Snapshots of mitochondrial fission imaged by cryo-scanning transmission electron tomography Open Access

Mitochondria undergo constant remodeling via fission, fusion, extension and degradation. Fission, in particular, depends on the accumulation of mitochondrial fission factor (MFF) and subsequent recruitment of the dynamin-related protein DRP1 (also known as DNM1L). We used cryo-scanning transmission electron tomography (cryo-STET) to investigate mitochondrial morphologies in MFF mutant (MFF^−/−) mouse embryonic fibroblast (MEF) cells in ATP-depleting conditions that normally induce fission. The capability of cryo-STET to image through the cytoplasmic volume to a depth of 1 µm facilitated visualization of intact mitochondria and their surroundings. We imaged changes in mitochondrial morphology and cristae structure, as well as contacts with the endoplasmic reticulum (ER), degradative organelles and the cytoskeleton at stalled fission sites. We found disruption of the outer mitochondrial membrane at contact sites with the ER and degradative organelles at sites of mitophagy. We identified fission sites where the inner mitochondrial membrane is already separated while the outer membrane is still continuous. Although MFF is a general fission factor, these observations demonstrate that mitochondrial fission can proceed to the final stage in its absence. The use of cryo-STET allays concerns about the loss of structures due to sample thinning required for tomography using cryo-transmission electron microscopy.

Obtaining three-dimensional (3D) information on cellular organelles enables advances in cell biology, inspiring the development of various advanced imaging methods (Glancy, 2020). These methods center largely on fluorescence microscopy (FM) due to the possibilities for genetic tagging with fluorescent proteins. Modern techniques bring the resolution of now-routine imaging modalities below 100 nm and into the domain of intracellular organelles (Jakobs and Wurm, 2014; Kleele et al., 2021; Landoni et al., 2024; Stephan et al., 2019). Although fluorescence provides a powerful means of molecular identification in the area of interest, the unlabeled cellular theater remains dark. In contrast, classical electron microscopy (EM) methods expose the cellular architecture. Still, the harsh protocols of specimen preparation – which typically include fixation, dehydration, and embedding – as well as the requirements for chemical staining, raise questions about the fidelity of the image representation. Cryo-microscopy takes an alternative approach by embedding the hydrated specimen in a vitrified state. Cryo-microscopy has most famously been applied in the field of structural biology, where cryo-transmission electron microscopy (cryo-TEM), combined with extensive image averaging, reaches near-atomic resolution for macromolecular structures. The cryo-TEM approach has been extended to tomography (cryo-ET) for studies of cells (for example, Frank, 2006; Young and Villa, 2023). In parallel, other 3D cryo-microscopy techniques have emerged for studying cells, notably soft X-ray tomography (Schneider et al., 2010) and serial surface imaging using a focused ion beam (FIB) with scanning electron microscopy (SEM) (Hoffman et al., 2020; Schertel et al., 2013). Each EM method has its strengths and limitations, which have been discussed elsewhere (see Elbaum, 2018; Varsano and Wolf, 2022). Cryo-ET requires specimens with a maximum thickness of a few hundred nanometers, which is significantly thinner than the typical scale of the cell and is even thinner than most mitochondria. The current practice to overcome this limitation is to prepare thin lamellae by FIB milling (Marko et al., 2007; Villa et al., 2013). An alternative approach, cryo-scanning transmission electron tomography (cryo-STET), offers specific advantages (Kirchenbuechler et al., 2015; Kirchweger et al., 2023a; Wolf et al., 2014, 2017) for studies at the scale of cytoplasmic organelles. Most notable is the ability to work with much thicker specimens than conventional cryo-TEM, reaching a micrometer or more, while retaining nanometer-scale resolution for structures not amenable to image averaging.

Several studies have used cryo-ET to study mitochondria in adherent fibroblast cells. For example, recruitment of the cytoskeleton to mitochondrial constriction sites has been found to correlate with depolarization of the membrane potential (Mageswaran et al., 2023). Such studies show many details of the ultrastructure of mitochondria and their surroundings. However, it has previously only been possible to study regions at the very edge of a cell or lamellae of ∼100–250 nm. Therefore, it has not been feasible to characterize intact whole mitochondria and their surroundings. In this work, we apply cryo-STET to mitochondria within their intact cellular milieu to study mitochondrial fission and the effect of knockout of mitochondrial fission factor (MFF).

