Mitochondria in disease: changes in shapes and dynamics

1Mitochondria (see Glossary) were discovered in the 1800s, and the advent of high-resolution electron microscopy (EM) techniques allowed Palade to identify their unique double membrane in the 1950s. Additional studies throughout the mid-1900s revealed the link between mitochondrial structure and function, and continual EM advancements have enabled 3D renderings of cellular ultrastructure. Across the 1990s, developments in plastic section electron tomography (ET), which remains clinically relevant today, allowed researchers to explicate the complexity of mitochondrial internal micro-compartments, known as cristae. Today, cryo-ET, using principles of rapid freezing, allows for 3D reconstruction of mitochondria and localizing associated large protein complexes (e.g., ATP synthase) in their native state, further elucidating the mitochondrial structure–function interplay.

2Recent advances in research have revealed mitochondria as dynamic organelles with heterogeneous shapes and networks both within and among cell types and tissues (Figure 1). For example, mitochondrial morphology can range from a single compact (e.g., in cardiac tissue) or swollen (e.g., in brown adipose tissue) structure to highly branched networks (as in skeletal muscle). Phenotypes can exist ubiquitously across tissue types, although certain tissue types or disease states are more likely to display certain morphologies: for example, liver mitochondria shift between spherical and donut- or c-shaped forms, cardiac muscle cells have elongated mitochondria arranged in parallel, and skeletal muscles have highly connected mitochondrial networks. These phenotypes are best analyzed via high-resolution 3D technology. For instance, low-resolution techniques may mistake nanotunnels for fragmentation and utilizing EM thin sections alone may falsely show cup- or stomatocyte-shaped mitochondria as mitochondrial donuts. Mitochondria can also display distinct morphologies and biochemical properties within a given cell: populations near the plasma membrane play essential roles in the functional coupling of ATP-guided ion channels, whereas perinuclear networks regulate gene transcription under hypoxic conditions by increasing reactive oxygen species (ROS) in the nucleus. The variation in structure between these subpopulations may reflect differences in function. Different intra- and intertissue mitochondrial networks and populations also display different responses to substrates, inhibitors, and oxidative stress, highlighting their complexity, diversity, and plasticity.

3In addition to producing energy in the form of ATP, mitochondria are involved in calcium signaling, apoptosis, and the oxidative stress response. Mitochondrial dynamics and heterogeneity are crucial for regulating these processes and maintaining cellular homeostasis both within individual cells and across an organism. Several mechanisms, including mitochondrial biogenesis, transport, mitophagy, fusion, and fission, regulate mitochondrial structure and function. Other cellular components, such as lipid droplets and the endoplasmic reticulum (ER), can also influence mitochondrial structure through interorganellar contact sites that modulate mitochondrial biochemical properties, functional capacity, and dynamics.

4Studies of 3D mitochondrial structures indicate that structural changes may occur during aging and that these changes may be associated with functional deficiencies and the development of mitochondria-dependent diseases. Additionally, a recent review summarized how mitochondrial phenotypic frequencies may change in response to stimuli. As we explore here, the differential responses to aging among disparate tissue types, which can include the presence of unique mitochondrial structures, may be a key pathological factor. However, the relative frequencies of these unique shapes, including MGs, nanotunnels, donuts, and elongated, branched, and compact morphologies (Table 1), must be further elucidated in disease states.

5Here, we first describe mitochondrial donuts, MGs, and tunnels in different cell types and in response to different stimuli before discussing how these shapes are related to the organization and orientation of their cristae. We then build on the overview of these internal and overall structures by describing how fusion and fission events, contacts with other organelles, and other structural regulators impact mitochondrial shape. Finally, we explore the potential clinical implications of improved knowledge of mitochondrial structures and dynamics.

6Ahmad et al. reported that donut-shaped mitochondria may be an early cellular stress marker. They showed that ROS production induced by complex I and III inhibitors resulted in the transient and reversible transition from tubular to donut-shaped mitochondrial structures, which are associated with increased mitochondrial ROS levels. The formation of these structures may also exhibit tissue dependency. Bleck et al. observed in healthy adult mice that, compared with oxidative muscle fibers, glycolytic muscle fibers showed a higher frequency of donut-shaped mitochondria. In contrast, donut-shaped mitochondria may arise more frequently in the liver due to metabolic stress. Donut-shaped mitochondria can also appear in disease states: for example, the memory impairment characteristic of Alzheimer’s disease occurs concomitantly with an increased prevalence of neuronal donut-shaped mitochondria.

