Keywords

1 Introduction

Mitochondria are key organelles in eukaryotic cells that carry out essential functions in cellular metabolism, signaling, and the execution of cell death. The central role of mitochondrial function is demonstrated by the multiple disorders associated with dysfunction of this organelle. Mitochondria present a complex structure, composed by a double membrane system: the mitochondrial outer membrane (MOM) and the mitochondrial inner membrane (MIM), separated by the intermembrane space and with the MIM enclosing the mitochondrial matrix. This sophisticated organization allows for compartmentalization of the different mitochondrial functions. Furthermore, mitochondria adopt a tubular structure that organizes into a dynamic network in the cytosol, due to continuous fission and fusion of the organelle, as well as transport along the microtubule cytoskeleton. Importantly, mitochondrial structure, dynamics, and intracellular localization are important for the organelle function and their alteration has also been linked with disease.

During the last decades, fluorescence light microscopy has contributed to advance our understanding of the relationship between mitochondrial dynamic organization and function. More recently, the developments of single-molecule fluorescence methods have pushed even further the experimental approaches available to investigate mitochondrial biology by making accessible the analysis of the structural organization and dynamic behavior of mitochondrial components with unparalleled temporal and spatial resolution. Here we discuss the principles and applications of single-molecule microscopy to mitochondrial research.

In order to study mitochondrial processes at the single-molecule level using fluorescence microscopy, fluorescently labeled mitochondrial proteins have to be first localized. This implies the use of a microscopy technique that reveals the localization of individual proteins below the diffraction limit, given by Abbe’s law:

$$ d=\frac{\lambda }{2\mathrm{NA}}=\frac{\lambda }{2\left(n\times \sin \alpha \right)} $$

here λ is the wavelength of light (monochromatic excitation of the fluorophore), n is the refractive index of the imaging medium and NA is the numerical aperture of the used objective.

One such technique is total internal reflection fluorescence (TIRF) microscopy, which is based on the principle of total internal reflection (TIR). TIR only works when the illumination laser encounters a boundary to a medium in the sample with a lower refractive index. Here, the refractive behavior of the excitation laser beam is described by Snell’s Law:

$$ {n}_1\sin {\theta}_1={n}_2\sin {\theta}_2 $$

Where, n1 is the higher refractive index and n2 the lower refractive index and θ is the angle of the excitation laser in medium 1 or 2. The angle θ of the excitation laser beam must be larger than the critical angle to be totally internal reflected. Total internal reflection of the excitation laser, in an inverted microscope produces an evanescent field/wave, exciting the fluorochromes in close proximity to the coverslip. The intensity I(z) of the evanescent field in the sample decays exponentially along the x-axis by the following equation:

$$ I(z)=I(0){e}^{-z/d} $$

Here, d is the distance of the fluorochrome to the coverslip. Furthermore, the penetration depth dp of the evanescent field can be calculated as following:

$$ {d}_p=\frac{\lambda }{4\pi }\ {\left({n}_1^2{\sin}^2\theta -{n}_2^2\right)}^{-1/2}, $$

where λ is the wavelength of the excitation laser, the angle θ of the illumination and the refractive index of the coverslip/glass n1 and the cultivation medium of the sample n2. Thus, TIRF microscopy limits the excitation of fluorescent molecules to about 200 nm along the axial axis of the sample, above the coverslip and into the sample. This lowers the background fluorescence resulting from out of focus, because the excitation only takes place in the evanescent field. Thus, the signal-to-background ratio (SBR) is increased compared to e.g. the Epi-fluorescence, resulting in a better identification of individual fluorophores as long as the single fluorescent signals fulfill the Rayleigh criterion. If the illumination angle is equal or larger than the critical angle, the microscope setup is used in the highly inclined optical (HILO) sheet mode [1] and if the illumination light follows the axial axis of the objective, epi-fluorescence is used. It is important to note that most of the mitochondrial network in a cell is above the evanescent field. Thus, the HILO illumination (also called pseudo-TIRF) is the illumination of choice to provide optimal SBR for single-molecule studies of mitochondrial processes resulting in a bright fluorescent signal with a low background level.