Mitochondrial homeostasis involves perpetual extension and fusion on one hand and fission and degradation on the other. Therefore, fission is essential for organelle biogenesis, quality control, metabolism and apoptosis (Friedman and Nunnari, 2014; Mattie et al., 2019). The core machinery involved in fission (Kraus et al., 2021) includes the GTPase dynamin-related protein 1 (DRP1, also known as DNM1L) (Labrousse et al., 1999), its primary outer mitochondrial membrane (OMM) adaptor proteins MFF (Gandre-Babbe and van der Bliek, 2008; Losón et al., 2013; Otera et al., 2010) and Mid49/51 (herein referring collectively to Mid49 and Mid51, also known as MIEF2 and MIEF1, respectively) (Palmer et al., 2011), together with other cellular structures such as the endoplasmic reticulum (ER) (Abrisch et al., 2020; Friedman et al., 2011) and cytoskeleton (De Vos et al., 2005; Ji et al., 2017). After recruitment to the OMM, DRP1 forms puncta at fission sites (Ji et al., 2015; Otera et al., 2010) and forms a ring around the constriction site to finalize the fission step (Kalia et al., 2018). DRP1 also binds to actin filaments, which increases fission activity (Hatch et al., 2016; Ji et al., 2015).

MFF forms puncta on the OMM (Otera et al., 2010), and its oligomerization is specifically required for DRP1 recruitment (Liu and Chan, 2015; Liu et al., 2021). Observations using FM reveal that deletion of MFF results in elongated mitochondria (Gandre-Babbe and van der Bliek, 2008; Liu et al., 2021; Losón et al., 2013; Otera et al., 2010) because fission is inhibited while fusion is still ongoing. In wild-type cells, the mitochondrial network can be disrupted by treatment with oligomycin (De Vos et al., 2005; Yang et al., 2018). Oligomycin binds to the proton channel in the c-ring of the ATP synthase and blocks protons from passing through, which increases the proton gradient (Lardy et al., 1958; Lee and O'Brien, 2010; Symersky et al., 2012). Therefore, we expected that oligomycin treatment of MFF mutant cells would reveal early intermediates in the fission process.

Using cryo-STET, we aimed to capture snapshots of mitochondrial constrictions together with surrounding cellular structures that might be involved in the process. To this end, we induced fission by the established protocol of oligomycin treatment, for MFF^−/− knockouts of both mouse embryonic fibroblasts (MEFs; Losón et al., 2013) and human bone osteosarcoma cells (U-2 OS; Liu et al., 2021). Under these fission-inducing conditions, wild-type cells displayed a disrupted mitochondrial network. In contrast, the mitochondrial network of MFF^−/− mutant cells remained elongated but appeared irregular in size and shape at the level of FM. Using cryo-FM, including super-resolution radial fluctuation under cryogenic conditions (cryo-SRRF; Kirchweger et al., 2023a), we located the stalled fission sites for detailed investigation by cryo-STET. Among the unusual morphologies detected, we could also observe characteristic features of normal mitochondrial fission, such as wrapping of the ER around the constriction site, the presence of microtubules and actin filaments nearby, the formation of narrow OMM tubes, and OMM constriction with inner mitochondrial membrane (IMM) separation. OMM scission, the final fission step, is a function of MFF that is a prerequisite to DRP1 oligomerization. Still, OMM scission can clearly occur even in the absence of MFF, perhaps using MiD49/51 or FIS1 as adaptor proteins, or by another process such as mechanical tension generated by the cytoskeleton. The tomograms provide complete 3D snapshots of intact mitochondrial constriction sites and the surrounding cytosol, highlighting the strength of cryo-STET as a tool in cell biology.

First, we confirmed the oligomycin sensitivity of the wild-type and MFF^−/− MEF cells using time-lapse lattice light-sheet fluorescence microscopy of live cells. In wild-type MEFs (Chen et al., 2003), the reticulum of serpentine mitochondria (Fig. 1A) was fragmented by oligomycin treatment (Fig. 1B), as expected (De Vos et al., 2005; Yang et al., 2018). In MFF^−/− MEF cells, elongated, less branched mitochondria were observed, which showed some fission events. In addition, large ‘balloon-like’ structures were observed (Fig. 1C). Under oligomycin treatment, the mitochondrial morphology changed drastically, with abundant pearling (Fig. 1D, 0 min), but mitochondria remained tubular, as previously reported (Gandre-Babbe and van der Bliek, 2008; Ji et al., 2017; Losón et al., 2013; Otera et al., 2010). The varicose irregularities observed suggest constriction sites (Fig. 1D), but the linear configuration argues against complete rupture (Fig. 1E, 4 h). At the resolution of conventional fluorescence, we can conclude that fission does not proceed efficiently in MFF^−/− cells, as in the wild-type cells. Still, the mitochondria of MFF^−/− cells did show constrictions, hinting that fission was initiated but not completed. We can hypothesize that these constrictions represent stalled intermediates in the fission process. Additionally, the mitochondria displayed a wide variety of morphologies. These included elongated mitochondria with a regular diameter, widened cylindrical morphologies and ‘beads-on-a-string’ morphologies, as well as several large balloon-like and fragmented mitochondria whose connectivity was impossible to evaluate due to the large differences in fluorescence intensity (Fig. 1D). Our aim in this work was to visualize these sites at a higher resolution, together with the cellular surroundings, using cryo-STET.

