Imaging interorganelle contacts at a glance

The evolution of membrane-bound organelles in eukaryotic cells allows for spatial separation of incompatible biochemical processes. Nevertheless, organelles must exchange materials and information for a cell to function as a coordinated unit. Over the past 20 years, there has been increasing interest in the role of membrane contact sites (MCSs) – sites of close apposition between two organelles – in intracellular communication. MCSs function in the exchange of ions, lipids and, in some cases, even proteins (Prinz et al., 2020). MCSs are also sites of organelle biogenesis: peroxisomes, lipid droplets and autophagosomes all arise from specialized endoplasmic reticulum (ER) subdomains (Joshi et al., 2017). Finally, organelle division can occur at MCSs; the ER has been implicated in fission of both mitochondria and endosomes, and lysosomes also regulate mitochondrial division (Wu et al., 2018). These findings highlight the growing importance of methods to visualize MCS structure and dynamics.

MCSs are typically defined as organelle membranes located within 10–80 nm from each other and tethered by one or more proteins in the absence of membrane fusion (Gatta and Levine, 2017), although some MCSs can occur at distances as large as 400 nm (Ping et al., 2016) (see poster). Because the distance between organelles is typically below the diffraction limit of light, visualizing MCSs has been challenging. Electron microscopy (EM) was the first method used to visualize MCSs (Bernhard and Rouiller, 1956) and remains the ‘gold standard’ for identifying bona fide contacts between organelles. However, EM is limited to use in fixed cells, and thus cannot be used to visualize MCS dynamics.

Recently, a variety of approaches that address these challenges have emerged. Each method has pros and cons making them uniquely suited to studying the structure and/or dynamics of MCSs in different contexts. In this Cell Science at a Glance article and the accompanying poster, we review methods for visualizing interorganelle MCSs. These include diffraction-limited approaches, such as widefield, confocal and multispectral imaging, proximity-based methods, light engineering and single-molecule localization super-resolution imaging methods, and two-dimensional (2D) and three-dimensional (3D) EM approaches.

Widefield and confocal microscopes have resolutions of ∼200–250 nm laterally and ∼500 nm axially, well above the typical dimensions of MCSs (see Box 1). Nevertheless, diffraction-limited approaches provide valuable information about the spatial organization of organelles (see poster and Box 2). Organelle proximity, which often correlates with the number and extent of MCSs, can be visualized by diffraction-limited microscopy and helps generate hypotheses about MCSs and organelle communication that can be further explored using other methods. These modalities are also widely available and compatible for use with live or fixed cells and with a variety of fluorophores. Widefield fluorescence microscopy is extremely sensitive and is frequently used for studies of small cells, such as yeast, where signals from endogenous proteins can be dim (Elbaz-Alon et al., 2014; Wozny et al., 2023). Additionally, fluorescence intensity can provide information on the number of molecules within MCSs; for example, widefield microscopy has been used to count the number of molecules for ER–mitochondria encounter structure (ERMES) components per MCS (Wozny et al., 2023). This approach has revealed that ER–mitochondria MCSs contain similar numbers of the yeast ERMES components Mmm1, Mdm12 and Mdm34.

For larger samples, including polarized mammalian cells, organoids and tissues, confocal microscopy provides useful optical sectioning capabilities. Point-scanning confocal microscopes can selectively target a laser to a region of interest (ROI), which makes them ideal for photomanipulation experiments. Photobleaching has been used to infer changes in the rate of transfer of a fluorescent fatty acid analog from the ER to lipid droplets in cells depleted of the lipid droplet biogenesis protein seipin (Salo et al., 2019). Several other variations of confocal microscopy are especially useful for studying MCSs, including multispectral imaging (MSI) and image scanning microscopy (ISM) (see poster). MSI uses multiple lasers simultaneously to excite fluorophores, combined with many spectral detector elements to collect an entire spectrum within each pixel in an image. Linear unmixing algorithms, which assume that a spectrum is formed by the linear combination of signal from each fluorophore detected in a pixel, analyze the contribution from each fluorophore to the total signal and computationally distinguish fluorophores with overlapping emission spectra (Cohen et al., 2018). Alternatively, fluorophores can be excited sequentially and unmixed based on their excitation spectra. We previously used MSI to visualize six organelle types simultaneously in live cells, revealing the ‘organelle interactome’ (Valm et al., 2017) (Box 2). Recently, MSI has revealed that functional multi-organelle units of mitochondria, ER, peroxisomes and lipid droplets regulate inflammatory lipid metabolism in macrophages (Zimmermann et al., 2024). Airyscan microscopy is a version of ISM that, like MSI, uses a multi-detector approach to increase detection sensitivity and resolution beyond that of standard confocal microscopy (see Box 1). Airyscan microscopy has been used to study mitochondria–lysosome contacts in sensory peripheral neurons expressing proteins with mutations associated with Charcot–Marie–Tooth disease (Wong et al., 2023). In addition, Airyscan complements single-molecule tracking methods and has recently been used to provide overall structural context of the ER and mitochondria while observing dynamics of ER–mitochondria tethering proteins (Obara et al., 2024).

