In the last decade, next generation sequencing has enhanced our understanding of the functions of genes, proteins, regulatory elements, and cellular processes. However, the need remains for this information to be carefully integrated into other layers of biology, such as the organisation of cells in complex populations and tissues. This is especially useful for heterogeneous tissues such as tumors, the brain, and the gastrointestinal tract. It is also particularly crucial for diagnostics and classifying disease state through phenotypic analysis. This has been partially achieved through single cell analysis - such as single-cell RNA-seq and multiplexed fluorescence in situ hybridisation (FISH) - but has not been as extensively exported to proteomics [4].

One of the ways researchers and clinicians investigate tissue biomarkers in situ is through immunostaining of cells or tissues. To increase the capabilities of these technologies, there have been many developments towards multiplexing in recent years, permitting the analysis of various biomarkers simultaneously, for example, during immunotherapy in cancer with 2-50 markers [5]. This review summarises the technologies available for multiplexing immunohistochemistry, and discuss how such developments are employed.

Immunohistochemistry (IHC) represents a powerful tool used for the identification of target proteins in the context of the cell or tissue analysed. This type of in situ analysis is effective in assessing the localisation of proteins in the cell, comparing the expression of a protein in a heterogeneous cell population, and can be used to identify sub-populations of cells based on the expression of specific markers. It is an important tool in the post-genomic era, necessary for uncovering protein function [6, 7].

Immunohistochemistry works through the following principles. Samples such as mammalian tissues are rapidly fixed (most commonly in formaldehyde) and embedded in paraffin wax - or frozen where this is not possible - to maintain their morphology. Samples are then thinly sliced into sections and mounted onto glass slides, and then dried. Paraffin is removed, and antigen retrieval is performed where required. Antigen retrieval reverses the masking of epitopes that can happen during sample processing. Blocking steps are then completed in order to inhibit the activity of endogenous enzymes and non-specific binding of antibodies to reduce background staining [8].

The sample is then incubated with the antibody of choice to detect the antigen. Usually, a primary antibody, followed by a secondary antibody conjugated to either a fluorescent marker, which can be detected on a fluorescent microscope or by an enzyme (referred to as chromogenic staining). In the latter case, when incubated with the appropriate substrate, the enzyme activity will lead to the production of coloured substance at the site of antibody binding. Sometimes, direct staining is used, where the primary antibody is directly labelled. However, this often leads to lower signal strength. Prior to imaging and analysis, counterstaining is carried out to provide contrast to the image, and can highlight other structures such as the nucleus (DAPI), or the cytoplasm (eoisin) [9].

In complex and heterogeneous tissues such as tumors and brain sections, it is challenging to use traditional immunostaining to identify target proteins or cell types. This is because classical immunohistochemistry relies on detecting one or two targets at a time. This is for various reasons, including the limited potential combinations of fluorophores that can be used in a single experiment, fluorescent bleed-through, difficulty in detecting colocalization in chromogenic staining due to colour mixing, and antibody cross-reactivity [10]. This means standard IHC methods present a technical hurdle for many applications where it is necessary to measure multiple biomarkers in individual cells.

However, in the last decade, various adaptations of IHC and related technologies have been developed through which several targets can be stained and visualised in one sample. Together, these applications are often referred to as “multiplexed IHC” [8]. These applications are hugely useful for detecting multiple biomarkers in various cell types in complex tissues such as the tumour microenvironment [5], neural tissues, and lung biopsies.

Some initial examples of so-called “low level” multiplexed IHC closely follow the principals of standard IHC [8, 11]. Such technologies involve the use of carefully designed antibody cocktails in an inverse colour scheme that localise to distinct cellular locations [11]. These approaches help to identify patterns in the distribution of groups of proteins in different cell types, where previously only one or two proteins could be assessed simultaneously.

In terms of the more advanced, higher-level applications of multiplexed IHC, there are various approaches relying on distinct principles. These can be loosely categorized into three flavours: stain removal, fluorophore inactivation, and DNA barcoding.