Next, we established the baseline for MFF^−/− MEF cells under normal growth conditions without oligomycin. Using FM imaging, the mitochondria were found to be very long and narrow; however, some constrictions were observed (Fig. 1C).

To visualize the effect of the MFF knockout, we imaged MFF^−/− MEF cells under normal conditions using cryo-STET (Fig. 2; Figs S1, S2). To capture a variety of mitochondrial morphologies, at first, a larger field of view was imaged (8 µm; Fig. 2A–C). Tubular mitochondria, with a diameter of 200–300 nm, could be observed crossing the entire field of view of 8 µm (Fig. 2A–C; Fig. S1A–D). Shorter mitochondria, in the range of 2–4 µm with an increased diameter up to 2 µm, could also be observed (Fig. S1E,F). Therefore, the elongated morphology is the first difference between wild-type MEF mitochondria imaged by others (for example, Mageswaran et al., 2023) and the mitochondria of MFF^−/− mutant MEFs. This is consistent with our FM observations (Fig. 1C) and those of others (Gandre-Babbe and van der Bliek, 2008; Losón et al., 2013; Otera et al., 2010). Amorphous calcium phosphate (CaP) matrix granules could be observed as dark spots in the mitochondria (dark blue arrowheads, Fig. 2A and Fig. S1A) (Wolf et al., 2017). Interestingly, in one cell, none of the mitochondria seen in the tomograms seemed to have CaP granules (Fig. S1C–F), even though the tomograms in Fig. 2A, Fig. S1A and Fig. S1C–F are taken from the same sample. Microtubules were observed near the mitochondria (Fig. 2A,B). The cryo-FM images show the elongated nature of mitochondria in the MFF^−/−mutant MEFs (Fig. S1G) and are overlayed on the tomograms (Fig. 2C; Fig. S1D,F).

Next, we increased the magnification and applied dual-axis tomography to obtain further insights into the cristae network. This revealed the organization of the cristae network (Fig. 2D,E; Fig. S1B, Fig. S2) and even individual protein densities in the cristae membrane (Fig. S2, magenta arrowheads). Interestingly, a mitochondria-derived vesicle (König and McBride, 2024; Sugiura et al., 2014) with a diameter of ∼130 nm was observed pinching off from the OMM of a swollen area of the mitochondrion depleted of cristae (Fig. 2D,E). An additional swollen mitochondrion more than 1 µm wide could be observed, with sparse and dense cristae co-existing. Another tubular mitochondrion is shown in Fig. S1B, with a well-ordered cristae network except in the vicinity of a large vesicle. Therefore, deletion of MFF in MEF cells does not disrupt the cristae structure as a whole, but the cells show elongated mitochondria with strong internal variability.

We induced fission using 10 µM oligomycin for 3 h to block the ATP synthase in MFF^−/− MEFs. The mitochondria were found to constrict, but characteristically, fission did not reach completion. To visualize the morphological changes and to understand which cellular compartments take part in stress-induced constriction, we targeted constriction sites for observation using cryo-FM and cryo-STET. These mitochondria showed a variety of morphologies (Fig. S3), which we describe below.

The first morphology is the pearling or ‘bead-on-a-string’ shape, as shown in Fig. 3. In the area imaged, the cell was ∼880 nm thick. Several mitochondria reach across the entire tomogram with a field of view of 6 µm (Fig. 3A,B). Segmentation of the tomogram gives an impression of the 3D nature of the data (Fig. 3B). The main bodies of the mitochondria have an oblate spheroid shape with a width of ∼600 nm to ∼1 µm. The mitochondria show degraded, rounded cristae compared to those of the untreated cells seen in Fig. 2C,D and Fig. S2. Interestingly, no CaP deposits are present in these mitochondria. The beads are connected via thin tubules with a diameter of 40–55 nm. A bundle of microtubules runs alongside the central mitochondria (Fig. 3B). Several constriction sites can be observed. To illustrate adjacent cellular components, we highlight three sites, which, due to their morphology, we consider as likely fission sites (Fig. 3C–E). Site 1 (Fig. 3C) shows the ER partly wrapping around the thin tubule, with a bundle of microtubules crossing the site. The tubule is ∼650 nm long and has a diameter of ∼55 nm. Site 2 (Fig. 3D) shows the ER wrapping around the site over a longer distance along the tubule, with no microtubules seen at this site. The tubule is 600 nm long and 40 nm in diameter. Site 3 (Fig. 3E) has no notable cellular features nearby. The diameter of the tubule is 50 nm, its length is ∼830 nm, and it appears that the IMM is already separated (red arrowhead).

