Bright-field illumination is one of the most widely used observation modes in optical microscopy. The illumination light is transmitted through the sample and contrast is generated by the absorption of light in dense areas of the specimen. This enables researchers to see the (sub)structures in the sample, for example, desmosomes in mucosa tissue (Raju et al., 2014). Due to the limited resolution of light (in theory, limited to 0.2 µm), light microscopy is often accompanied by electron microscopy (EM) images to confirm the findings. White light microscopy makes it efficient to get a fast and general overview of the intercellular junctions. Even though the resolution is lower, the field of view is larger, and the surrounding structures are visible. It is an easily accessible, low technical demanding technique that is widely used and available in every research laboratory.

Not only the ultrastructure of intercellular adhesions is of interest, but research has also been conducted to unravel the complex composition of the adhesion complexes, including the different proteins. This can be done by using fluorescence microscopy or EM combined with the labeling of the individual junctional proteins. For instance, protein labeling when using light and electron microscopic techniques was a critical part of the discovery of the area composita (a mixed type of junction at the intercalated disc of cardiac muscle cells) (Borrmann et al., 2006; Franke et al., 2006).

Imaging of cell junctions with light has challenges due to the limited spatial resolution of fluorescence microscopes (250 nm in xy and 600 nm in z). Another challenge is the size of the junctions, which is close to or below the resolution limit. Therefore, it is difficult to distinguish between different proteins within a junction, and proteins can appear co-localized when in reality their organization is distinct (Rayleigh 1879). Recent advances made it possible to push the lateral (xy) resolution of fluorescent imaging beyond 200 nm, up to the two digits nanometer range (Table 1).

Confocal microscopy is a fluorescence imaging technique that uses lasers to illuminate and scan the tissue at a certain depth (called the focal plane) while pinholes are present to physically block out-of-focus light and thus eliminate or reduce background information away from the focal plane (Figure 3A and Figure 5). Emitted fluorescence is recorded, and optical sectioning of the specimen is obtained. Confocal microscopes may reach a resolution of 170–250 nm laterally (xy) and around 500 nm axially (z) (Ooshio et al., 2010; Elliott 2020; Valli et al., 2021) and relatively thick samples (5–30 µm) can be imaged. Thicker samples do produce more out-of-focus light because the objective lens does not have sufficient depth of focus resulting in the detection of light from sample planes above and below the focal plane. However, many new methods have emerged to overcome this and give better results in signal-to-noise ratio (SNR) and resolution, including pixel reassignment (Airyscan) (Murray 2011; Elliott 2020; Valli et al., 2021).

One example out of thousands of imaging intercellular junctions by confocal microscopy: research showed that claudins and junctional adhesion molecules (JAMs) are major cell adhesion molecules at tight junctions, whereas cadherins and nectins are major adhesion molecules at adherens junctions. Claudins and JAMs are associated with zona-occludens (ZO) proteins, whereas cadherins are associated with β- and α-catenins, and nectins are associated with afadin. To investigate how tight junction components are recruited to the apical side of adherens junctions during polarization of epithelial cells, researchers used confocal laser scanning microscopy to study the roles of afadin and ZO-1. Researchers saw that nectins first form cell-cell adhesions by recruiting the cadherin-catenin complex in order to form adherens junctions. This is then followed by the recruitment of the JAM-ZO and claudin-ZO complexes to the apical side of adherens junctions in order to form tight junctions (Ooshio et al., 2010). Before the formation of tight junctions, ZO-1 interacts with afadin. However, during and after formation of tight junctions, ZO-1 dissociates from afadin and is associated with JAM-A. Confocal microscopy was also useful to see that disruption of afadin impaired the formation of both adherens junctions and tight junctions while knockdown of ZO-1 only impaired the formation of tight junctions and not of adherens junctions (Ooshio et al., 2010).

Spinning disk technology, a fast confocal microscopy technique, employs a parallel array of pinholes on a rotating disk. With a similar resolution as in confocal microscopy (Zubkovs et al., 2018), this technology showed that epithelial apical junctions (tight and adherens junctions) of contractile tissue display a periodic assembly of bipolar non-muscle myosin II filaments that interlace with the peri-junctional actin and α-actinin (Figure 3A1) (Ebrahim et al., 2013). In that way, the structure forms a continuous belt of muscle-like sarcomeric units around each epithelial cell. The spinning disk illustrates that the sarcomeres of adjacent cells appeared to be precisely paired across the junctional line to create a transcellular contractile network as seen with transmission electron microscopy (TEM; see below) (Ebrahim et al., 2013).