Today, most TIRF microscopes are inverted microscopes, and the excitation laser beam must be focused to the back focal plane of the TIRF objective. Once the fluorophores in the sample are excited, they emit photons. The fluorescent light is collected by the TIRF objective and passes optical filters and is guided to a highly sensitive EMCCD or sCMOS camera installed in the emission light path of the objective. Important to note is that the pixel size of a camera used for single-molecule localization fulfills the Nyquist-Shannon theorem [2, 3]. If single fluorescent molecules are excited in the sample and if their fluorescence is bright enough to be distinguishable from the noise of the camera and the background of the sample, the light of the fluorescent molecules can be recorded as single signals by the camera.

Therefore, bright fluorochromes are the fluorophore of choice in single-molecule microscopy. The brightness of a fluorochrome can be calculated as follows:

$$ \mathrm{brightness}=\varepsilon \times \Phi, $$

where ε is the extinction coefficient of the fluorochrome and Ф is its quantum yield. Today many fluorescent molecules can be used in order to perform single-molecule microscopy at mitochondria. Nevertheless, fluorescent dyes are in most cases brighter than fluorescent proteins. Thus, choosing the best fluorescent label possible is fundamental for single-molecule localization microscopy (SMLM). In order to record single fluorescent signals of individual molecules, several different techniques can be used, here we will discuss the most common once: (1) fluorescent photoactivation localization microscopy fPALM/PALM [4,5,6], (2) stochastic optical reconstruction microscopy (STORM) [7,8,9], (3) DNA-Based Point Accumulation for Imaging in Nanoscale Topography (DNA-PAINT) microscopy [10] and (4) if SMLM is done in living cells, also single-particle tracking (SPT) [11, 12] can be performed. fPALM/PLAM uses photoactivatable or photoconvertable fluorescent proteins. STORM exploits the blinking of fluorescent dyes, which can be bound to the protein of interest (POI) via a tag or an anti-/nanobody. In contrast, DNA-PAINT uses the on- and off-state of binding and unbinding of a fluorescently labeled single-stranded DNA, often called the imager strand, to another single-stranded DNA, the handler strand, coupled to the POI. The handler strand in a DNA-PAINT experiment can also be coupled to the POI via a tag or an anti-/nanobody. Here, a very interesting approach is to use the ALFA-tag [13] or tags like SnapTag and HaloTag [14, 15]. Single-molecule localization microscopy is also the basis for SPT, and in this case highly photostable fluorescent tags with minimal dark states are preferred for the tracking of the signal. Important for all these techniques is that, in each frame of a movie, distinguishable individual signals have to be recorded, as already mentioned above. The minimal distance between two single fluorescent signals to be localized as two different singals is described by the Rayleigh criterion dr:

$$ {d}_r=\frac{1.22\lambda }{2\mathrm{NA}}=\frac{0.61\lambda }{n\times \sin \alpha }=\frac{0.61\lambda }{\mathrm{NA}} $$

Additionally, a low uncertainty in the localization of single fluorescent signals is improving the result in all mentioned techniques above. This uncertainty σ can be calculated as following,

$$ \sigma =\sqrt{\frac{s^2}{N}+\frac{a^2/12}{N}+\frac{4\sqrt{\pi {s}^3{b}^2}}{a{N}^2}} $$

where N is the number of gathered photons of the single fluorophore, s is the standard deviation of a Gaussian function for fitting the point spread function (PSF) of the fluorophore, a is the pixel size of the camera used, and b is the background of the image.