One of the most commonly used approaches of multiplexing IHC is stain removal [6, 8, 10]. Stain removal technologies (sometimes called dye cycling) refer to a class of protocols that rely on a cycle of antibody staining, image capture, removal of stain, and re-staining to greatly increase the number of markers that can be tested in a single experiment. There are a variety of ways in which a stain or signal can be removed; two examples include multiepitope-ligand cartography (MELC) and sequential immunoperoxidase labelling and erasing (SIMPLE). Others include multiplexed immunohistochemical consecutive staining on the same slide (MICSSS) [12].

One of the first examples of multiplexed IHC based on stain removal technologies is called multiepitope-ligand cartography (MELC). MELC relies on sequential rounds of antibody staining and photobleaching to remove fluorescent signals, facilitating the imaging of hundreds of different proteins in one sample [6, 13-15]. This procedure is fully automated. Maps showing protein organisation can be generated by representing the data as binary vectors, facilitating the visualisation of complex interactome networks. MELC has been frequently used to study the intricacies of the immune system [16-18]. However, photobleaching can only be applied to the field of view, limiting the depth of information gathered in one round of cycling. The protocol is also fairly time-consuming and requires an expensive robotic set up coupled to a fluorescent microscope.

Another technique incorporating sequential wash steps was published in 2009 by Glass et al [10], where they describe SIMPLE. SIMPLE relies on sequential rounds of staining, enabled by using the alcohol-soluble red peroxidase substrate 3-amino-9-ethylcarbazole (AEC). Briefly, in line with standard procedures, the sample is fixed and embedded. The sample is then subjected to counterstaining, and a counterstain-only image is acquired. Antigen retrieval is then carried out, removing the counterstain. Next, immunohistochemical staining with AEC is performed, and imaging takes place. The AEC precipitate is removed by washing in 95% ethanol, and the antibody is removed in an elution solution. This process is repeated, and a single composite image with pseudocolors can be generated.

SIMPLE has been applied for probing the immune system [19-21], often in the context of cancer [19, 20, 22]. This highlights the effectiveness of SIMPLE in investigating highly complex and heterogeneous cell populations. However, it is limited to the testing of five antibody labels per section.

Multiplexed fluorescence microscopy method (MxIF) was introduced in 2013 [23], where it was described as a multiplexed fluorescence microscopy method for analysing multiple markers in a quantitative manner in fixed tissue samples. The technology aimed to overcome the limitations associated with standard immunofluorescence or immunohistochemistry techniques. The principle of the technique is similar to the stain removal techniques where pH, denaturation or photobleaching is used to remove a stain, but instead, this technique relies on inactivation of the fluorophores. The fluorophores are inactivated through alkaline oxidation chemistry, which quickly eliminates cyanine-based dye fluorescence, and was patented by the research group (US patent 7,741,045).

Briefly, background autofluorescence tissue images are acquired, and the sample is stained with fluorescently labelled primary antibodies. Images are captured, and the fluorescent dye is inactivated. Then, a new background autofluorescence image is required, and the sample is stained with the next set of primary antibodies. The cycle is repeated for all antibodies, and the images are then processed and analysed. The authors demonstrated the use of this technique by analysing 61 different proteins in single tissue samples [23], and later used the technology to demonstrate heterogeneity in ductal carcinoma – a type of breast cancer [24].

Cyclic immunofluorescence (CycIF) follows the principles of the technology patented by Gedes et al [23], bringing it into the public domain [25, 26]. The protocol also benefits from the use of standard reagents and open-source software.

CycIF has variants [25], but the general principles involve staining with fluorophore-conjugated antibodies and imaging, followed by fluorophore inactivation by oxidation (with hydrogen peroxide), alkaline pH, and light, and a final wash step. Subsequent rounds of staining, imaging, and removal are then performed. This is usually done with four-channel imaging, where 3 channels are used for different antibody-conjugated fluorophores, and the 4th channel is used for a counterstain, such as Hoechst.