7Our understanding of donut-shaped mitochondria has evolved alongside the techniques that allow us to model them, including EM and live-cell imaging. Donut-shaped mitochondria form when the opposite ends of a mitochondrion fuse, although the mechanism underlying this process remains largely unknown. While it has been suggested that donut-shaped mitochondria can form independent from fission, these fission-independent models likely show collapsed mitochondria that phenotypically appear as cup shaped, which represent a fission-independent response to mitochondrial membrane potential loss. Donut mitochondria have been observed under hypoxic conditions, and are associated with reduced mitochondrial fusion events and changes in osmotic pressure that lead to diminished neuronal synapses. When returned to nonstressed conditions, donut-shaped mitochondria readily revert to other morphologies, suggesting that the donut shape may represent an adaptive response for increasing mitochondrial surface area for organelle contacts or resisting mitophagy. Donut-shaped mitochondria maintain the overall mitochondrial membrane potential (generated by an electrochemical gradient across the inner mitochondrial membrane). This suggests that donut formation may protect against mitochondrial function loss by reducing the Gibbs free energy of the system, which inversely increases membrane potential. However, it is unclear if other mitochondrial structures depend on Gibbs free energy, as well as associated thermodynamic temperature. While past results have found that even under uncouplers, mitochondria donut diameters tend to be approximately 1.3 μm, it is unclear if donut size may determine the bending energy, a major barrier to donut formation that may be counterbalanced by stress-inducible mitochondrial osmotic pressure. Furthermore, as the study of mitochondrial donuts continues, advanced methods must be utilized, as fluorescence and thin EM techniques may mistakenly identify discoid-, curved-, and vase-shaped mitochondria as donuts.

8Mitochondria that swell to two to three times their typical size (from an average length of ~5 μm up to 10 μm) while maintaining outer membrane integrity are characterized as MGs. Although these phenotypes can reach the size of the nucleus, ultrastructural or 3D reconstruction techniques are useful to identify them. MGs form in response to diseases, such as nonalcoholic fatty liver disease, and environmental or cellular stressors. However, the effect of MG formation on cell function is unknown.

9The mechanism underlying MG formation is not fully understood, but mitochondrial fission and fusion defects are likely contributors. During stress, fusion increases mitochondrial complementation and maximizes oxidative capacity as a compensatory mechanism. MG formation may also result from stress-induced hyperfusion, the accumulation of damaged or dysfunctional mitochondria due to fission or fusion defects, and mitochondrial membrane potential–dependent changes in mitochondrial fission and fusion.

10The ER-associated degradation (ERAD) pathway also regulates mitochondrial size, and ERAD deficiency leads to MG formation, suggesting that MG formation may be regulated through inter-organelle mechanisms. Given there is a known association between ER stress and mitochondrial elongation, Zhou et al. investigated how ERAD deficiency affects mitochondria and found that deficiency in the ERAD component Sel1L promotes MG formation in brown adipocytes, possibly due to inhibited fission or accelerated fusion. This finding suggests that MG formation may be affected by changes in expression of specific proteins at the mitochondrial–ER contact sites (MERCs).

11Investigations of several diseases have uncovered various specialized mitochondrial shapes. In Alzheimer’s disease, 3D EM revealed a novel elongated mitochondrial phenotype known as ‘mitochondria-on-a-string’ – later termed nanotunnels. As reviewed, mitochondrial nanotunnels are thin (<200-nm diameter) double membranes that connect adjacent mitochondria (up to 30 μm in distance) and can transport proteins and molecules across the mitochondrial membrane into connected organelles. Nanotunnels have been detected in diseased tissue and are hypothesized to result from incomplete fission, although it is unclear whether their formation requires specific types of fission (e.g., midpoint or endpoint). The effect of nanotunnel formation on the distribution of dynamic mitochondrial proteins and the structure dependence of protein localization also requires investigation. The spatial orientation of mitochondrial proteins may reveal nanotunnel formation mechanisms, although the specific pathways involved in healthy and disease conditions remain underexplored.