Additionally, confocal microscopy elegantly showed that desmosomes stay intact and are crucial for the apoptotic cell extrusion in a monolayer of epithelial cells (Figure 3A2) (Thomas et al., 2020). Confocal imaging showed that upon apoptosis the formation of actomyosin cables occurred in the vicinity of desmosomal junctions and that they subsequently deviated from desmosomal junctions during its constriction, which coincided with a loss of straightness of the desmosomal junction.

Confocal microscopy is thus an ideal technique to study the localization of proteins in and around the intercellular junction complexes to see how specific proteins function and how they interact with each other. Confocal imaging is a very integrated technique, not technically demanding and readily available in almost every research lab. It is often used as a first approach to understand or unravel complex intercellular junctions.

The progresses made to overcome the optical diffraction limit in the field of optical microscopy over the last few decades has greatly advanced the resolving power. This has opened windows for researchers to observe objects with much higher spatial resolution. Compared with wide-field fluorescence microscopy, super-resolution techniques offer a superb resolution to study the protein organization, protein dynamics and protein co-localization at nanoscale in macromolecular complexes. The resolving power of 20–120 nm (Table 1) (Hell and Wichmann, 1994; Neil et al., 1997; Betzig et al., 2006; Rust et al., 2006; Valli et al., 2021) also enables to distinguish between homogenous mixing and sub-junctional clustering, to see the relative position of proteins within the junction (proximal or distal to the membrane) and allows a precise identification and measurement of the size, distribution and composition of junctional clusters. The type of super-resolution microscopy depends on the researcher’s needs, which includes the spatial resolution, temporal resolution and ease of sample preparation. Thicker samples are still challenging to image due to light scattering and optical aberrations. Cautious data interpretation is also warranted when analyzing data at a nanometer resolution level. For example, researchers need to be aware that antibodies around 10 nm in size leave a 20 nm cloud of uncertainty around the object of interest (when labeling with primary and secondary antibodies). Other labeling techniques, including direct labelled primary antibodies, Fab fragments, nanobodies, affimers and chromobodies, might be considered as they reduce the distance between the protein and fluorophore (Bates et al., 2007; Carrington et al., 2019).

A variety of super-resolution microscopic techniques exist. Below we give a short description of the techniques, with their pros and cons, and some elegant examples of imaged adhesions (Figure 3C and Figure 5). Super-resolution techniques can be broadly split into two categories: super-resolved ensemble microscopy techniques, which improve the resolution of overall structures, and super-resolved single-molecule localization microscopy techniques (SMLM), which use localizations of individual fluorescent molecules to build up an overall structure (Valli et al., 2021).

Intercellular junctions are complex multi-protein structures that are densely packed to provide inter- and intracellular communication as well as adhesion and structural integrity for proper homeostasis. To resolve these junctions in detail, often researchers want to study their exact interaction with one another. Super-resolution techniques provide the necessary lateral (xy) and axial (z) resolution to accomplish this, when combined with proper labeling and cautious interpretation of the data. Proximity assays and Forster resonance energy transfer (FRET) imaging are two techniques that have been used to unravel the protein network of tight junctions, adherens junctions and desmosomes.