CycIF can be combined with live-cell imaging [26], and has been used to study the immune microenvironment of brain tumors [27] and SARS-CoV-2 viral infection in macaque lungs [28]. An extension of CycIF was published in 2018 that is applicable to fixed tissues, called t-CycIF [29, 30]. CycIF is cost-effective, relies on standard reagents and commercially available antibodies. This makes it more accessible and applicable to a wide variety of contexts.

Multiplexing techniques relying on image cycling are not without their caveats. For example, it has been suggested that cycles can be incomplete, or could change the binding affinity of the target protein to the antibody [31]. Also, the tissue can become damaged or degraded throughout cycles of staining, which could be especially cumbersome when studying tissues with an intricate morphology.

One of the problems with detecting multiple markers in a sample is that sometimes the epitopes are present at a low copy number. Multiplexed signal amplification methods aim to address this need, providing strong signals for imaging even small amounts of protein. Tyramide signal amplification represents one such method.

Tyramide signal amplification (TSA) aims to overcome some of the hurdles associated with dye cycling and fluorophore inactivation methods such as the cost associated with robotics attached to microscopy systems and the time taken to carry out imaging. TSA is an enzyme-mediated signal amplification method that can be employed to detect proteins present in low amounts in the cell. A secondary antibody is conjugated to an enzyme such as horseradish peroxidase (HRP) or alkaline phosphatase. The enzymatic reaction is used to catalyse the covalent binding of tyramide-labelled fluorophores to the sample at the target site. The amplification comes from the substantial amount of fluorophore that is deposited at the target site, producing an intense signal [32-34]. To facilitate multiplexing, microwave treatments can be performed between rounds of staining to remove signals before the application of new antibodies. This process can also be automated. Parra et al shared their experience after staining 4000 FFPE tumor samples using the multiplex TSA system for translational studies in a recent article [35].

TSA has been used to characterize immune environments in multiple cancers, such as the tumour microenvironment in Hodgkin’s lymphoma [36], and breast cancer [37]. Very recently, TSA was used to identify distinct subsets of macrophage in mammalian tissue [38].

Another approach to multiplexing is using the innate properties and specificities of DNA itself. These technologies take advantage of DNA’s sequence-specific binding properties to generate highly specific staining which is intrinsically applicable to multiplexing and mapping.

DNA exchange imaging (DEI) builds upon a previously reported super-resolution technology that permitted multiplexed imaging of DNA in situ or in vitro [39] and adapts the technique to more complex cell populations and tissues. Briefly, the antibodies that bind to markers of interest are labelled with a unique DNA barcode and are applied to the sample. To give a signal that can be captured through microscopy, so-called “imager” strands that fluorescently labelled and bind to the DNA target are sequentially added, imaged, and removed through buffer exchange after each round of imaging [4]. This technology was later built upon to create SABER, which will be discussed in detail below [2].

[enlarge]

Figure 1. CODEX schematic from Goltsev et al, 2018 [1]. Antibodies are conjugated to a barcode with a corresponding dye-labelled reporter. The sample is stained with an antibody cocktail with as many as 50 antibodies. A subset of signals is revealed by the addition of the first cycle of reporters. Then, imaging of spectrally distinct dyes (in this case Cy3 and Cy5) takes place for the first cycle. Then, these reporters are removed, and the next cycle with the second index nucleotides is carried out, and the process is completed.

One of the major advantages of DNA exchange imaging is its speed, overcoming the need for multiple cycles of imaging over several hours. The technology can also be adapted to standard imaging platforms and is applicable to super-resolution microscopy such as SIM, STED and STORM [4, 40, 41]. The authors showed this technology to be effective in complex structures such as mouse retina tissue samples [4].