12In human skeletal tissue samples from a patient with mitochondrial disease and large-scale mitochondrial DNA (mtDNA) deletion, Vincent and colleagues identified an 18.1-fold increase in nanotunnel formation compared with healthy samples, suggesting a correlation between nanotunnel formation and mtDNA mutations, although other mechanisms may be involved. Mitochondrial fragmentation due to excessive fission indicates cell damage, and nanotunnels may form in response to cell damage caused by mtDNA mutations. There are other examples of proteins involved in nanotunnel formation: ryanodine receptor 2 (RYR2)-dependent calcium homeostasis is known to regulate nanotunnel formation, and activating transcription factor 4 (ATF4) and other ER stress proteins could play a role in this regulation; little else is known about these regulatory mechanisms and would benefit from further exploration.

13One potential complication of studying nanotunnels is that they are among the mitochondrial shapes that arise under hypoxic conditions and therefore may result from improper, hypoxia-inducing fixation techniques. Thus, the standardization of fixation and quantification techniques must be improved to ensure that mitochondrial structures represent altered metabolic conditions rather than improper fixation.

14The diversity and heterogeneity of mitochondrial shapes, including those described in the preceding text, affect the spatial orientation of cristae and are modulated by the mitochondrial contact site and cristae organizing system (MICOS) complex, as well as mitochondrial signaling pathways (Figure 2A). Although cristae have unique dynamics independent of mitochondrial structures, common fusion proteins, including optic atrophy 1 (OPA1), which is epistatic to the MICOS complex, and YME1L, regulate their dynamics through crista junction closing and opening. Cristae are tubular or lamellar under normal conditions; abnormal shapes, such as spherical and onion-ring cristae (Figure 2A), arise in abnormal conditions, such as mitochondrial dysfunction and mitochondrial disorder, respectively. These abnormal shapes may optimize intracristal spaces, but their functional implications remain understudied. In some mitochondrial phenotypes, such as MGs, a paucity or abnormality of cristae is common. However, while the cristae structural differences may confer functional changes, the interdependence between cristae shape and mitochondrial phenotype remains poorly explicated.

15Cristae harbor the machinery for oxidative phosphorylation and ATP production, and their organization is intricately linked to energy production efficiency and overall mitochondrial health. Advanced imaging techniques, such as 3D MINFLUX nanoscopy, have revealed the complex organization of cristae and the MICOS complex, which is involved in structural cristae remodeling. Cristae organization disruptions and alterations are implicated in neuro-degenerative, metabolic, and cardiovascular diseases.

16Notably, the interplay between crista shapes and the mitochondrial permeability transition (mPT) must be further explored. Though the molecular mechanism is unknown, elevated calcium may open an mPT pore to an unspecified channel, resulting in ruptured cristae and eventual cell death. As mPT pores are regulated by matrix cyclophilin D, they may consist of ATP synthase dimers; however, this suggestion remains controversial. Beyond the role of mPT in disease, the mPT has clear roles in modulating membrane potential. This offers implications in mPT regulating membrane potential-dependent mitochondrial structure, such as the formation of cup-shaped cristae upon loss of membrane potential. Because individual cristae can have distinct membrane potentials, and because ATP synthase dimerization regulates crista morphology, the interplay between mPT, ATP synthase dimerization, cristae structure, and mitochondrial structure must be explicated, possibly through cryo-ET, which enables visualization of ATP synthases.

17Cristae undergo ongoing and dynamic membrane remodeling events, which may be implicated in mPT pore formation or shape tendency. In these continuous cycles, similar to OPA1-mediated cristae junction flux, cristae are continuously remodeled by transverse, x-type, y-type, and other types of cristae fusion and fission (Figure 2B). However, this process remains poorly elucidated. For example, crista dynamics may be determined by the spatial orientation of the mitochondrial MICOS complex, which may depend on the mitochondrial shape. Yet, it is unclear whether the relative rates of types of crista remodeling events affect their membrane potential, tendency toward crista organizational types, or mitochondrial shape, especially in the case of pathological conditions.