The ‘bead-on-a-string’ morphology was also seen in U-2 OS DRP1–GFP MFF^−/− cells in the presence of 5 µM oligomycin. Fig. 4 shows a slice through a tomogram with several mitochondria of differing morphologies near rough ER, microtubules and lipid droplets (Fig. 4A,B). Two mitochondrial constrictions to a diameter of ∼70 nm can be observed in the figure. An expanded view of one of the constriction sites (Fig. 4C) shows a microtubule running across the site and a lipid droplet nearby. Protein densities can be observed around the OMM. The presence of a CaP granule in the second constriction site suggests that transport was still active (Fig. 4D). When overlaying the cryo-FM image onto the tomogram, the DRP1–GFP signal can be seen to closely resemble that of the mitochondrial marker mitoBFP (Fig. 4G).

The second morphological category is balloon-like mitochondria. Fig. 5 shows a cryo-STET tomogram of an MFF^−/− MEF (Fig. 5A) and the corresponding 3D segmentation (Fig. 5B) containing a 2 µm-wide and 750 nm-thick mitochondrion with a degraded cristae network. At the end of the mitochondrion, a tubule is seen with a diameter of 40 nm extending further into the cytosol with an unclear ending. This mitochondrion does not contain any CaP granules.

Two other morphological features that we observed using FM are mitochondria with a regular diameter and with a significantly increased diameter. These features can be seen in greater details in the cryo-STET data from MFF^−/− MEFs (Fig. 6). Two neighboring dual-axis tomograms (Fig. 6A,B) show several mitochondria, including a 10 µm-long mitochondrion dividing into two (M1 and M2 in Fig. 6A,B). The mitochondrion extends from the first tomogram into the neighboring one (Fig. 6), as shown by the overlay of the tomograms onto the cryo-FM data (Fig. S4A). This mitochondrion can be separated into three parts, displaying two distinct constriction sites (Fig. S4B). In the first part (Fig. S4B, green box), an increased diameter of ∼750 nm with a degraded cristae network is observed at one end (Fig. 6A). Roughly in the center of the first tomogram, the ER is seen to wrap around the mitochondrion (Fig. 6A, green arrowheads). Careful inspection reveals presence of the ER tube at the top and bottom of the mitochondrion (Figs S4B and S5B). This marks the beginning of the second part, where the diameter is reduced (Fig. S4B, blue box). Interestingly, hardly any cristae can be observed in this narrower part of the mitochondrion. The second part stretches further into the neighboring tomogram (Fig. 6B) until an advanced constriction site is observed, which marks the beginning of the ‘third part’ (Fig. S4B, red box). This third part (M2 in Fig. 6B) has a diameter of 250–400 µm and shows well-ordered cristae (Fig. 6C,D). A longer mitochondrion (M3) with degraded cristae also reaches into the first tomogram (Fig. 6A,B). Smaller mitochondrial fragments are observed (Fig. 6A,B, M4–M9), some entirely depleted of cristae (M4, M9).

Having described the changes in morphology, we now highlight some contacts between mitochondria and other cellular components observed in the MFF^−/− MEF tomograms. In the following, we highlight some examples, even though more can be found when scrolling through the 3D volumes. Two microtubules are seen to interact with the mitochondrion at the fission site, separating M1 from M2 (Fig. 7A) and crossing on the top and bottom sides of the fission site. This fission site is particularly interesting because the IMM has already separated at two positions (Fig. 7A), although the OMM is still shared. This suggests that it is a fission site at a late stage. Additionally, actin bundles run alongside the mitochondrion from one tomogram into the second (Fig. 6C,D, pink bundles). At the third part (M2, Fig. 7B), whose morphology is closest to normal, these actin bundles directly interact with the mitochondrion before and after the fission site and at the very tip of M2 (Fig. S6).

The ER is wrapped around the long mitochondrion at several positions in these two tomograms (Fig. 6; Figs S4, S5A–D). One such contact site appears at the tip of M1 (Fig. 6A; Fig. S5A). Two additional ER tubules are seen on top of the mitochondria (Fig. S5B, z39 and z48), one of which wraps around the contact (Fig. S5B, z156) in the aforementioned early-stage fission site. Another ER tubule reaches below the same mitochondrion in the second part (Fig. S5C), and yet another appears above the center of the third part (Fig. S5D).