The technique of TEM is based on an electron beam that passes through the sample (thin sections of 60–80 nm of tissue or cultured cells) such that the beam will be absorbed and scattered by the sample, producing contrast that can be imaged (Winey et al., 2014) (Figure 4A and Figure 6). Using electrons as a source instead of light has the advantage that imaging can be performed at a much higher spatial resolution. The wavelength of an electron beam is much shorter than that of visible light and images with a resolution up to 2 nm can be obtained. Therefore, EM can be applied for high-resolution (ultra)structural analysis of whole tissues. Disadvantages are the limited sample depth the electrons can penetrate, imaging is done under vacuum conditions and there can be radiation damage. Therefore, microscope design, sample preparation, imaging and image processing must be carefully optimized. The quality of an image is mainly defined by three factors: contrast, resolution, and SNR [reviewed in (Franken et al., 2020)]. Staining of the sample with heavy metals is performed to increase the scattering events caused by the electrons when they interact with the sample, leading to a higher contrast of the image. Because the specimen chamber is under vacuum, samples need to be highly processed and are no longer in the natural state. Samples can be prepared by chemical fixation, dehydration, and staining with heavy metals which enhances contrast as mentioned above and protects somewhat against dehydration and radiation damage. Another way to preserve the tissue is cryo-fixation (cryo-EM); i.e. samples on a grid are frozen (either by plunging into a liquid cryogen or placing under high pressure, called plunge-frozen or high-pressure frozen respectively), using liquid ethane or ethane/propane to avoid the formation of ice crystals. Only thin samples (less than ≈10 µm) can be plunge-frozen, while specimens up to 300 µm can be high-pressure frozen, to achieve sufficient freezing of the sample, including the center (Dubochet et al., 1988). Cryo-EM can be performed without heavy metal staining and the contrast is then based on phase contrast only. The advantage is that it resembles the most native state and results in a high resolution. Thin samples are obtained by a process called ultramicrotomy, which can be done at room temperature or at sub-zero temperatures which can be challenging (Richter et al., 2007). In order to create thin sections from thicker samples, recent developments including cryo-EM of vitreous sections (CEMOVIS), or cryo-focused ion beam (FIB) milling can be used to create thinned areas within cryo-frozen specimens [reviewed in (Franken et al., 2020)]. Imaging of cryo-frozen samples require imaging in a cryogenic environment, which is technically demanding. High-pressure frozen samples can be stained by freeze-substitution, embedded in plastic resin, cut by ultramicrotomy and imaged at room temperature. More information about TEM of biological samples can be found in a book chapter (Mielańczyk Ł and Romuald, 2015).

EM made a major contribution to the visualization of intercellular structures by unraveling the adhesions in high detail. Ideally, the connection of the junctions to the cell’s cytoskeleton and membrane elements are imaged together to observe how they make an active junctional complex (see below). Additionally, individual proteins can be localized in context of the 3D volume.

Overall, researchers see that desmosomes are electron-dense and surrounded by a fuzzy area while adherens junctions are less electron-dense with the same intermembrane distance. For tight junctions however, there are different observations. In chemically fixed intestinal mucosa of the rat, tight junctions were seen as electron-dense apical membrane contacts forming a continuous belt-like attachment over a 200–500 nm distance by bringing the adjacent cell membranes in very close proximity (kissing points) (Figure 4A1) (Farquhar and Palade, 1963). Later, FFEM showed that tight junctions form interconnecting strand networks between cells that vary in number and morphology among different tissues. The frozen biological samples (with possible chemical fixation prior to freezing) are physically broken apart, typically fracturing at weak hydrophobic sites such as membranes, and a platinum-carbon replicate of the fracture plane is then analyzed by TEM, exposing the intercellular junctions between cells (Figure 4A3) (Chalcroft and Bullivant, 1970; Claude and Goodenough, 1973; Staehelin 1973; van Deurs and Koehler, 1979; Furuse 2010). FFEM has a limited spatial resolution due to the inherent property of metal atoms to crystallize. Recently Krystofiak et al. used a combination of amorphous carbon replicas with phase-contrast EM (Danev et al., 2014) to overcome this (Krystofiak et al., 2019). With this approach, they found that tight junction strands have an antiparallel double-stranded morphology.

When looking at the connection with the cytoskeleton, early observations using TEM of chemically fixed tissue section demonstrated that tight junction-associated actin filaments occur predominantly at sites of intercellular membrane apposition. These actin filaments are decorated by the actin binding region of myosin and appear to insert directly into the submembrane tight junction space (Madara and Pappenheimer, 1987). However, TEM of quick-freeze, deep-etch, rotary shadow replicas made it possible to distinguish between two different actin populations in epithelial cells (Höflinger 2014). Actin at the adherens junctions is organized as a ring and is composed of filaments that run parallel to the plasma membrane. This differs from the actin filaments found just beneath the tight junction membrane; these are organized as a meshwork of filaments (Hirokawa and Tilney, 1982; Hirokawa et al., 1983). Later, similar TEM-based techniques showed different features of the adherens junctions in a single layer of epithelial cells (chemically fixed and frozen). When sections were cut perpendicular to the cell membrane it revealed that adherens junctions have a 15–25 nm space between both membranes (Figure 4A2). With freeze-etching, it became clear that they are made up of a cytoplasmic macromolecular complex that consists of rod-like structures extending from the extracellular surface into the intercellular space, presumably catenin proteins of which the extracellular part may cant at about 60° with the plasma membrane (Miyaguchi 2000; Sluysmans et al., 2017). The standard TEM of chemical fixed cultured cells grown on a grid confirmed that the actin filaments run parallel to the adherens junctions (Buckley et al., 2014).