An alternative technique for multiplexed imaging based on DNA barcoding is called CODEX, shortened from CO-Detection by indEXing [1]. CODEX works through the following protocol, reliant upon the actions of DNA polymerase: antibodies are labelled with unique double stranded DNA sequences with 5’ overhangs. Samples are stained with all the tagged antibodies simultaneously. Samples are then incubated with a pool of nucleotides containing two “index” nucleotides, and two fluorescent “labelling” nucleotides. The index nucleotides fill in the first index position across all antibodies, but the oligo labels are constructed such that only the first two antibodies can be labelled with one of the fluorescent nucleotides, and only if the overhang was first filled in by the index nucleotide. Those two antibodies and the resulting fluorescent signal are then imaged. Then, the fluorophores are cleaved by the reducing agent TCEP, and washing steps take place. The sample is then subject to a second cycle of staining, where it is incubated with a different indexing nucleotide. Finally, the images are reconstructed. The CODEX assay platform is now commercialized and owned by Akoya Biosciences, and has been renamed PhenoCycler by Akoya in 2022. A schematic is shown in Figure 1, taken from the initial publication [1].

CODEX was shown to be effective against 66 antigens in this publication, and with the premise of the technology, could permit the analysis of many more. CODEX is relatively cost-effective and quick, requiring a standard 3 channel fluorescence microscope, but does require the addition of an automated fluidic setup. The effectiveness of CODEX at characterizing the architecture of complex cell populations was demonstrated in its initial publication, with the authors utilizing the technology to show changes in spleen morphology and cell-cell interactions with the progression of auto-immune disease [1]. EK Neumann et surveyed renal cell diversity with CODEX for 23 antigens [42]. PA Szabo et al profiled the lung tissues from SARS-Cov2 patients with Akoya Biosciences Opal 7-Color IHC Kits [43]. AR Palla et al used CODEX imaging to identify the source of 15-PGDH in aged muscle tissue [44].

One of the shortcomings of technologies such as DEI and CODEX is that DNA oligos are directly conjugated to primary antibodies (in other words they do not employ a secondary antibody step for signal amplification). As a result of this single layer approach, fluorescent signals can be quite low. In 2019, a tool called SABER (Signal Amplification By Exchange Reaction) was developed that aimed to resolve this issue [2, 45]. The technology is applicable to both immunostaining [2] and FISH [45].

[enlarge]

Figure 2. Immuno-SABER schematic from Saka et al 2019 [2]. (a) Immuno-SABER relies on PER where primer sequence ‘a’ binds to a single-stranded ‘a*’ sequence on a hairpin. (2) Next, extension occurs via the actions of a polymerase enzyme. The hairpin possesses a stop sequence that releases the polymerase. (3) Branch migration occurs, displacing the new sequence. (4) The hairpin and extended primer dissociate. (5) The final product is generated through repetition of this process. (b) (1a) The immuno-SABER process begins with antibody staining, where multiple antibodies containing different DNA tags bind multiple targets. (1b) Concatemers are produced by PER as described in Figure 2a. (2) Concatemers are then hybridized to the unique DNA sequence (in blue) on each antibody. (3) Fluorophore (purple star)-labelled DNA strands (referred to as ‘imagers’ hybridize to the repeated sites on the concatemers. (c) For further multiplexing, Exchange-SABER can be performed. (1) Different proteins (t.1, t.2, t.n) are bound by antibodies labelled with distinct DNA oligo sequences (b.1, b.2, b.n) during sample staining. (2) The amplified (i.1, i.2, i.n) concatemers are hybridized to the antibody-labelled markers (b.1*, b.2*, b.n*). (3) The fluorophore-labelled imagers complementary to target t.1 (i.1*) are hybridised, and images captured. (4) To image multiple targets in the same sample with the same range of fluorophores, the imagers can be exchanged, such that imagers complementary to target t.2 (i.2*) are hybridized (and i.1* are dehybridized). These exchange cycles can be repeated until all targets have been images. (5) Finally, a final image can be reconstructed using pseudocolours to overlay the images and visualise different targets in one image.