A cradle-like membrane contact is seen to coincide with disrupted OMM and disrupted cristae (Fig. 8A). Some protein densities making contact between the ER and the mitochondrion M7 are highlighted in Fig. 8A. An unknown filled tubular structure (light red tube in Fig. 6D) that makes contact with the end of M2 (Fig. 6C,D) appears close to M7 and makes contact with the ER (Fig. 8A). This tube has a diameter of 70–80 nm. Atypically for ER, it is filled with some density. We suggest that it might be a tubulated form of a degrative organelle (Bohnert and Johnson, 2022). Round degradative organelles resembling multivesicular bodies or endosomes (Fig. 8B; Fig. S5E) appear adjacent to mitochondria that lack cristae (M4, M9) and show ruptured OMM. Some densities are observed in the area where the OMM disintegrates. Interestingly, these small mitochondrial fragments retained green fluorescence (Fig. S4A), suggesting that the mitochondrial matrix had not yet been acidified due to its contact with the degradative organelle (Sargsyan et al., 2015).

In summary, in this study, we highlight correlative cryo-STET, which enables us to image thicker areas of whole cells without the need for complicated thinning approaches. Using the cryo-FM information as a guide, we targeted snapshots of mitochondrial constriction sites to learn more about fission and the different cellular structures involved in the process. Due to the increased field of view of cryo-STET, we imaged a myriad of interactions between mitochondria and other cellular structures.

Mitochondrial morphology is a classic topic in cell biology (Lewis and Lewis, 1915; Monzel et al., 2023; Preminger and Schuldiner, 2024). Mitochondria undergo constant fission and fusion, dependent on protein factors for membrane remodeling (Kraus et al., 2021; Pagliuso et al., 2018). In the case of fission, MFF was previously identified as an early-stage adaptor to the dynein-related scission protein DRP1 (Otera et al., 2010). MFF is thought to oligomerize on the OMM, reducing its diameter to roughly 100 nm (Liu and Chan, 2015; Liu et al., 2021), at which point DRP1 can form rings that ultimately cleave the membrane constriction (Kalia et al., 2018). Other factors, such as Mid49/51 (Palmer et al., 2011), INF2 (Hatch et al., 2014), the ER (Friedman et al., 2011), membrane tension due to cytoskeleton pulling (Mahecic et al., 2021), or other mechanical forces (Helle et al., 2017; Romani et al., 2022), have also been implicated in constricting the future cleavage site (Kraus et al., 2021). Using correlative light and electron microscopy (CLEM; Schwartz et al., 2007)-guided cryo-STET imaging, we found that OMM constriction can progress to a very exaggerated stage even in an MFF knockout background. We propose that our observations reflect the dysfunction of the last step in fission, that is, DRP1-mediated scission of the OMM, in the absence of MFF.

FM imaging revealed that MFF^−/− MEF cells possess an extended mitochondrial network. In untreated MFF^−/− cells, mitochondria are highly elongated and appear somewhat narrower than normal (Fig. 1C, Fig. 2; Fig. S1). The cristae pattern seems well ordered along much of the mitochondria length but is degraded in swollen areas (Fig. 2C,D; Fig. S1B, Fig. S2). Although shorter and swollen mitochondria are observed (Fig. S1E), constrictions of mitochondria are difficult to find. Upon treatment of the MFF^−/− cells with oligomycin, the frequency and morphology of the constrictions change drastically. These changes were apparent already in the first recorded image (Fig. 1D, 0 min) and persisted at least to the 4 h time point (Fig. 1D, 4 h), as reported in earlier publications (Yang and Chan, 2024; Yang et al., 2018). By cryo-FM and/or cryo-SRRF, we targeted constriction sites after 3 h. Cryo-STET was then used to visualize the snapshots of the constriction sites and to elucidate the contributions of other intracellular structures at different stages.

