panels for multiple tumor types and downstream analysis pipelines, more knowledge could be gained about differences in TME between patients and tumor types. This can, for instance, lead to more insights into the role of the TME in antitumor response to specific treatments. This may even give rise to new biomarkers to predict factors such as response to treatment and expected survival. Overall, this can enable multiplex IHC to become a clinical tool to aid with clinical decision-making, in a personalized medicine approach. Admittedly, more steps of the analysis procedure should probably be automated and standardized for it to be feasible for use in a daily diagnostic setting, so as of yet, it is mostly a futuristic perspective.
Analysis of multiple markers on a single sample slide can be a very powerful tool in spite of its technical challenges. With standardized experimental protocols and a robust analysis method, as we described here using ImmuNet, the quantification of multiple markers makes it more informative than classical IHC, while multiplex IHC remains relatively high-throughput compared to novel higher plex experimental methods.
The authors have nothing to disclose.
The PhenoImager HT was purchased through funding provided by the Radboud University Medical Center and Radboud Technology Center for Microscopy. CF is financially supported by a Dutch Cancer Society grant (10673) and ERC Adv grant ARTimmune (834618). JT is financially supported by an NWO Vidi grant (VI.Vidi.192.084). The authors would like to thank Eric van Dinther and Ankur Ankan for their assistance in creating workflows to store multiplex IHC data and Bengt Phung is thanked for instructions on how to implement multiplex IHC data in QuPath for ROI drawing.
anti-CD14 | Cell Marque | 114R-16 | section 3, clone EPR3653 |
anti-CD163 | Cell Marque | 163M-15 | section 3, clone MRQ-26 |
anti-CD19 | Abcam | ab134114 | section 3, clone EPR5906 |
anti-CD1c (BDCA1) | Thermo Scientific | TA505411 | section 3, clone OTI2F4 |
anti-CD20 | Thermo Scientific | MS-340-S | section 3, clone L26 |
anti-CD3 | Thermo Scientific | RM-9107 | section 3, clone sp7 |
anti-CD303/BDCA2 | Dendritics via Enzo Lifesciences/Axxora | DDX0043 | section 3, clone 124B3.13 |
anti-CD56 | Cell Marque | 156R-94 | section 3, clone MRQ-42 |
anti-CD66b | BD Biosciences | 555723 | section 3, clone G10F5 |
anti-CD68 | Dako Agilent | M087601 | section 3, clone PG-M1 |
anti-CD8 | Dako Agilent | M7103 | section 3, clone C8/144B |
anti-Foxp3 | Thermo Scientific | 14-4777 | section 3, clone 236A/E7 |
anti-Gp100 | Dako Agilent | M063401 | section 3, clone HMB45 |
anti-HLA-DR, DP, DQ | Santa Cruz | sc-53302 | section 3, clone CR3/43 |
anti-MART-1 | Thermo Scientific | MS-799 | section 3, clone A103 |
anti-pan cytokeratin | Abcam | ab86734 | section 3, clone AE1/AE3 + 5D3 |
anti-SOX10 | Sigma Aldrich | 383R | section 3, clone EP268 |
anti-Tyrosinase | Sanbio | MONX10591 | section 3, clone T311 |
anti-XCR1 | Cell Signaling Technologies via Bioké | 44665S | section 3, clone D2F8T |
antibody diluent | Akoya BioSciences | SKU ARD1001EA | section 3, from Opal 7-Color Automation IHC Kit 50 slide (can optionally also be replaced by TBST with 10% BSA) |
Bond Aspirating Probe | Leica Biosciences | S21.0605 | section 3 |
Bond Aspirating Probe Cleaning | Leica Biosciences | CS9100 | section 3 |
Bond Dewax Solution | Leica Biosciences | AR9222 | section 3 |
Bond Objectglas label + print lint | Leica Biosciences | S21.4564.A | section 3 |
Bond Research Detection System 2 | Leica Biosciences | DS9777 | section 3 |
Bond RX autostainer | Leica Biosciences | – | section 3, automated platform |
Bond TM Epitope Retrieval 1 – 1 L | Leica Biosciences | AR9961 | section 3 |
Bond TM Epitope Retrieval 2 – 1 L | Leica Biosciences | AR9640 | section 3 |
Bond TM Wash Solution 10x – 1 L | Leica Biosciences | AR9590 | section 3 |
BOND Universal Covertile | Leica Biosciences | S21.4611 | section 3 |
Bond(TM) Titration Kit | Leica Biosciences | OPT9049 | section 3 |
Coverslip 24 x 32 mm #1 (0.13-0.16 mm) | Fisher Scientific | 15717592 | section 2 |
coverslip 24 x 50 mm | VWR | 631-0146 | section 2 |
DAPI Fluoromount-G | VWR | 0100-20 | section 3, whenever monoplex slides need to be quickly checked, not for official analysis, then DAPI is stained seperately for better results |
Eosine | section 2, home made | ||
Ethanol 99.5% | VWR | 4099.9005 | section 2 |
Fluoromount-G | VWR | 0100-01 | section 3 |
haematoxyline | – | section 2, home made | |
ImmuNet | – | immune cell detection and phenotyping pipeline | |
inForm software 2.4.10 | Akoya BioSciences | – | section 4 & 6 |
OPAL 480 reagent pack | Akoya BioSciences | FP1500001KT | section 3 |
OPAL 520 reagent pack | Akoya BioSciences | FP1487001KT | section 3 |
OPAL 570 reagent pack | Akoya BioSciences | FP1488001KT | section 3 |
OPAL 620 reagent pack | Akoya BioSciences | FP1495001KT | section 3 |
OPAL 690 reagent pack | Akoya BioSciences | FP1497001KT | section 3 |
OPAL 780 reagent pack | Akoya BioSciences | FP1501001KT | section 3 |
Opal 7-Color Automation IHC Kit 50 slide | Akoya BioSciences | NEL821001KT | section 3 |
PhenoChart 1.1.0 | Akoya BioSciences | – | section 5 |
PhenoImagerHT | Akoya BioSciences | CLS143455 | section 4, digital pathology imager with slide viewer and imaging software (formerly known as Vectra Polaris) |
Quick-D mounting medium | Klinipath | 7280 | section 2 |
QuPath 0.3.2 | whole slide image analysis software platform | ||
R 4.1.1 | |||
Slide boxes | VWR | 631-0737 | section 1 |
SuperFrost Plus | Thermo Scientific through VWR | 631-9483 | section 1 |
Vectra Polaris software 1.0.13 | Akoya BioSciences | – | section 4 |
Xylene | VWR | 4055-9005 | section 2 |
.