SABER is based on a method of single-stranded DNA synthesis called the primer exchange reaction (PER), which permits the generation of a concatemer based on a short primer sequence. The kinetics of this reaction can be controlled by changing parameters such as magnesium concentration [46], meaning the amplification process is highly controlled by the user. PER means oligo-based FISH probes can be enhanced through the extension of the sequence, followed by hybridisation to the substrate and downstream imaging. The paper, published in Nature methods, demonstrated that SABER allows the FISH probe signal to be amplified up to 450-fold, and used against 17 targets in one experiment [45].

In the case of immuno-SABER, multiplexed immunohistochemistry is performed without the need for in situ enzymatic reactions, making it highly controllable for use in a range of targets. Immuno-SABER first requires staining of a sample with DNA barcoded antibodies of interest. Using PER, single-stranded DNA concatemers are then hybridized to the sample. Fluorophore-labelled single strand oligos then bind to repeated signals in the concatemer, generating an amplified signal. In order to maximise the number of targets that can be imaged at once, rounds of sequential hybridisation and dehybridisation of fluorophore labelled oligos can be employed in a process referred to as “Exchange-SABER”. The Immuno-SABER process is presented in Figure 2, in a schematic taken from Saka et al [2].

In this initial publication, the effectiveness of SABER was demonstrated in a range of sample types including cultured cells, cryosections, and fixed tissues [2]. They also used the technology to analyse 10 markers in cryosections of the mouse retina, and SABER was even adapted for super-resolution platforms. Importantly, SABER has been cited by the NIH Common Fund Human Biomolecular Atlas Program (HubMAP) as a transformative technology among those that will be used to map a 3D, single-cell resolution atlas of the human body [47].

Aside from traditional microscopy imaging, flow cytometry can also be used to analyse biomarkers in cell populations [48]. However, similar limitations apply with regards to the limited number of fluorophores that can be exploited in a single experiment. Mass cytometry is another example of a non-microscopy-based method of quantifying multiple biomarkers, that allows users to carry out multiplexing [3, 49]. More detailed discussion on mass cytometry is available here.

[enlarge]

Figure 3. Schematic of mass cytometry taken from Di Palma & Bodenmiller, 2014 [3]. Cells are cross-linked and permeabilized before being stained with antibodies labelled with metal isotopes. Cells are then nebulized, ionized, and atomized before being introduced into the mass cytometer. The composition and quantity of the metal isotopes are measured and correlated with the presence of marker proteins in single cells [3].

First, as in traditional imaging or flow cytometry-based methods, samples are stained with antibodies complementary to targets of interest. However, these antibodies are tagged with metal isotopes rather than fluorophores. These isotopes act as the reporter in the assay. When the labelled samples are analysed by the mass cytometer (in single-cell droplets), the isotopes can be identified by their atomic mass, and their abundance can be determined. This abundance correlates with the amount of the target protein present in the sample [3, 50].

This mass cytometry technology and instrument were first described in a report from 2009, Bandura et al [50] where they introduce the Cytometry by Time Of Flight (CyTOF) mass cytometer. The technology is based on inductively coupled plasma time-of-flight mass spectrometry (ICP-TOFMS), and the workflow is shown in Figure 3, taken from Di Palma & Bodenmiller, 2014 [3].

Since its conception, there have been several advances on mass cytometry, enhancing its throughput and application to screening. For example, one method called Mass-tag Cellular Barcoding (MCB) [51], based on fluorescent cell barcoding [52], greatly increases the multiplexing capabilities of mass cytometry. MCB hugely increases the throughput of mass cytometry, because individual samples are labelled with a unique binary combination of tags before samples are pooled prior to antibody staining [3].

One dimension of information missing from mass cytometry is spatial information, as the technology relies on cells being in a suspension. Spatial information is important in intricate tissues, tumour microenvironments, and in analysing cell-cell interactions. Imaging mass cytometry represents another adaptation of mass cytometry, where it is coupled with immunocytochemistry, immunohistochemistry, and importantly, a high-resolution laser ablation system [31].