Diffraction-limited microscopy methods cannot determine whether two organelles are physically touching and should generally be combined with other approaches, such as timelapse imaging. If two organelles are observed to co-traffic for extended periods of time, they are more likely to be in physical association. For example, confocal live-cell imaging has revealed that the ER makes extended contact with motile mitochondria and endosomes (Friedman et al., 2010), and widefield microscopy has revealed that peroxisomes, lipid droplets and ER can ‘hitchhike’ on early endosomes in a fungal model system (Guimaraes et al., 2015; Salogiannis et al., 2016). Both widefield and confocal microscopy can be used to detect the probes described in the proximity-based methods section below. Putative MCSs observed by diffraction-limited microscopy can also be further validated using super-resolution or EM techniques.

Proximity-based methods have revolutionized the study of MCSs by leveraging biochemical interactions to enable the visualization of interactions between proteins and organelles with high specificity and sensitivity (Box 2). Such approaches can be used in combination with many imaging techniques and offer a resolution independent of the limitations of diffraction-limited microscopy. These methods include approaches based on the transfer of energy between molecules in close proximity, the dimerization or complementation of fluorescent proteins and approaches that detect antibodies targeted to different organelles only when they are in contact. An important caveat when using genetically encoded probes is that overexpression of reporter proteins can potentially alter the dynamics of MCSs, distort their structure or even artificially induce contacts, resulting in artifacts and false positives. Nevertheless, when used correctly, proximity-based systems minimally perturb the endogenous dynamics of organelles, allowing for accurate observation of natural processes at MCSs. Some of these techniques are appropriate for use in live cells, whereas others only work in chemically fixed cells. Additionally, some are low affinity and easily reversible, allowing the visualization of rapid MCS dynamics, whereas others are high affinity, and therefore less reversible, meaning they are capable of providing a snapshot of MCSs in time but less suitable for studying MCS dynamics over short time scales.

Fluorescence resonance energy transfer (FRET) relies on overlapping excitation and emission spectra of fluorescent proteins to detect protein–protein interactions. When sufficiently close (1–10 nm), an excited donor molecule can transfer energy to an acceptor, and this non-radiative transfer can be detected as emitted fluorescence (see poster). FRET is both non-destructive and non-invasive; thus, by fusing fluorescent proteins to membrane proteins or tethers, this technique can be used in living cells (Rizzuto et al., 1996; Zimmer, 2002). FRET has been applied to study MCSs between various organelles (Marx, 2017; Pietraszewska-Bogiel and Gadella, 2011), including ER–mitochondria (Csordás et al., 2010; Naon et al., 2016), mitochondria–lysosome (Wong et al., 2018), ER–plasma membrane (Poteser et al., 2016) and ER–trans-Golgi network contacts (Venditti et al., 2019), demonstrating its broad applicability. Moreover, combination of FRET with other techniques can enable detailed explorations of mechanisms at MCSs and temporal dynamics of membrane proximity. Noteworthy examples are its combination with total internal reflection microscopy to study ER–plasma membrane MCSs (Chang et al., 2018; Chen et al., 2017; Poteser et al., 2016) and its combination with fluorescence lifetime imaging to explore intracellular Ca²⁺ levels at ER–mitochondria contacts (Bastiaens and Squire, 1999; Elangovan et al., 2002; Miyawaki et al., 1997; Romoser et al., 1997; Sekar and Periasamy, 2003). Despite its functionality, FRET has several drawbacks. Importantly, the efficiency of energy transfer is influenced by the orientation and distance between donor–acceptor pairs and their spectral overlap. This can lead to high background fluorescence and low signal-to-noise ratio, complicating detection. Moreover, effective FRET requires equimolar expression of donor and acceptor pairs. A FRET-based probe that expresses donor and acceptor fluorophores as a single mRNA with a self-cleavable TAV2a sequence between them has recently been developed to overcome this issue (Naon et al., 2016).