2 Single-Molecule Localization Microscopy

Single-molecule localization microscopy is the term used to encompass a number of microscopy methods, including e.g. STORM, PALM, and DNA-PAINT. The basic principle for these stochastic functional techniques is the imaging of scattered groups containing single molecules with the distance between them higher than the Rayleigh criterion. Given that biological samples are densely labeled with fluorescent proteins or dyes, precise localization of all the fluorophores in the sample is not directly accessible with standard illumination strategies due to the signal overlay and an ensemble readout. Single-molecule localization microscopy can help to solve this issue by controlling rate of fluorophors emitting a fluorescent signal or in DNA-PAINT the ON/OFF-state of the fluorophore. Currently, the minimum distance between two fluorophores that is possible to distinguish in the range of 5–10 nm. However, the highest resolution can be obtained only if fluorophores are separated in time and space. Thus, using the correct concentration of the fluorophores and the ON/OFF-switching of the fluorescent signals in the sample and the correct exposure time enables to reveal the biological structure with highest resolution possible.

2.1 PALM and STORM of Mitochondrial Proteins

Super-resolution imaging techniques like PALM and STORM are based on detection and time-resolved localization of randomly excited singe fluorophores [4,5,6, 9]. During a measurement, fluorophores can be photoswitched between a dark state and a fluorescent state [16]. Under certain laser illumination conditions, a subset of the fluorophores can be activated and their emission is recorded (and later used for localization), over many cycles of excitation/emission so that optimally all fluorophores of a biological structure are detected. The subsequent reconstruction of a super-resolution image is obtained in a similar manner with PALM, STORM and DNA-PAINT, by localizing the emitted signal with a 2D-Gaussion fit and reconstruction of the biological structure by combining the localizations of the individual fluorophores (Fig. 1).

Fig. 1
figure 1

Scheme of the STORM approach. (a) left – The illumination protocol excites only a subset of individual fluorophores in the population while others stay in the dark state and the fluorescence is collected in each frame. In the next cycle, another subset of fluorophores turn to the excited state for N number of frames collected. For image analysis, the individual fluorophores in each frame are localized with high accuracy and combined to build the super-resolution image on the right. (b) Representative reconstructed super-resolution image of GFP-Bax in HeLa cells stained with AF647-anti-GFP nanobodies. The image was acquired on fixed cells 3 h after induction of apoptosis. Scale bar, 5 μm. (Figure adapted from: [17])

Full size image

To ensure that only a small subset of fluorophores is switched into the fluorescent state, both the fluorophore photophysical properties and the illumination protocol need to be optimized. Super-resolution microscopy with the accurate localization of single molecules is based on the separation of the signal of individual fluorescent molecules. The light intensity for illumination during the experiment is adjusted in a way that the sparsely activated fluorophores are not closer than the Rayleigh criterion, so that the position of each fluorophore can be determined with an accuracy of up to 1 nm. PALM is based on the use of e.g. photoactivatable GFP, which is normally stable in a dark state, and a fraction of the fluorophores can be switched into the fluorescent conformation by UV light. In STORM, fluorophores with strong blinking properties are used, like Atto647, so that they can be brought into a dark state from which only a small subset of them will be excited in a controlled manner and emit photons.

STORM has been used to reveal the organization of the proapoptotic protein Bax and its assembly into line, arc and ring structures during mitochondrial apoptosis, in situ [17] (Fig. 2). Furthermore, STORM can be upgraded to all three dimensions to perform 3D STORM, by installing a cylindrical lens into the emission light path of a TIRF microscope. 3D STORM of mitochondrial proteins allows the reconstruction of entire mitochondrial networks with the additional information of single mitochondrial proteins and their organization in all three dimensions [8].