Cryo-EM of desmosomes in high-pressure frozen tissue sections showed that the desmosomal intercellular space has been variously reported as between 20 and 35 nm wide (Al-Amoudi et al., 2004). It is characterized by the presence of a dense midline with cross-bridges extending to the plasma membrane. When injecting the extracellular space of desmosomes with the electron-dense tracer lanthanum, researchers saw that the cross-bridges extending to the plasma membrane appear to alternate on opposite sides of the midline and that the desmosomal cadherin extracellular domains are ordered in all 3 planes (x, y, and z) (Rayns et al., 1969). So, the high resolution of EM and related techniques, including tracers and freeze-etching, allows researchers to distinguish the extracellular domains of cadherins in the intercellular junctions (Miyaguchi 2000; Al-Amoudi et al., 2004).

An evolved technique of rotary shadowing EM to reveal the surface topography of a sample is platinum replica EM (PREM). PREM is a specific type of EM where a cell on a glass coverslip is extracted or unroofed, followed by fixation and critical point drying. Then platinum is evaporated on a 3D sample at an angle during rotation of the sample after which carbon is reinforced on it (Heuser 1981). This reveals the topography and makes it an ideal technique for the structural analyses of the cytoskeleton (Svitkina and Borisy, 1998; Svitkina 2016). PREM achieves the high resolution typical for EM and the cytoskeletal structures are well preserved with visibility of single filaments, even if they are densely packed. Sample preparation for PREM is fast and efficient making it an inexpensive and accessible technique. However, only thin samples can be imaged as only the surface is made visible. The technique requires samples to be attached to glass surfaces which is not ideal to study membrane structures. Because of its informative value for cytoskeleton studies, PREM has been used and revealed the double-stranded morphology of the tight junction intermembrane fibrils (Krystofiak et al., 2019). In combination with light microscopy and immunolabeling for vascular endothelial cadherin with immunogold, it revealed the relationship between the parallel extensive branched actin network and the junctions between two endothelial cells (see Figure 4A4) (Svitkina 2017; Efimova and Svitkina, 2018). Specifically, the vascular endothelial cadherin colocalizes with the Arp2/3 complex of the F-actin network at different adherens junction types and is positioned at the interface between two oppositely oriented branched networks from adjacent cells. It is also proposed that the actin cytoskeleton at adherens junctions is a dynamic push-pull system (Efimova and Svitkina, 2018).

TEM allows imaging of intercellular junctions and ultrastructure of tissue or cultured cells at a very high (nanometer) resolution. Junctions in a cell layer are imaged perpendicular to the cell surface layer and with some techniques (like freeze-etching and PREM), structures just beneath the cell surface can be exposed and imaged. This is a suitable technique for visualizing the morphology of junctions, unraveling the relationship between the actin cytoskeleton and the intercellular complexes—with PREM it is even possible to see individual filaments in a 3D manner—and for measuring the intermembrane space between cells or see the extracellular domains of cadherin proteins. PREM requires some technical handling of the samples and optimization of the sample preparation.