Bioluminescence resonance energy transfer (BRET) is a variant of FRET that uses a bioluminescent luciferase enzyme as the donor and a fluorescent protein as the acceptor (Pfleger and Eidne, 2006). Luciferase catalyzes a bioluminescent oxidation reaction and energy is transferred to the acceptor if protein–protein interaction occurs (see poster). Unlike FRET, BRET does not require donor excitation, which reduces phototoxicity when applied in living cells. Combined with a genetically encoded Ca²⁺ indicator, BRET has been used for ratiometric measurement of inter-organellar Ca²⁺ dynamics (Cho et al., 2023). Another example is ‘mitochondria–ER length indicator nanosensor’ (MERLIN), which uses ER-linked Renilla Luciferase 8 (RLuc) as the donor and mitochondria-linked mVenus as the acceptor. MERLIN has been utilized to analyze distances between the ER and mitochondria with several modular linkers spanning distances between 0 and 24 nm, maintaining optimal signal-to-noise ratios without forcing artificial interorganelle contacts (Hertlein et al., 2020). A few technical aspects need to be considered when using BRET. Environmental factors, such as pH and ATP levels, can affect luciferase activity, and changes in environmental conditions can lead to inconsistent signals. Moreover, BRET requires the addition of luciferase substrate, which can potentially introduce cell-to-cell variability in signal.

Dimerization-dependent fluorescent proteins (ddFPs) are genetically encoded reporters that can detect protein–protein interactions in living cells through the low affinity binding of two protein monomers, one that is weakly fluorescent and one that is non-fluorescent. When the monomers come into close proximity, a brightly fluorescent heterodimeric complex is formed (see poster) (Ding et al., 2015). The original ddFP system used a homodimeric variant of the red fluorescent protein DsRed; green and yellow ddFPs have subsequently been developed, offering a range of emission spectra and improved brightness (Alford et al., 2012a). The ddFP system leverages non-covalent interactions between monomers, making it suitable for studying rapid MCS dynamics without artificially inducing contacts. ddGFP has been used to tag the ER membrane and outer mitochondrial membrane, producing a stronger fluorescence signal when the distance between these membranes is less than 20 nm (Alford et al., 2012b). Additionally, we recently generated a toolkit to visualize MCSs between various organelles including lipid droplets, ER, mitochondria, peroxisomes, lysosomes, plasma membrane and caveolae, further expanding the molecular ‘palette’ for studying MCSs (Miner et al., 2024). In this palette, red and green ‘A’ monomers can each dimerize with the non-fluorescent ‘B’ monomers; thus, ddFPs can be used to simultaneously visualize two different MCSs that share one organelle. Despite the advantages of this low-affinity system, the fluorescence signal produced by ddFPs is generally lower than that produced by methods such as bimolecular fluorescence complementation (BiFC, discussed below), which can limit its sensitivity.

Bimolecular fluorescence complementation (BiFC) utilizes super-folder fluorescent proteins, which are more efficiently and constitutively fluorescent, that are split into two non-fluorescent fragments targeted to separate organelles. For example, green fluorescent protein (GFP) can be split into GFP1–10 (residues 1–214) and GFP11 (residues 214–230). When these fragments come together at MCSs, the GFP reassembles and the resulting fluorescence signal is detectable as distinct puncta (see poster) (Cabantous et al., 2005; Cieri et al., 2018; Magliery et al., 2005; Shekhawat and Ghosh, 2011). Engineered GFP mutants, such as enhanced GFP (eGFP) and Venus, are often employed in BiFC, as are red variants such as mCherry and FusionRed (Lahiri et al., 2014; Toulmay and Prinz, 2012). Because split fluorescent proteins are available in multiple colors, BiFC can also be used to image two contacts that share one organelle (Vallese et al., 2020). Split-Venus has been used in yeast to systematically detect MCSs among ER, mitochondria, peroxisomes, vacuoles, lipid droplets, and plasma membrane (Shai et al., 2018). BiFC systems have also helped uncover new MCSs and describe the remodeling of ER-mitochondria contacts under different cellular conditions (Lahiri et al., 2014; Shai et al., 2018; Vallese et al., 2020). Systems that add flexible spacers, such as SPLICS short (8-10 nm) and SPLICS long (40-50 nm), can additionally allow determination of distances between membranes at contact sites (Cieri et al., 2018). Advantages of BiFC are that it produces easily quantifiable fluorescence signals and is less expensive and time-consuming compared to other proximity-based techniques. However, a drawback is that overexpression of high-affinity split fluorescent proteins can artificially induce MCSs, potentially generating false-positive identification of MCSs or distorting normal organelle morphology (Bishop et al., 2019; Tashiro et al., 2020). Expression levels should be carefully considered when studying dynamic processes and properly optimized. One recent study tackled this issue by generating split fluorescent probes with low self-affinity that can be reversibly associated by the addition and subsequent washout of a fluorogen compound (see poster) (Li et al., 2024).