Oligomycin treatment transformed the narrow mitochondrial cylinders of MFF^−/− cells into a varicose, beads-on-a-string morphology of bulges interspersed by very narrow tubules. We propose that these reflect a stalled or abortive intermediate in the fission process; therefore, we call them ‘fission tubules’. The cristae network in the bulges is less regular in than that seen in the tubes of untreated cells, showing degraded circular cristae (Fig. 3A,B). Very thin, narrow fission tubules connect the bulges (Fig. 3C–E). These appear, at least in some cases (Fig. 3E), to be composed only of OMM, leaving a closed IMM on either side. It has been reported previously that the constriction of the IMM and OMM can take place independently (Cho et al., 2017; Malka et al., 2005). This is in contrast to normal mitochondrial fission (Mageswaran et al., 2023), where the two membranes pinch together simultaneously in a cusp-like shape with no intervening tubular extension. We suggest that the fission tubules in the MFF^−/− cells might be drawn out from the isolated pinch point by a pulling force due to a bundle of microtubules running alongside the mitochondria (Fig. 3A,B) and passing across one of the fission sites (Fig. 3C). Rupture of the fission tubule leaves an isolated mitochondrion, for example as seen in Fig. 5, where the tubule remains attached on one side.

Two other constriction sites are reported here, one in a very early stage, where a change in morphology, colocalization with the ER, and a crista perpendicular to the OMM was observed (Fig. 6A; Fig. S5B, between parts one and two), and another one at a later stage, where the IMM had already clearly separated while the OMM remained continuous (Figs 6B and 7A). Additionally, in MFF^−/− DRP1–GFP U-2 OS cells, DRP1 appeared to be distributed across the whole mitochondrion and not localized specifically to the fission sites (Fig. 4E–G), consistent with earlier reports (Liu et al., 2021; Otera et al., 2010).

Whereas MFF is also considered to be an early-stage factor in mitochondrial fission by recruitment to ER–mitochondria contact sites (Abrisch et al., 2020; Friedman et al., 2011), we observe that constriction progresses toward fission even in its absence, stalling only before OMM scission. Early functions might be fulfilled by redundant factors, in particular Mid49/51. Even without MFF, fission is initiated by constricting the mitochondria by recruiting the ER and microtubules to the fission sites. The process continues, with the IMM already separated. This is consistent with a model in which severing of the OMM by DRP1 does depend on MFF in an obligatory way. In the absence of scission, membrane tension could explain the transformation of the pinch site into a long fission tubule. The final rupture would then be affected by a mechanical force, where via an anchoring factor, one end of the mitochondria gets fixed to the cytoskeleton – for example, microtubules (Fig. 3) or actin (Figs 6 and 7B) – while the other end gets pulled away. This bears a similarity to the formation of mitochondrial nanotunnels (Vincent et al., 2017; Wang et al., 2015). We have summarized our observations in Table S1 and our model in Fig. S7. We note that the snapshots of the constriction sites are taken from cells under oxidative stress in order to increase the constriction frequency; therefore, further imaging of constriction sites under physiological conditions is needed.

Technologically, the correlative cryo-FM–cryo-STET workflow combines the specificity of FM with the possibility of imaging the whole cellular environment to a thickness of ∼1 µm in a near-native state without the need for sample thinning and staining. Locating the rare constriction sites without guidance from the cryo-FM would be challenging. Moreover, the fission tubules are so narrow that the probability of finding them in a FIB-milled section of 100 nm is exceedingly slim. Tubules that end on one side, such as the one seen in Fig. 5, might also be lost in sectioning, so such features would have to be observed multiple times for confidence in interpretation. Here, the requirement for a large statistical sampling is replaced by the ability to image through the entire cytoplasm. The lower resolution of cryo-STET, on the other hand, makes it impractical for sub-tomogram averaging so far, at least with the simple bright-field modality employed here. In recent years, we have enhanced scanning transmission EM (STEM) technology (Kirchweger et al., 2023b; Wolf et al., 2014) by adding 3D deconvolution (Waugh et al., 2020) combined with dual-axis tomography (Kirchweger et al., 2023a) to the toolbox. Further developments are underway, especially with four-dimensional STEM methods (Seifer et al., 2024), and we can look forward to enhanced resolution in the foreseeable future.

Wild-type MEF (Chen et al., 2003) and MFF^−/−MEF cells (Losón et al., 2013) expressing EGFP in the matrix of the mitochondria with a SU9 mitochondrial targeting sequence were a gift from the lab of David Chan (California Institute of Technology, CA, USA). Wild-type U-2 OS cells and U-2 OS DRP1–GFP MFF^−/− cells (Ji et al., 2017; Liu et al., 2021) were gifts from the lab of Harry Higgs (Geisel School of Medicine, Dartmouth College, NH, USA). MEF and U-2 OS cells were maintained in DMEM (Gibco, 41965) with 10% fetal calf serum (FCS; Biological Industries, 04-007-1A), 4 mM L-glutamate (Biological Industries, 03-020-1B), 1% Pen/Strep (Biological Industries, 03-031-1B) and 1 mM sodium pyruvate (Biological Industries, 03-042-1B) at 37°C and 5% CO₂ in a humid environment. The U-2 OS cells were transfected with 800 ng of mitoBFP (amino acids 1–22 of COX4 N-terminal to BFP; a gift from the Higgs Lab) in 1 ml volume in a 3.5 cm dish (Ji et al., 2015) using jetPRIME (Polyplus) according to the short DNA transfection protocol. To induce fission, the cells were incubated with 5–10 µM oligomycin (Sigma, O4876) for ∼3 h before imaging or vitrification.