18Discussions of mitochondrial dynamics often include a common set of core proteins, including OPA1, dynamin-related protein 1, mitofusin 1, and mitofusin 2. However, the current understanding has evolved beyond these factors (Figure 2C). In addition to its role in cristae remodeling, OPA1, when in its long, active form, plays an important role in fusion. Upstream, OPA1 can be inhibited through cleavage by YME1L and OMA1, resident proteases that modulate mitochondrial dynamics. Similarly, actin constricts to enact fission, and other mitochondrial dynamics proteins, including MiD49 and MiD51, also play a role in fission. Notably, actin constriction prefission is orchestrated by ER-associated IFN2, highlighting the interconnectedness of MERCs and mitochondrial fusion–fission dynamics.

19Although past studies have suggested that mitochondrial networks are determined entirely by fusion and fission dynamics, the involvement of various fission factors, such as mitochondrial fission factor (MFF) and mitochondrial fission 1 protein (FIS1), in determining mitochondrial phenotype remains unclear. Kleele et al. demonstrated that distinct forms of fission occur based on spatial orientation of FIS1 and MFF, suggesting that protein distribution may contribute to mitochondrial morphology. A recent study found that FIS1 overexpression results in increased donut formation, but whether MFF overexpression has similar effects has not yet been explored. Although previous studies have shown that mitochondrial fission events are integral to mitochondrial morphology outcomes, the specific contributions of unique 3D phenotypes, such as MGs, and contact sites require further exploration.

20Estrogen, which declines with aging, is associated with changes in mitochondrial morphology. In ovariectomized monkeys, donut-shaped mitochondria are a pathological indicator of age-related Alzheimer’s disease, and normal mitochondrial structure and function can be rescued by estrogen treatment. Notably, an in vivo mouse study showed that 17β-estradiol (E2) in mitochondrial membranes decreases membrane viscosity, which can have adverse effects given that optimal membrane fluidity is necessary for proper mitochondrial function. Mitochondrial membrane potential changes may occur concomitantly with alterations in membrane fluidity, and mitochondrial cristae have functionally independent membrane potentials. Therefore, it could be informative to examine whether E2 inner-membrane localization occurs, and, if so, whether such localization is cristae-dependent or E2 binds to specific areas of the inner mitochondrial membrane. Furthermore, sex-dependent differences in mitochondrial structure across aging remain unclear, which has implications for uncovering if estrogen may modulate mitochondrial structure and confer a protective effect.

21Nucleoids contain mtDNA, which encodes several proteins and RNAs (tRNA and rRNA) involved in protein synthesis, mechanistically enabling submitochondrial organization and maintenance of mtDNA. As reviewed, nucleoids, dictated by tissue type and age, often have integral roles in modulating metabolism. Lewis et al. convincingly demonstrated that mtDNA signaling involves MERCs. Nucleoid transportation at MERCs is further determined by the MICOS complex, suggesting the potential involvement of cristae. Although fusion and fission are understood to contribute to mitochondrial genome integrity through the maintenance of nucleoids distributed to daughter mitochondria, it is unclear how different mitochondrial shapes affect nucleoid replication and distribution. For example, the large size of MGs may affect MERC formation and consequently impact nucleoid distribution. The mechanisms underlying the formation of different mitochondrial shapes may also impact the number and distribution of nucleoids; for example, if MGs are formed from multiple fusion events, they may contain more than one nucleoid, which may impact their function. Thus, further study is needed to understand how changes in mitochondrial structure affect mtDNA and vice versa. Utilizing new techniques, such as Mitomate tracker, for the automated quantification of nucleoid distribution in tissues where mitochondrial phenotypes are naturally found (e.g., liver cells for MG), may offer greater insight into this topic.