Finally, the proximity ligation assay (PLA) is a unique method designed to detect interactions between two proteins in close proximity, typically within a maximum of 40 nm (Benhammouda et al., 2021; Fredriksson et al., 2002). This technique uses single-stranded oligonucleotides conjugated to antibodies that bind the target proteins. When the targets are in proximity, the oligonucleotides are ligated to form a template for rolling-circle amplification (RCA), producing long single-stranded DNA that is then detected by hybridization with fluorophore-labeled oligonucleotides (see poster) (Fredriksson et al., 2002; Schallmeiner et al., 2007; Soderberg et al., 2008). This generates a discrete fluorescence signal indicating the location of the protein–protein interaction. PLA can be used to target probes to specific tether proteins at MCSs. For example, the ER–mitochondria interface was visualized by probing the interaction between voltage-dependent anion channel 1 (VDAC1) in the outer mitochondrial membrane and inositol 1,4,5-triphosphate receptor (IP3R; also known as ITPR1) at ER–mitochondria MCSs (Atakpa et al., 2018; Gomez-Suaga et al., 2017; Tubbs and Rieusset, 2016). This system has also been used to study the role of the ER–mitochondria tethering complex VAPB–PTPIP51 (PTPIP51 is also known as RMDN3) in regulation of autophagy (Gomez-Suaga et al., 2017) and to demonstrate dysfunctional ER–mitochondria contacts in motor neurons in a mouse model of amyotrophic lateral sclerosis with frontotemporal dementia (ALS-FTD) (Stoica et al., 2016). PLA has also been used to identify clusters of IP3R at ER–lysosome MCSs, which facilitate local Ca²⁺ delivery from the ER to lysosomes (Atakpa et al., 2018). The main advantage of PLA is that it does not require protein overexpression, enabling the direct visualization of endogenous proteins in their natural context. Although PLA can detect low-abundance proteins and transient interactions, its use can be limited by the availability of the antibody pairs, and cross-reactivity can lead to false-positive signals. In addition, PLA requires chemical fixation and is thus incompatible with live-cell imaging.

Super-resolution microscopy overcomes the resolution limitation of light microscopy, allowing accurate visualization of the dynamics and distribution of MCSs and the proteins within them (Box 1). Structured illumination microscopy (SIM), stimulated emission depletion (STED) and single-molecule localization microscopy (SMLM) approaches have each provided unique insights into MCSs.

SIM is compatible with conventional fluorophores and can be integrated with a range of complementary methods to derive additional information about MCSs, such as their structural and temporal dynamics (see poster). Offering fast imaging speed and low excitation intensity, SIM enables high-throughput and full-field live-cell nanoscale imaging. Grazing incidence SIM (GI-SIM), which uses incomplete internal reflection of the excitation light to improve the signal-to-noise ratio, has been applied to resolve fine ER structures and image organelle interactions of highly dynamic ER tubules. These structures were found to regulate mitochondrial fission and fusion via ER–mitochondria MCSs and to facilitate hitchhiking interactions using molecular motors at ER–organelle contacts, enabling organelle reshaping and rearrangement (Guo et al., 2018). GI-SIM has also helped visualize three-way contact sites between ER, mitochondria and endosomes mediated by the tethering protein PDZ domain containing 8 (PDZD8) (Elbaz-Alon et al., 2020). 3D-SIM captures fine details in all three spatial dimensions and is excellent for comprehensive structural studies of MCSs but is limited by imaging speed compared to 2D-SIM. 3D-SIM has been used to identify ER–mitochondria protein tethers (Swayne et al., 2011; Van Alstyne et al., 2018) and ER–endosome MCSs (Raiborg et al., 2015). Dual-color SIM, which provides precise colocalization information for two different fluorescent labels with simultaneous super-resolution imaging, has also been applied to image the dynamics of mitofusin 2-driven contacts between mitochondria and ER (Gottschalk et al., 2022) and different types of lysosome–mitochondria MCS interactions (Han et al., 2017).