For FM imaging (Fig. 1), the MEF cells were grown in an 8-well slide. The images were recorded on a Lattice Lightsheet 7 (Zeiss, Germany), equipped with an Orca Fusion camera (Hamamatsu, Japan), with a 30 µm×1000 nm light sheet. The fluorescence was excited with the 488 nm laser, and the emission was recorded using the LP565 emission filter. The cells were kept in an on-stage incubator at 37°C and 5% CO₂ in a humid environment. Images were recorded for ∼1 h before inducing fission by adding oligomycin. After adding oligomycin, a focus and settings check was performed. Thereafter the cells were observed for 3 h, by scanning every 3 min. Please note that 0 min in the movies deposited at Zenodo (doi:10.5281/zenodo.14959968) refers to the timeline of the movie, not the time of the oligomycin treatment. After data collection, the data were deskewed without further post-processing. For maximum z-projection, the z-slices showing in focus information were chosen. Of the MFF^−/− and wild-type MEFs, several independent FM images were recorded; however, only one is shown here.

For specimen preparation, the cells were grown on either R3.5/1 200 mesh grids (Quantifoil), onto which 10 nm of continuous carbon was floated (Fig. 2A–C; Figs S1A–C, S2, S4D), or G200F1 R2/2 SiO2 grids (Quantifoil) (remaining figures). The grids were glow discharged and coated with human fibronectin (FAL356008, Corning; 50 µg µl⁻¹) in PBS without Mg²⁺ or Ca²⁺ for ∼1 h. Subsequently, 2 ml of growth medium was added, and the cells were seeded on the grids. The cells were grown in the incubator to a density of ∼1–2 cells per grid square, usually overnight. Oligomycin was added ∼3 h before chemical fixation or plunging. Before plunge freezing, MEF cells in Fig. 2D, Fig. 6 and Fig. S1B were chemically fixed using 4% paraformaldehyde and 0.1% glutaraldehyde in PBS buffer. The specimens for the remaining figures were unfixed. The U-2 OS cells were transfected and plunge frozen after ∼18 h of transfection. The U-2 OS cells were not chemically fixed. Vitrification was performed as described previously (Kirchweger et al., 2023a). In brief, we used a Leica EM GP plunger (Leica Microsystems, Vienna, Austria; Resch et al., 2011). The chamber was set to 37°C and 95% humidity. Before blotting, 3 µl of growth medium (front side) and 1.5 µl of homemade 15 nm gold fiducial markers (back side) were added. Subsequently, the grids were blotted from the back side for 5 s and plunged into liquid ethane.

As a summary, Figs 3, 5, 6, S3A–C and S3D represent five independent experiments on the MFF^−/− MEFs under fission-inducing conditions. Fig. 2A–C (and Fig. S1A,C–F) represent two independent experiments. For the MFF^−/− DRP1–GFP U-2 OS cells (Fig. 4), only one experiment is shown.

Cryo-FM grid maps were recorded on a Leica Cryo CLEM (Leica Microsystems) using the GFP channel as described previously (Kirchweger et al., 2023a; Schorb et al., 2017). The microscope is equipped with a Hamamatsu Orca-Flash 4.0 sCMOS camera. A focus map was recorded in the GFP channel, and the subsequent grid map was recorded as a z-stack of typically 20 µm in the bright-field and GFP channels. Cryo-SRRF data were acquired as described previously (Kirchweger et al., 2023a). In short, an area of interest was chosen, and 500 frames of ∼1024×1024 pixels of different z-heights were recorded. The data was analyzed using the NANOJ-SRRF (Culley et al., 2018; Gustafsson et al., 2016) plugin in FIJI (Schindelin et al., 2012) using a Python script to find the best SRRF settings by splitting the stack into even and odds and calculating the Fourier ring correlation (FRC) of the final SRRF images. The script is available on Github (https://github.com/PKirchweger/SRRF-macro).