22The MICOS complex is involved in aging and, as described in the preceding text, cristae organization (Figure 2A), and may also affect overall mitochondrial shape. As previously reviewed, the MICOS complex contains multiple proteins, although their independent roles and necessity for MICOS complex functionality remain unknown. One such protein, mitofilin, is important in nucleoid organization and metabolism, suggesting that nucleoid distribution may involve the MICOS complex or that its various components play essential roles beyond metabolism. The MICOS complex also interacts with the sorting and assembly machinery complex, but the necessity of these two complexes for the formation of specific mitochondrial shapes remains unclear.

23Tissue-dependent deletion of the MICOS complex and its specific proteins may advance our understanding of how the MICOS complex affects the gross mitochondrial structure and cristae spacing. Tools have recently been developed to enact overexpression of MICOS complex components. However, floxed models for the deletion of MICOS complex components are lacking, and it remains unclear whether the MICOS complex subunits can be deleted in a tissue-specific manner to better identify the specific functions of individual components.

24As this section highlighted, our understanding of mitochondrial modulators must continue to evolve. One promising avenue is the study of mitochondrial transporters, specialized proteins that mediate the movement of essential small molecules that are critical in maintaining cellular energy production, metabolism, and redox balance (Box 1). Mitochondrial processes involving the internal diffusion of metabolites and ions are influenced both by the shape of the inner membrane, in turn, affected by crista dynamics, and transport across the membrane, making transporters novel potential upstream regulators of mitochondrial structure. Understanding their integrated functions in mitochondrial biology may lead to the identification of therapeutic targets and novel modulators of mitochondrial structure in various diseases.

25Mitochondrial structures should be considered in the context of the entire organellar connectome, including MERCs and contacts with lipid droplets and autophagic materials. As noted previously, recent findings have shown that MERCs have regulatory roles in both fission and fusion, thereby modulating mitochondrial morphology. However, it is unclear whether different types of contact sites with unique biochemical roles exist. For example, centers of donut-shaped mitochondria are often filled with sarcoplasmic reticulum (SR), ER, or lipid droplet contact sites (Figure 2D). Tangentially, mitochondria form contacts with lipid droplets to perform fatty acid oxidation. Studies of these lipid droplets show that their mitochondrial contact sites can either be anchored or formed in ‘kiss-and-run’ interactions, each of which has distinct functional implications and formation pathways. As organelle recruitment may differ across distinct structures, future studies should examine how different mitochondrial shapes affect MERCs.

26Beyond MERCs, deeper investigation is needed into how mitochondrial interactions with other organelles, such as recycling machinery, may dictate morphology. Notably, artificial transgene tethers have arisen as a potential therapy to regulate MERC formation, which may thereby alter mitochondrial morphology. This application represents a potentially valuable opportunity to reduce cardiac modeling. However, no transgenes have been identified as potential tethers for other mitochondrial contact sites.

27MERC formation may facilitate the interorganellar transfer of mtDNA, Ca2+, or other ions. Lavorato and colleagues reported that decreased RYR2 activity, which is associated with SR Ca2+ imbalances, increased nanotunneling without changing fission rates. This phenomenon may be connected to ATF4-mediated pathways, which are associated with mitochondrial Ca2+ homeostasis. Thus, nanotunnel formation represents a compensatory cellular response to Ca2+ dysregulation, which is further supported by the increased rate of nanotunnel formation in disease states, suggesting an interplay between impaired MERC formation and the emergence of structures representative of disease states.

28The contributions of different crista types and the involvement of mitochondrial proteins, such as the MICOS complex or fission factors, in the formation of contact sites, also remain unclear. However, the current understanding of how MERCs dictate MICOS complex orientation may help explain how MERCs can dictate cristae structure.

29mtDNA changes were first considered to be functionally linked to myopathies in the 1980s, and the role of mitochondria in medicine has continued to evolve ever since. However, despite the known association of mitochondrial structure with pathologies, it still has limited clinical applications, and the functional impacts of different mitochondrial shapes, including their relative frequencies in different disease states and tissues, remain unclear. Certain mitochondrial phenotypes, such as donut mitochondria, may serve as markers of disease, and recent studies have noted changes in the prevalence of certain mitochondrial phenotypes in response to specific pathological conditions. Frequency changes for specific mitochondrial shapes could further indicate cellular stress, inflammation, or metabolic alterations. Notably, mitochondrial mass, stress conditions, and motility dynamically affect fusion and fission rates, resulting in heterogenous micropopulations. Further elucidation of how these factors affect the drivers of fission, such as MFF or FIS1, may support clinical diagnosis.