In 2D it is difficult to interpret the complexity of the junctions; 3D imaging gives more accurate data. ET is an EM-derived technique and is based on a tilt series of 2D images acquired at different viewing angles, which are subsequently aligned and combined digitally into a 3D volume (see Figures 4B, 6). For an ideal reconstruction, tilt series contains projections from a ±60–70° rotation of the object to include projections from all possible directions [reviewed in (Ercius et al., 2015)]. This gives a 3D view through the depth of the specimen at a high resolution. Single-axis ET can cause imaging artifacts (known as the missing wedge problem) such as elongation of structures in the z-direction and loss of information in the x-direction (perpendicular to the tilt axis). Developments of dual-axis imaging partially overcome this problem (Guesdon et al., 2013). Sub-tomogram averaging can additionally be used to reduce the imaging artifacts and decrease the SNR. By making several images at different angles the resolution of ET is higher compared to EM, particularly in z, and especially when tomograms are acquired from viewing at two different axes. Cryo-ET is often used to preserve the sample in its native state. In combination with sub-tomogram averaging, it results in an atomic resolution (5–10 Å) and allows for single particle analysis (Nakane et al., 2020; Yip et al., 2020; Zhang et al., 2020), but it has not yet been reached with imaging of biological cells or tissue. This technique provides molecular resolution 3D images of cellular landscapes and can visualize the junction from different angles in the tissue, not just perpendicular to the cell membrane. ET is, as TEM, restricted in volume because the sample must be thin sections (limited to about 300 nm) to render them electron transparent. Thicker samples can be imaged with the of more sophisticated approaches and instrumentation (Lučič et al., 2013).

Freeze-substituted ET revealed that desmosomes are more loosely arranged (He et al., 2003) (Figure 4B1) compared to those seen using CEMOVIS from a ca. 150 μm thick fully vitrified biopsy (Al-Amoudi et al., 2004). However, using CEMOVIS data results differ from what has been seen before. It has to be noted that those types of samples are typically larger by 15% or more compared to fixed, dehydrated specimens [reviewed in (Al-Amoudi et al., 2004)]. Cryo-ET experiments combined with sub-tomogram averaging and atomistic molecular dynamics simulations demonstrated that cryo-ET is a powerful method for high-resolution imaging. It was able to reveal discrete groups of cadherin molecules with different plausible cadherin arrangements, and their cis and trans interactions in desmosomes (Sikora et al., 2020).

Using ET, researchers made a 3D reconstruction of an intercalated disc, a component of the connection between murine cardiac muscle cells where mainly gap junctions, desmosomes and a mixed type of junction (area compositae) are located (Borrmann et al., 2006; Franke et al., 2006). This revealed the pattern and the presence of desmosomes and mixed junctions in 3D (Leo-Macías et al., 2015).

ET can obtain higher resolution than traditional TEM and also has the advantage of imaging samples in 3D. ET can be used to study the filament arrangement within the junctions and unravel the binding interactions (cis or trans) between substructures of the intercellular junctions. It can also be used as a quantitative tool of the intermembrane distances in 3D. The sample preparations, technical experience and microscope availability are similar to that of TEM.

Volume scanning EM (volume-SEM) offers the opportunity to image larger volumes with a greater sample depth compared to ET. The technique is based on altering between scanning the tissue surface with a focused electron beam followed by cutting off a section of the en-bloc tissue (Figure 4B and Figure 6). This generates a large stack of 2D images on which segmentation is performed to render 3D reconstructions of the cell with its subcellular structures. Cutting the section is done by either a diamond knife (serial block-face (SBF)-SEM) or an ion beam (focused ion beam (FIB)-SEM), enabling the cutting of sections with a thickness of 25–50 nm and 5–20 nm respectively. The resulting high-resolution voxel sizes are up to 10 × 10 × 25 nm³ for SBF-SEM (Wanner et al., 2016) and 5 × 5 × 5 nm³ for FIB-SEM (Knott et al., 2011; Boergens and Denk, 2013). The maximum sample width for SBF-SEM is around 1 mm, while FIB-SEM is limited to a much smaller imaging field of 20–100 µm width (Titze and Genoud, 2016). The volume SEM can image depends on the available imaging time; the more volume you image the longer it takes. The sample preparation needs optimization per sample type, and often includes heavy metal staining for contrast and embedding in plastic resin. On top, coating of the sample with a thin layer of conductive material (i.e., gold) is performed to reduce the charging effects. Both SBF-SEM and FIB-SEM have been used to study the intercalated disc between cardiac muscle cells in 3D, along with the desmosomes and gap junctions of murine cardiac tissue (Figure 4B2) (Vanslembrouck et al., 2018; Vanslembrouck et al., 2020). The heavily folded intercalated disc is clearly visible, where two different regions can be distinguished with a different pattern of junctions on them. Moreover, FIB-SEM has revealed that in diseased mice, the intercellular junctions behave differently in a region-specific manner.