Fig. 2
figure 2

STORM image revealing non-random structures of Bax in apoptotic mitochondria. (a) STORM image of Bax in a HeLa cell (dotted line) 3 h after apoptosis, induced by staurosporine. (b) Representative Bax structures on mitochondria during apoptosis, revealing supramolecular assemblies classified into dots, aggregates, lines, double lines, arcs and ring structures. Scale bars: (a) 5 μm, (b) 100 nm. (Figure adapted from: [17])

Full size image

2.2 DNA-Based Point Accumulation for Imaging in Nanoscale Topography (DNA-PAINT)

A different approach for switching the fluorophores between an ON- or OFF-state on the observed target is DNA-PAINT. As with PALM and STORM, DNA-PAINT microscopy is also based on the fact that the centroid of the PSF of a fluorescent signal can be precisely determined. The methods above based on the fluorophore localization use the photophysical switching molecules between excited and relaxed states, negotiated by the chemical reactions or by the photon’s excitation. The DNA-PAINT approach is different. The target in this case is labeled with a single-stranded DNA (ssDNA). The complementary ssDNA labeled with the fluorophore can then quickly hybridize to the docking DNA from the target, thereby leading to efficient labelling of the POI (Fig. 3).

Fig. 3
figure 3

Scheme of the DNA-PAINT approach. (a) left: The OFF-state, a signal is not detected, unbound imager strands carrying a fluorophore are defused away; right: ON state, the imager strand hybridized to the docking side, and the signal is detected. (b) comparison of different labelling probes: antibody, HaloTag, GFP Nanobody and SNAPTag. (c) DNA-PAINT overview image of NUP96-SNAP in U2OS cells. Scale bar: 5 μm. (Figure adapted from: Schlichthaerle et al. 2019)

Full size image

The lengths of the imager strands and the sequence are selected to have a short lifetime in the bound state with the docking strands – a single-stranded DNA extension from the sample. During the binding event – hybridization of docking and imager strands – the fluorophore on the imager strand is exited during ON-time τb (bright) while the signal is recorded by a camera. The association rate kon is calculated from the lifetime τd (dark) via \( \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{${\tau}_d$}\right.={k}_{\mathrm{on}}c \), where c is the concentration of imager strand. The dissociation rate \( {k}_{\mathrm{off}}=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{${\tau}_b\ $}\right. \) depends on the duplex length. Adjusting the temperature, sequence, length and concentration/ratio of a DNA duplex allows optimization of the ON- and OFF-states of the fluorophores, so that they can be excited for a short enough time to avoid bleaching but long enough to record a signal with a high SBR. Since the fluorescent single-molecule signal can be detected and localized only when the fluorophore is located in the focal plane, the effect of blinking fluorescence similar to STORM is created. The unstable duplex allows multiple hybridisation events. Thus, one advantage of DNA-PAINT over STORM and PALM is that bleached fluorophores are exchanged within the sample by non-bleached imager strands, thereby extending the possible measurement time as well as the achieved spatial resolution. Furthermore, multicolour imaging of different proteins of interest can be achieved with dyes of the same wavelength by simply exchanging the imager strand, which are then measured in a sequential manner. Thus the same dye can be used for different proteins by using different DNA duplex constructs. 

Anti- or nanobodies can be attached to ssDNA by modifying antibodies through the usage of click chemistry [18, 19]. This modification opens the possibility to use the DNA-PAINT approach for a variety of targets in the cell, similar to anti- or nanobodies in STORM but with the already mentioned advantages. With the given advantages mentioned above this opens the field for highly multiplexed imaging in biological samples with nanoscale resolution [20]. Important to note is that, immunolabeling for SMLM done with IgG antibodies cause large linkage errors (~10 nm, or ~20 nm for indirect immunolabeling) [21]. These errors are less when using nanobodies.

Recently, DNA-PAINT has been combined with SPT, which allows monitoring of single molecules for extended durations [22]. However, the use for mitochondrial processes has not yet been shown and will be a challenge for the future.