30Different mitochondrial distributions have been established for subsarcolemmal, perinuclear, and intermyofibrillar regions of mammalian cardiac muscle, but whether these distribution differences affect mitochondrial function has not been determined. Further, various mitochondrial populations have displayed differences in relative energetics, and variations in the aging rates of tissues may indicate tissue- or population-dependent mitochondrial regulation mechanisms contributing to pathology. However, many studies have focused on murine models; a comprehensive understanding of mitochondrial structures and functions will require the characterization of mitochondrial phenotypic frequencies across various tissue and cell types from various organisms, including humans, to account for potential tissue-specific effects.

31Research must examine the processes and pathways affected by mitochondrial genomes to elucidate how ancestry affects population health studies. The presence of several different alleles within one patient, known as mtDNA heteroplasmy, may also contribute to differences in mitochondrial shape as mtDNA may affect both mitochondrial shape and organelle–organelle interactions. However, it is unclear whether mtDNA alterations affect mitochondrial shape across ancestry, race, ethnicity, and geographical order. mtDNA has a high mutation rate, and some mutations are deleterious, predisposing carriers toward certain mitochondrial diseases. Moreover, depletion of mtDNA induces mitochondria fission factors, impacts 3D formation, and changes the actin cytoskeleton. Therefore, identifying female germline mtDNA mutations, learning how these mutations become predominant, and determining the possible pathways impacted are important avenues for understanding mtDNA-related inheritable diseases and interactions with mitochondrial structures.

32A program using deep learning processes and large-input, open-source software could determine whether mitochondrial shapes contribute to disease prognosis and associated diagnosis. Future clinical applications will likely require machine learning techniques to render 3D mitochondrial structures. For example, the recently developed MitoEM program has successfully segmented mitochondrial images in both human and mouse tissues, representing a promising avenue in accelerating the speed of mitochondrial reconstruction and subsequent quantification. However, this pipeline has not been tested in clinical applications, and EM technique standardization is still lacking. Moreover, current machine learning techniques may struggle to accurately survey mitochondrial phenotypic frequencies in human biopsy tissue without a dedicated database or atlas. While drastic improvements must be made in EM techniques, not limited to validation, technique standardization, and dedicated databases for accurate phenotypic surveys in human biopsy tissues, as these technologies continue to rapidly improve, clinical applications may be a promising future avenue.

33Future studies of 3D mitochondrial structures will be facilitated by developing microscopy techniques, such as serial block-face scanning electron microscopy (SEM), correlative light EM, transmission EM, focused ion beam-SEM, cryo-ET, and mass spectrometry imaging (Box 2). Using these techniques in tandem to define protein distribution and the 3D structures of mitochondria and cristae may clarify how different protein concentrations regulate mitochondrial phenotypes, such as through specific concentrations of proteins in distinct geographies. These methods may be further combined with techniques, such as mitoRACE, that provide information on energy metabolism. Although these techniques are not yet widely used in clinical settings, they may reveal useful information about mitochondrial ultrastructure for clinical contexts.

34We believe that accurate quantification of mitochondrial shapes is essential for characterizing their roles in cellular functions and disease, and that we must define parameters for classifying distinct phenotypes, including the distances of nanotunnels and the sizes of donuts and MGs. Importantly, our understanding of mitochondrial structure must consider the potential pluralistic hallmarks of disease states (Figure 3, Key figure). Future research may explore how the presence and frequency of these structures differ between tissues and investigate the mechanisms of abnormal mitochondrial morphology formation in various diseases (see Outstanding questions for additional topics of study). For example, recent studies have shown that the MICOS complex is mechanistically implicated in cardiac dysfunction, but the clinical impacts of MICOS complex expression changes in different disease states must be further explored. In conclusion, future study of mitochondrial shape, arrangement, and heterogeneity, in addition to identifying novel regulators, will be crucial to understanding their roles in cellular homeostasis and how they are altered to respond to the physiological state and energy demands of the cell.