FIB-SEM is a useful, modern technique which is shown by the study that looked at the detailed location of the SARS-CoV2 virus particles in human lung epithelial cells to accurately quantify virus density and surface curvatures. Results of the FIB-SEM imaging and 3D reconstructions of interactions between the SARS-CoV-2 virus and the cell revealed a tight junction-mediated contact between adjacent cells with a dramatic surface viral density difference on either side of the tight junction (Figure 4B3) (Baena et al., 2021).

Volume-SEM is a great tool to visualize intercellular junctions and the connecting membrane in a voluminous and quantitative way at high resolution in 3D. It can be used to image the relationship between different junctions in a volume, their localization on the membrane surface, and the relationship to the other substructures of the cell. Immunogold labeling of tissue imaged by SEM has been done previously, but not yet for intercellular junctions. This could however, with sample optimization, be a very helpful approach to image individual proteins in their 3D imaged junctions (Goldberg and Fišerová, 2016). However, the main disadvantage of volume-SEM is the cost and technical handling of the machine, together with a need for proper optimization of the sample preparation. Even though there is no broad availability of this technique, the microscopes are often part of a service facility that helps the researcher with all the technical aspects, sample preparation and data interpretation.

X-rays can be used as an illumination light source while a wide range in resolution can be achieved using different energies with corresponding X-ray optics (Rawson et al., 2020); i.e. microtomography with a spatial resolution from 1 µm up to 100 µm and nano-imaging with a resolution down to 10 nm voxel size. The principle of X-ray tomography is based on taking projection images using X-rays from different angles around a sample. This is followed by reconstruction and segmentation to obtain 3D volumes and volumetric quantifications (Bradley and Withers, 2016; Guo and Larabell, 2019; Töpperwien et al., 2020) (Figure 6). Using X-rays generated at a synchrotron, images with a nanometer resolution (50 nm or better) can be generated. Because of the destructive nature of X-rays, samples are plunge frozen and thus imaged as close to the native state as possible due to cryofixation of the samples (Krenkel et al., 2015). SXT images the sample in the water window (Weinhardt et al., 2019), meaning that X-rays are absorbed by carbon and nitrogen in biological tissue more than by oxygen in water, resulting in grayscale data that can be used for quantitative analysis. However, limited penetration of soft X-rays results in a limited sample thickness of 10–20 µm (Carzaniga et al., 2014; Duke et al., 2014).

SXT has mainly been used to image adherent cells cultured on 2D surfaces or in cell suspension (Elgass et al., 2015; Harkiolaki et al., 2018). The resolution of this technique allows researchers to study the adhesion complexes between cells, as the plasma membrane is clearly visible in several studies (Peter Szekeres et al., 2020). However, the focus of most of these projects is the nucleus, lipid droplets and intracellular structures. Often, 3D reconstructions of single cells are the main goal with SXT, but imaging of multiple cells has been done (Loo Jr et al., 2001; Loconte et al., 2021) and efforts are evolving to image cell-cell connections between adherent cells (own unpublished results). For example, researchers saw multinucleated cells in SARS-CoV-2 infected embryonic kidney cells which could have points of interaction made visible with SXT, as seen for desmosomes in the FIB-SEM images from SARS-CoV-2 infected human lung cells (Baena et al., 2021; Loconte et al., 2021).

SXT allows the researchers to image intercellular junctions at nanometer resolution in the near-native state, and it is possible to reconstruct the ultrastructure of the cell in 3D. Even though imaging of samples is at cryogenic temperatures, only thinner samples (max of 10 microns) can be efficiently frozen as SXT is up to date limited to plunge freezing. To our knowledge, it has not been used yet to image desmosomes, adherens junctions or tight junctions. However, with some sample preparation optimization, we are convinced that the resolution of the technique allows the researcher to image these substructures. Correlation with light microscopy is also an option (Ekman et al., 2019), which makes it a good technique for studying cell-cell adhesion complexes.