We followed the previously reported protocol for alignment and set-up of the microscope (Kirchweger et al., 2023a,b). Bright-field (BF) cryo-STET data were collected on a Titan Krios G3 (Thermo Fisher Scientific), operated at 300 kV, equipped with an X-FEG electron source, a dual-axis stage and a Fischione high-angle annular dark-field (HAADF) detector. We used a semi-convergence angle of 1.2 mrad, a 70 µm condenser aperture and spot size 9, which resulted in a current of ∼20 pA of the factory-calibrated projection screen. BF cryo-STET collection on the HAADF was achieved by inserting a 20 µm objective aperture to limit the collection angles to 3 mrad and, using diffraction alignment, moving the beam off-axis onto the sensitive area of the HAADF (Kirchweger et al., 2023a). Cryo-STET data were collected using the STEM Tomography software (Figs 2A, 3 and 5) (Thermo Fisher) or SerialEM (Figs 2C, 4 and 6) (Mastronarde, 2003, 2005; Resch, 2019).

In the case of the STEM Tomography software, a low-resolution Atlas was recorded, and positions for tilt-series data collection were chosen using the fluorescent map as a guide. Tilt series were recorded in a bidirectional manner, from −20° to 60°, and then from −22° to −60°, with an image of 2048×2048 pixels at a magnification of 21,000× or 29,000×, with sampling steps of 4.212 nm or 3.001 nm, respectively. The holder calibrations were used to skip autofocusing during the data collection.

In the case of SerialEM, cryo-STET data collection has been described previously (Kirchweger et al., 2023a,b). Briefly, a low-resolution map was recorded in LM-STEM mode (260× magnification), to which the fluorescence map was registered. Medium-resolution square maps at 3500× magnification were recorded based on how the cell was positioned on the grid square and the fluorescence of the mitochondria. Then, low-dose mode was activated, and tilt series positions were defined using Anchor Maps and recorded in View and Preview settings. Tilt series were recorded starting at 0°, using the dose symmetric tilt scheme up to ±60° (Hagen et al., 2017). Tilt series with 4096×4096 pixels were recorded at 29,000× or 41,000× magnification, resulting in sampling steps of 1.447 nm and 1.021 nm, respectively.

A SerialEM script was developed for dual-axis cryo-STET data collection (deposited at the Nexperion SerialEM Script Repository; https://serialemscripts.nexperion.net/script/81; version 1). The script involves several steps. First, a low-resolution Record image of the center of the grid is recorded and saved as a map in the navigator. Then, the grid is rotated by 90°, and a second image at the center is recorded and saved as a navigator map. Then, the Align-with-rotation procedure is called, and the two center images are compared and aligned. The resulting rotation and shift are applied to all navigator items, including the Map images. After terminating the script, the tilt-series data collection is set up for the second axis. Realign-to-item should work well with the rotated Anchor Maps if the tilt series position is not too close to the square's edge.

Tilt series alignment was performed in IMOD (Kremer et al., 1996; Mastronarde and Held, 2017). Patch tracking was used for the tomogram in Fig. 3, where no fiducials were present. Other tilt series were aligned using fiducials. All tomograms were reconstructed to an image size of 2048×2048 pixels. For visualization and dual-axis tomogram combination, tomograms were reconstructed using the SIRT-like filter with constant=30, while the weighted back projection with a HammingLikeFilter set to 0.0 was used for subsequent deconvolution (entropy regularized deconvolution, ERDC; Kirchweger et al., 2023a).

For ERDC, the densities of gold beads were removed from the reconstructed map using the findbeads3D pipeline in IMOD. ERDC was performed using the recently published Python script as described previously (Croxford et al., 2021; Kirchweger et al., 2023a; Waugh et al., 2020). The script is available on GitHub (https://github.com/PKirchweger/CSTET_Deconvolution).

Amira 3D software v2022.2 was used for segmentation and 3D representation of the reconstructed data (Thermo Fisher Scientific). Distance analysis was performed using the ‘Surface Distance’ module in Amira 3D, which measures the distance between two triangulated surfaces. The module calculates the nearest point on the opposite surface for each vertex and outputs vector magnitudes. These distances are visualized with color intensity representing the distance at each node.

ChatGPT and Microsoft copilot were used for grammar correction. After using these services, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Wild-type and MFF^−/− MEF cells were a gift from the lab of David Chan, California Institute of Technology, CA, USA. The U-2 OS cells were a gift from the lab of Harry Higgs, Geisel School of Medicine, Dartmouth College, NH, USA. The authors thank Tal Ilani and Deborah Fass, Department of Chemical and Structural Biology, Weizmann Institute of Science, for support during the experiments and extensive discussion on the manuscript. In addition, we thank Tatjana Smirnova and Joseph Addadi from the Department of Life Science Core Facilities, Weizmann Institute of Science, for the Lattice Lightsheet data collection.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.263639.reviewer-comments.pdf