3D volumetric STED imaging provides outstanding lateral and axial resolution, surpassing that of 3D-SIM (see poster). This approach has been used to distinguish regulatory mechanisms at mitochondria–ER contacts (Cardoen et al., 2024). Although STED reaches lateral resolutions high enough for structural analysis of MCSs, several factors limit its capability for studying MCS dynamics. First, not all fluorophores are efficiently depleted by the STED laser, and STED-compatible dyes must be membrane-permeable, restricting the range of suitable fluorescent dyes. Furthermore, imaging highly dynamic organelle processes with STED is challenging because of the phototoxicity arising from the high-power laser and the longer image acquisition time required. However, several studies have overcome these limitations by using different labeling strategies for multi-color live-cell STED imaging to track ER–mitochondrial MCS dynamics (Bottanelli et al., 2016; Liu et al., 2022).

SMLM approaches include stochastic optical reconstruction microscopy (STORM) and photoactivated localization microscopy (PALM) (see poster). These two related approaches differ in the types of fluorophores used. STORM uses small, fluorescent organic dyes (e.g. cyanines, rhodamines and Alexa Fluor 647) paired with specialized oxygen-scavenging imaging buffers to achieve stochastic photoactivation or photoswitching of fluorescent dyes between ‘on’ (bright) and ‘off’ (dark) states (a behavior referred to as ‘blinking’), which allows a sparse subset of fluorophores to emit fluorescence at a given time. In contrast, PALM exploits photoactivatable (e.g. PA-GFP, PA-mCherry), photoconvertible (fluorescent proteins that already emit fluorescence in non-converted state and change emission properties irreversibly upon light exposure; e.g. mEOS) or photoswitchable (fluorophores that can be reversibly toggled between different fluorescent states; e.g. Dronpa) fluorescent proteins. By tracking large numbers of single molecules in live cells, two-color STORM has revealed the ultrastructural dynamics of mitochondrial fusion and fission and ER remodeling (Shim et al., 2012). An analogous experimental strategy is single-particle-tracking PALM (spt-PALM), which has been used for high-speed molecular tracking of the ER–mitochondria tether protein VAMP-associated protein B (VAPB) to demonstrate that ER–mitochondrial MCSs exhibit exceptional plasticity compared to pseudo-stable ER–plasma membrane MCSs (Obara et al., 2024). SMLM achieves the highest resolution currently available among the super-resolution methods, providing up to 20-nm lateral and 50-nm axial resolutions, but sacrifices temporal resolution, because thousands of frames are needed to reconstruct a single super-resolved image, restricting its ability to analyze MCS dynamics in real-time. Moreover, complex sample preparation, including the use of specialized fluorophores and imaging media and intense data post-processing requirements, pose additional technical challenges, making this approach less accessible compared to other super-resolution methods.

EM has long been considered the ‘gold standard’ for visualizing the ultrastructure of MCSs in fixed cells. Multiple studies, as discussed below, highlight the strength of EM for exploring the nanoscale structural details of MCSs without the need for fluorescent labeling. However, EM approaches cannot provide information about dynamic changes of MCSs over time. Several significant technical challenges related to sample preparation, low-throughput capacity and the requirement for specialized equipment and expertise also limit its general accessibility to researchers. Conventional EM requires chemical fixation and dehydration that can damage specimens, potentially affecting the native architecture of MCSs. In recent years, EM techniques using cryofixation to preserve samples in their natural state have gained popularity for studying MCSs because they allow close-to-life visualization of inter-organelle ultrastructure in a native cellular context. Here, we will overview conventional and cryo-EM approaches that can be used to examine different structural aspects of MCSs.

Transmission EM (TEM) uses an electron beam passed through a thin (50–200 nm) specimen to generate high-resolution images of structures spanning atomic to micron scales (see poster). Despite the unparalleled spatial resolution offered by TEM, it is low throughput, can only be applied to very thin specimens and fails to detect rare contact sites unless combined with other approaches, such as correlative light EM (CLEM). CLEM initially locates an ROI with light microscopy based on fluorescent labels and then uses TEM to acquire ultrastructural information from the same sample, making it especially useful for detecting rare or transient MCSs. TEM can also be combined with immunogold labeling to localize specific MCS proteins while still providing rich structural information. Although TEM enables assessment of qualitative and quantitative features of MCSs, such as the distance between membranes or area of contact, it is unable to capture the entire extent of contact sites because its output is limited to a single plane.