3 Single-Particle Tracking (SPT)

SPT is based on the single-molecule localization of individual fluorescent signals and in their reidentification in the following frames of a recorded movie, in order to track a fluorescent signal over time (Fig. 4a–c) [23,24,25,26,27,28,29]. In contrast to fPALM/PALM, STORM and DNA-PAINT, SPT works best with high frame rates, in order to gain a high time resolution, which allows accurate tracking of the POI and thereby reveals the accurate spatio-temporal behavior of the POI (Fig. 4d–f). Using a high frame rate results in less photons per frame of the individual signals, which is counterproductive to the localization of the fluorescent signal. Thus, especially in SPT, bright fluorophores with a high photostability are advantageous. Thereby photostable fluorophores result in long trajectories. This allows to study the spatiotemporal organization of an induvidual POI in different microcompartments. In order to gain single fluorescent signals, the POI can be genetically fused to a photoactivatable protein resulting in sptPALM [26, 30]. Thus, a subset of fluorophores can be activated and their fluorescent signal can be recorded. Another approach is to fuse the mitochondrial POI to a HaloTag or SNAPtag and label only a subset of the tagged proteins in order to gain single signals [23]. An important factor for this method is a posttranslational, substoichiometric labeling of the mitochondrial proteins of interest via a HaloTag or SnapTag with membrane-permeable dyes.

Fig. 4
figure 4

Scheme of single-particle tracking. (a) The fluorescent signals are localized by determining the center of their PSF. (b) Reidentification of localizations of the same POI at different time points results in (c). Chronological series of single particles results in their trajectories (red and blue). (d) A Higher frame rate result in more localizations of a POI. (e/f) High frame rates improve the tracking result of the POI by providing more steps and thus a better resolved spatiotemporal information (purple versus orange)

Full size image

Single-molecule tracking allows the study of mitochondrial proteins in isolated yeast mitochondria [31]. Furthermore, SPT of mitochondrial proteins can also be performed in situ, [23, 32] in correlation with the metabolic activity of mitochondria [33]. Additionally, 3D SPT is possible by engineering a double-helix point spread function (DH-PSF) [34, 35]. Multicolor SPT of different mitochondrial proteins at the same time is also possible [36, 37].

SPT of mitochondrial proteins via this method is done by using the HILO illumination method, which allows illumination of fluorophores above the evanescent field. Nevertheless, SPT of mitochondrial proteins performs best in the cell periphery. Together with the use of bright dyes, a high SBR can be achieved and results in a high localization precision. Thereby SPT allows the study of protein organization and micro-compartmentalization in the different mitochondrial locations [23, 38] (Fig. 5). Mitochondrial proteins of the MOM show a random movement along the MOM, while proteins of the oxidative phosphorylation system (OXPHOS) follow the MIM shape revealing the orientation of cristae. Furthermore, proteins that are part of the mitochondrial inner membrane organization system/mitochondrial contact site complex (MINOS/MICOS), such as Mic60, show a confined diffusion. The same is true for the translocase of inner membrane subunit 23 (Tim23). Especially in the case of proteins of the MIM, SPT of mitochondrial proteins reveals functional micro-compartments of continuous mitochondrial membranes [32, 39].

Fig. 5
figure 5

Single-particle tracking of mitochondrial proteins. (a) Trajectories of the outer mitochondrial membrane proteins human fission factor 1 (hFis1) and translocase of outer membrane subunit 20 (Tom20). (b) Trajectory maps of proteins of the OXPHOS complex: the cytochrome c oxidase subunit CoxVIIIa (CIV) and F1FO ATP synthase subunit-γ (CV). (c) Trajectories of the MIM proteins, translocase of inner membrane subunit 23 (Tim23) and MINOS/MICOS complex protein 60 (Mic60). (Figure adapted from: [38])

Full size image

Besides single-molecule studies of mitochondrial proteins in living cells, SPT can also be done in supported lipid bilayers (SLBs), in vitro. This provides the advantage of studying individual protein–protein interactions directly in model membrane systems. Here, the assembly and regulation of proteins and their interaction partners can be studied [40]. This allows the analysis of the stoichiometry of Bax oligomers and the activation of Bax by cBid and the dissociation of Bax oligomers by Bcl-XL [40]. Furthermore, single lipid organization can be studied in a SLB [41].