Briefly, samples are stained with antibodies of interest using standard immunostaining protocols. These antibodies are tagged with metal isotopes as in mass cytometry. The sample is then ablated with a laser beam, “spot by spot and line by line”, before being transferred to the CyTOF instrument [31]. The measured reporter signals are then mapped using the coordinates of each laser spot, and an overlay image is generated based on this data [3, 31]. Over 100 markers can be detected simultaneously with this system.

Using this technique, researchers studied tumour heterogeneity in breast cancer samples and were able to identify specific subpopulations [31]. Since its publication, imaging mass cytometry has been applied to a variety of systems including two recent studies analysing the heterogeneity of multiple sclerosis lesions [53], and the pathophysiology of type 1 diabetes [54].

Around the same time as the initial publication of imaging mass cytometry, another similar technique was described, called multiplexed ion beam imaging (MIBI) [55]. This technology also uses antibodies tagged with metals, but analysis is carried out using secondary ion mass spectrometry (SIMS) [56].

Samples are incubated with antibodies as standard. To detect the isotopes, the tissue is rasterized, “pixel by pixel” using an ion beam that liberates the antibody-bound metal, which is subsequently analysed by SIMS [57]. Using a MIBI adaptation taking advantage of time-of-flight mass spectrometry (MIBI-TOF) [57], Keren et al probed the relationship between cell identity and tissue architecture, providing a multi-layered for organisation of the immune microenvironment [58].

One feature all of these methods have in common is the use of primary antibodies. Therefore, proper validation to ensure specificity and reproducibility is key [63]. This is especially crucial for multiplexed imaging, where often many antibodies are used in tandem. It is necessary to assess the order in which the antibodies are used for staining, and whether interference between antibodies exists when they are pooled [64], or whether antigenicity of certain markers changes across cycles [63].

Fortunately, most techniques are based on similar staining principles, so if one antibody is effective in – for example – imaging mass cytometry, it is likely to also be effective in CyC-IF with minor modifications, taking into account detection methods [63]. This makes it easier for researchers to begin building their panel of antibodies. A robust protocol for testing antibodies for these applications was published in 2019 [63]. Some commercial platforms such as CODEX also have lists of validated antibodies that are tried and tested on their system.

With the huge amounts of data generated from multiplexed imaging, there is the requirement for streamlined analysis protocols that allow researchers to efficiently process the data, ask the correct questions, and properly interpret the output. There are a huge number of computational tools available for the analysis of this data (reviewed by Liu et al, 2019 [65] ), including fairly standard, open source tools such as ImageJ [66] and CellProfiler [67]. One pipeline that was developed specifically for the analysis of multiplexed imaging datasets is called Cytokit [68], which allows users to process and analyse large and complicated imaging datasets, such as those from CODEX or SABER.

Scientists’ understanding of molecular biomarkers that define normal and disease states is constantly evolving. To accurately characterise cell populations in research and diagnosis in a high-throughput and high-resolution manner, novel imaging platforms, assays, and technologies are required. Multiplexing immunohistochemistry is one avenue increasing the capabilities of clinical research and diagnostics, providing researchers with the tools to investigate complex and heterogeneous tissues in a powerful manner.

There are a multitude of multiplexing platforms and chemistries that researchers can choose from to match their applications. Some technologies offer the benefit of being able to assay over 50 biomarkers in one sample, other systems focus more on high sensitivity, or speed and low cost. Some of the technologies reviewed here are highly novel, and further advances and applications in the field are inevitable. Likely, the next generation of technologies will involve more interdisciplinary collaborations, with the involvement of clinicians, molecular biologists, chemists, engineers, computer scientists, and more [34]. Enhanced data analysis and automation are likely to be at the forefront of the clinical application of multiplexing technology.