In parallel with fluorescence imaging (see part 2.2), immuno-labeling with a certain tag/antibody can also be performed when imaging with electrons. Immuno-EM is a very useful technique to localize individual proteins at a nanometer resolution. It has the advantage that the ultrastructure of the junctions is still visible at the same time and in a single image. Immuno-EM has, however, limited molecular specificity, meaning that the antigen/antibody binding must be strong and specific in which proper controls are highly desirable to eliminate non-specific binding. Immunolabeling in TEM and SEM is a widely used technique in which different sizes of gold nanoparticles have been used to label different proteins resulting in multi-“colored” images (Koster and Klumperman, 2003; Straub et al., 2011; Boykins et al., 2016; Goldberg and Fišerová, 2016; Jones 2016). It can be, however, technically challenging and proper sample preparation is required (Griffiths and Lucocq, 2014). Immunolabeling can generally be done in two ways depending on the embedding process: pre- and post-embedding. During pre-embedding, the labeling is carried out on micrometer-thick cryostat sections of frozen tissues or chemically fixed tissue followed by embedding in resin followed by sectioning. Pre-embedding has the advantage of strong immunolabeling and preserving the antigenicity but is poor in preserving ultrastructural details. Post-embedding requires the antibody to bind on the surface of the ultrathin section and requires thawing of frozen sections with cryoprotection and hydrophilic resins, like the Tokuyasu method (Tokuyasu 1973). This technique is a better choice to preserve the ultrastructure. Recent developments exist that combine the benefits of both techniques (Jones 2016; Kim et al., 2021).

Immuno-EM helped already in the early ages of junctional research in numerous studies to identify the individual proteins located at the intercellular junctions as well as their distance to both the transmembrane and peri-junctional part of the cell (Stevenson et al., 1986; Jesaitis and Goodenough, 1994). For example, immuno-EM labeling was crucial for the identification of occludin and claudin protein at freeze-fractured tight junctions (Furuse et al., 1994; Furuse et al., 1998).

In a more recent study that investigated endocytosis in epithelial cells (enterocytes of mice), antibodies against occludin and caveolin-1 with a different size of gold particle were used. They observed that both proteins were present at the apical junction complex, with occludin at the tight junction while caveolin-1 was detected at the adherens junction and, to a lesser extent, at the tight junction. They also noticed that the co-localization pattern is disturbed after pharmacological intervention resulting in loss of the tight junction barrier (Figure 3D) (Marchiando et al., 2010). Colocalization immuno-EM was also a crucial method to study the area compositae at the intercalated disc of cardiac muscle cells. Using this technique, researchers identified the mixed type of junction where proteins previously belonging to one type of junction are seen together, such as plakoglobin (Franke et al., 2006).

An alternative method to localize junctional proteins in EM by gold-conjugated antibodies, is the use of genetic tags that generate EM contrast on a specific protein or subcellular structure of interest. This has the advantages that it does not require permeabilizing treatments and is a robust method. Examples are APEX2 and mini-SOG. Specifically, APEX (enhanced ascorbate peroxidase) is an enzyme that catalyzes the H₂O₂-dependent polymerization of DAB into an insoluble polymer. This DAB polymer is osmiophilic, which becomes EM-visible after treatment with OsO₄. APEX does not need any light, has minimal diffusion of the reaction product, works in presence of glutaraldehyde fixation and yields excellent preservation of the ultrastructure (Martell et al., 2017). In a recent study, researchers used APEX2 (a better and more recently developed APEX mutant) with EM imaging in combination with quantitative proximity proteomics. It revealed the molecular and spatial organization of the apical junction complex (mix of tight and adherens proteins) and the junction-associated polarity network in fully polarized epithelial cells. The researchers showed using APEX labelling that the Crumbs complex (module) localizes to a distinct cortical region just apical of tight junctions (Tan et al., 2020). MiniSOG contains 106 amino acids, less than half the size of green fluorescence protein. Illumination of miniSOG generates sufficient singlet oxygen to locally catalyze the polymerization of diaminobenzidine into an osmiophilic reaction product resolvable by EM (Shu et al., 2011).

Immuno-EM has the main advantage that it can label individual proteins and see the ultrastructure of the sample at a nanometer resolution. Even though it is a destructive technique that needs sample optimization, it is accessible to many researchers. Immuno-EM is, together with super-resolution microscopy, often used to confirm protein (co)localization and interactions within and around the intercellular junction of interest. The technique that researchers use depends on the availability of the microscope.