4 Summary

Investigations of the mitochondrial proteins and their function to understand mitochondrial processes are a central aspect of cell biology. Recent technological advances in fluorescence microscopy approaches have opened a new range of methods for mitochondrial research. Nevertheless, visualization of protein behavior by single-molecule microscopy remains challenging. Here, a variety of super-resolution microscopy methods enable the protein localization with a precision of ~1 nm [21]. DNA-PAINT, STORM, and fPALM/PALM rely on the accurate localization and the detection of single fluorophores. The localization precision depends on the number of gathered photons, which responsively depends on the brightness of fluorophores and the SBR, as well as on the exposure time used [6, 10, 42]. This implies the use of bright fluorescent labels and long exposure times and a low background, but also an accurate determination of the PSF center. Interestingly, tagging the POIs to a SNAPtag or HaloTag as well as labeling via a nanobody allows the use of organic fluorescent dyes with best performance in STORM, DNA-PAINT, SPT, or even MINFLUX [43, 44]. Furthermore, in STORM and MINFLUX blinking behavior of the used dye can be adjusted and optimized. Comparably, in DNA-PAINT the ON and OFF rate of the imager strand can be changed by using different DNA-sequences. In addition, DNA-PAINT depends on the binding and unbinding of the imager strand and therefore allows long-term imaging of a sample, regardless of bleaching of single imager strands. Beside its possibility to perform highly multiplexed imaging, this is one of the main advantages of DNA-PAINT experiments as it results in a high resolution of a biological structure.

However, revealing the protein organization via STORM and DNA-PAINT is limited to fixed cells. In contrast, SPT is a powerful tool to visualize the spatio-temporal dynamics of mitochondrial processes and the involved proteins in living cells. SPT allows the study of the dynamics of individual proteins over time, while techniques like fluorescent recovery after photobleaching (FRAP) and fluorescent correlation spectroscopy (FCS) can only provide average information about the observed area of the cell and millisecond temporal resolution [45]. Nevertheless, a key aspect is that SPT has to be correlated with cellular processes such as apoptosis, metabolism, membrane dynamics, or signaling pathways, in order to answer biological questions of mitochondrial processes and proteins or lipids involved in it. An interesting option here is to use Dual-Color SPT in order to reveal co-diffusion of proteins as it was done with interferon receptors in the plasma membrane [46]. Thus, the interaction of mitochondrial proteins and their interplay with mitochondrial processes could possibly reveal in living cells.

The combination of different techniques, like STED-PAINT, opens new possibilities to increase the image resolution of fixed cells [47]. Another imaging approach combining two super-resolution techniques is called MINFLUX. This method provides three-dimensional resolution at the size of the fluorescent molecule. The “torus” or “doughnut” shaped excitation beam can be used in fixed or living cells. It creates a movable excitation beam featuring an intensity minimum. The fluorophores are then switched individually like in STORM or fPALM. It was demonstrated that fast three-dimension imaging and simultaneous two-color registration is possible, which is interesting for live cell applications [48].

Taken together, when planning the study of proteins involved in mitochondrial processes, it is important to determine what the specific imaging requirements are before choosing which microscopy method to use. To study protein organization with super-resolution at the single-molecule level, STORM and DNA-PAINT are still the methods of choice, since both techniques reveal protein complexes with high resolution. It is best to compare the performance of both methods in a protein complex of the nucleus – the nuclear pore complex (NPC) [49] consisting of around 30 different proteins called the nucleoporins (NUPs) [50]. Here, DNA-PAINT and STORM have successfully revealed the octameric structure of the single NUPs [14, 15, 51]. DNA-PAINT can reach tens of thousands of photons per blinking event, which is hard to achieve using STORM [14, 15]. In contrast, STORM is so far the only method that revealed the 3D architecture of the NPC [49, 51, 52]. Theoretically, this should also be possible with DNA-PAINT.