By contrast, electron tomography (ET) and focused ion beam-scanning EM (FIB-SEM) allow visualization of larger 3D volumes of specimens at high resolution and can therefore facilitate 3D reconstructions of MCSs; however, these require specialized expertise and/or equipment and are computationally intensive. In ET, a series of TEM images of a specimen at different tilt angles are acquired and computationally aligned to reconstruct a 3D volume (see poster). ET is thus also limited by the thickness of the sample and requires serial sectioning, which can be laborious and technically challenging. In addition, the practical limitations of tilt range for ET imaging can produce artifacts in the reconstructed tomogram. Thus, ET is most suited for investigating MCSs at the periphery of cells, such as ER–plasma membrane MCSs (Hoffmann et al., 2019).

In FIB-SEM, a highly focused ion beam is applied to sequentially mill away thin layers of the surface of a sample while a scanning electron microscope simultaneously images the newly exposed surface, enabling reconstruction of a 3D tomogram from a stack of 2D images (see poster). FIB-SEM allows for imaging of samples with a wider range of thicknesses compared to TEM or ET and is suitable for the study of large-volume MCSs within the interior of cells. For example, MCSs between the ER and multiple other organelles have been systematically investigated by FIB-SEM in murine neurons (Wu et al., 2017). Recently, deep learning-based methods for automatic segmentation were trained using whole-cell FIB-SEM images of 35 distinct cellular organelles, offering comprehensive quantitative insights and facilitating a more integrated understanding of organelle interaction (Heinrich et al., 2021).

Finally, in situ cryo-ET and cryo-FIB-SEM allow for the direct visualization of MCSs within the intact cellular context, revealing the structural arrangement of organelle interactions and insights into their physiological roles. For example, cryo-ET has been used to obtain the 3D ultrastructure of ER–plasma membrane and ER–mitochondria MCSs (Fernandez-Busnadiego et al., 2015). These techniques have also helped examine the structure and function of proteins found at MCSs. In situ structural analysis of autophagic membranes suggests that yeast autophagy-related protein 2 (ATG2) bridges the distance of phagophore–ER contacts to facilitate lipid transfer (Bieber et al., 2022). Similarly, rod-shaped bridges formed by vacuolar protein sorting 13 homolog C (VPS13C) at ER–endosome or lysosome contacts have been revealed by in situ cryo-FIB-SEM and cryo-ET, supporting a model of VPS13C-mediated lipid transfer at MCSs (Cai et al., 2022). Recently, the combined use of in situ cryo-CLEM, cryo-FIB-SEM and cryo-ET has been used to build a comprehensive molecular organization model of the ERMES protein complex across contact sites in yeast (Wozny et al., 2023).

In recent years, a variety of tools have advanced the visualization and quantification of the structure, composition and dynamics of MCSs in multiple ways. Diffraction-limited and super-resolution microscopy approaches allow for quantification of organelle proximity and co-tracking of organelles in timelapse imaging experiments. EM allows for the measurement of precise distances between organelles, and EM as well as MSI also facilitate characterization of organelle interactomes among multiple interorganelle contacts. Finally, proximity-based methods allow for the quantification of the number and extent of MCSs and can act as length sensors if appropriately designed. Proximity-based methods can also be used for high-throughput screens to identify proteins that modulate MCSs (e.g. Castro et al., 2022; Shai et al., 2018). Each of these tools has advantages and disadvantages and, generally, multiple approaches must be used in tandem to comprehensively characterize MCSs. However, it is important to note that none of the methods described here test MCS function. Even EM approaches offering the highest available resolution do not provide information about whether a MCS is active in the exchange of material. In some cases, MCS function can be gated by post-translational modifications of tether proteins, such as phosphorylation (Kors et al., 2022; Miner et al., 2023; Prinz et al., 2020) or redox state (Yoboue et al., 2018). Thus, tools for visualizing MCSs should be combined with functional assays that test for the active exchange of ions, lipids or proteins. Alternatively, functional assays that interrogate the role of MCSs in organelle biogenesis or division could be employed. By combining such functional assays with the current and growing toolbox for visualizing interorganelle contact sites, we anticipate many exciting new discoveries in this field in the years to come.

We thank Dr Christopher Obara and Dr Aubrey Weigel for providing editable images and for helpful feedback.