Interestingly, MINFLUX microscopy is also capable of revealing this octameric organization of NUPs [44]. MINFLUX imaging achieved a lateral localization precision below 3 nm and an axial localization precision below 2 nm. Thus, MINFLUX allows the highest localization precision in all three dimensions compared to other SMLM techniques; DNA-PAINT, STORM, and SPT. However, MINFLUX is still at its infancy and further work will be needed to bring this method to the cell biologist. Recently, it has been shown that dual-color 3D MINFLUX nanoscopy allows to reveal organellar substructures with a 3D localization precision of ∼5 nm in mitochondria, and indicates that multiple Mic60 molecules arrange in a ring-like structure with a diameter of 40–50 nm, suggesting that Mic60 surrounds individual crista junctions [53]. Here, it was demonstrated to what extent MINFLUX will permit a closer look in the field of mitochondrial processes in respect of mitochondrial protein organization. MINFLUX allows to improve the temporal resolution of imaging and the 3D resolution. Thus, it is the most suitable technique to study protein assemblies and to reveal their formation due to cellular processes in all three dimensions via its nanometer-scale isotropic localization precision. Additionally, SPT with MINFLUX is possible [54, 55]. Here, MINFLUX allowed the study of the diffusion of the DPPE-ATTO 647 N lipid at RT in an SLB with a temporal acquisition with a mean of 117 μs per localization. This demonstrated what is technically possible with SPT by MINFLUX. One of the next challenges is to investigate the spatio-temporal dynamics of mitochondrial proteins or lipids by MINFLUX, in order to make use of the big advance in fluorescent microscopy provided by MINFLUX, namely, both the high localization precision and high temporal resolution of MINFLUX and thus to provide knowledge about e.g. the assembly of proteins into multiprotein complexes. Here, it will be of great interest to track multiple signals over time simultaneously.

In contrast to the mentioned techniques, another not yet mentioned technique here is structured illumination microscopy (SIM). SIM was used to reveal cristae dynamics in long-term experiments by using Hessian SIM [56]. Here, even the fusion events of two individual cristae into one was revealed. Hessian SIM allows to use less laser power compared to the conventional SIM resulting in less bleaching of the fluorescent label with high frame rates enabling to use a high spatio-temporal resolution of 88 nm and 188 Hz [56]. Thus, Hessian SIM is a promising microscopy method to investigate the organization of proteins and mitochondrial sub-compartments over time. This can allow to reveal mitochondrial processes with a higher temporal resolution even in longer time windows. The possibility to record mitochondrial processes in long time ranges with high temporal resolutions is important in biological processes in order to understand the full mechanism in living cells.

Finally, one of the next challenges in the coming future will be to determine the formation and dynamic of mitochondrial protein organizations of large macromolecular complexes like the MINOS/MICOS complex, the apoptotic pore, OXPHOS protein complexes and supercomplexes, or the TOM-TIM machinery. Here, the use of STORM, DNA-PAINT, MINFLUX, and SIM in fixed and living cells, in 2D and 3D, hold strong promises to shed new light on the nanoscale organization of mitochondrial proteins in relation to the mitochondrial processes in which they participate, in order to answer new biological relevant questions. Today the competition between the described methods results from their individual advantages in the triangle of spatial resolution, temporal resolution and sensitivity/phototoxicity to answer key questions in cell biology of high biological relevance. Thus, today a combination of different techniques enables to consider mitochondrial processes from all sides is still the best to fully understand the behavior of mitochondrial proteins and structures. Therefore, correlative light-electron microscopy is a very interesting method to reveal mitochondrial membrane structures and the localization of mitochondrial proteins within as well as their influence on mitochondrial substructures with nanometric resolution. Here, the next step can be to combine super-resolution microscopy and electron microscopy to enable a more detailed view on mitochondrial structures and mitochondrial sub-molecular protein organizations.