Volume electron microscopy in axon regeneration research: insights into mitochondria, endoplasmic reticulum, and membrane contacts

Abstract

Analysing organelle structures in tissue samples that preserve the in vivo environment, rather than in isolated cells, can provide valuable information on the mechanisms underlying nerve recovery and repair. However, determining organelle ultrastructure within tissue samples remains challenging with conventional microscopic techniques. To address this limitation, volume electron microscopy, particularly focused ion beam/scanning electron microscopy (FIB/SEM), has provided novel insights into organelle morphology, distribution, and membrane contacts, in three-dimensions and even within intact tissues. This review highlights the application of FIB/SEM for exploring the three-dimensional organization of mitochondria in motor neuron cell bodies and along the axon initial segments (AIS), where simultaneous investigation of intracellular and extracellular environments is difficult using other approaches. These analyses have revealed novel findings regarding mitochondrial distribution under healthy conditions and its dramatic alteration following injury, as well as microglial attachment around the AIS. Furthermore, FIB/SEM has enabled detailed characterization of the complex endoplasmic reticulum (ER) architecture within motor neuron cell bodies. Three-dimensional reconstructions have demonstrated a distinct uneven distribution of the ER in healthy neurons and revealed disruption of this organization following injury. In addition, ER–plasma membrane (ER–PM) contacts have been characterized as sheet-like structures, and quantitative analyses have shown significant increases in ER–PM contacts after injury. The novel findings obtained through FIB/SEM provide new perspectives on the cellular mechanisms underlying neuroregeneration and highlight the value of volume electron microscopy in advancing our understanding of nerve repair processes.

1. Introduction

Peripheral nerve injury is well known to recover after injury, whereas recovery following central nervous system (CNS) injuries, such as spinal cord injury, remains limited (He and Jin, 2016). Understanding the mechanisms underlying peripheral nerve regeneration could provide valuable insights into approaches for promoting CNS neuroregeneration. In particular, surrounding factors, such as the extracellular matrix and immune cells, strongly influence nerve recovery processes. In this context, tissue-level analysis is important for understanding these mechanisms without cell isolation.

Organelle responses are crucial for nerve recovery. For example, mitochondria frequently undergo structural changes, such as fusion and fission, to regulate energy production, redox homeostasis, and cellular apoptosis (Charrasse et al., 2024; Youle and van der Bliek, 2012; Giacomello et al., 2020; Friedman and Nunnari, 2014; Preminger and Schuldiner, 2024). Disruption of these regulatory mechanisms can trigger several pathological conditions, including neurodegenerative diseases (Berman et al., 2009; Bilsland et al., 2010; Chan, 2012; Knott A. B. et al., 2008; Sirois et al., 2026). In particular, mitochondrial transport between the cell body and axon is required for nerve recovery; therefore, the mechanisms regulating mitochondrial transport and the accompanying morphological changes are important research targets in axon regeneration (Cho et al., 2009; Kiryu-Seo et al., 2010; Wang et al., 2021; Kiryu-Seo et al., 2016). The endoplasmic reticulum (ER) also exhibits a close relationship between morphology and function (Kors and Schlaitz, 2024; Friedman et al., 2010; Sawyer et al., 2024; Voeltz et al., 2006; Schroeder et al., 2019; Bragulat-Teixidor et al., 2024; Westrate et al., 2015; Puhka et al., 2012; Heinrich et al., 2021; Obara et al., 2023; Arruda and Parlakgül, 2023). The ER forms a continuous membrane network and plays essential roles in cellular homeostasis. Disruption of ER morphology, has been implicated in various diseases, highlighting the close relationship between ER structure and function (Westrate et al., 2015). Furthermore, membrane contact sites (MCSs) between the ER and other organelles, including mitochondria and the plasma membrane (PM), regulate crucial processes related to signal transduction and membrane dynamics (Miyazono et al., 2018). Disruption of MCSs has been associated with diverse neuronal pathologies (Watanabe et al., 2016; Saheki and De Camilli, 2017; Prinz et al., 2020).

One of the latest methodological trends in studying organelle morphology is fluorescence-based imaging with super-resolution microscopy (Nixon-Abell et al., 2016; Goujon et al., 2019; Schroeder et al., 2019). These methods provide a vast amount of information, particularly for isolated or cultured cells. However, for studying nerve recovery mechanisms, such samples may not fully reflect in vivo conditions because cell isolation dramatically alters the surrounding environment and may activate numerous physiological responses within the cells themselves. To obtain a more accurate understanding of neuroregeneration in vivo, electron microscopy (EM), particularly volume electron microscopy (vEM), is a highly suitable tool (Tamada, 2023), as it provides three-dimensional ultrastructural information without requiring isolation from surrounding tissues.

vEM (Denk and Horstmann, 2004; Merchan-Perez et al., 2009; Ohta et al., 2012; Hayworth et al., 2015; Narayan and Subramaniam, 2015; Xu et al., 2017; Peddie et al., 2022; Tamada, 2023), including focused ion beam/scanning electron microscopy (FIB/SEM) (Knott G. et al., 2008; Xu et al., 2021; Heinrich et al., 2021; Denk et al., 2012), the serial block face–SEM (SBF-SEM) (Denk and Horstmann, 2004; Wilke et al., 2013), and array tomography (Hayworth et al., 2015; Morgan et al., 2016), acquires multiple electron micrographs that are stacked to generate three-dimensional ultrastructural reconstructions. Among the techniques, FIB/SEM is particularly suitable for studying organelle architecture because its z-resolution, achieved through gallium ion beam milling, enables accurate reconstruction of sheet-like structures and membrane contact sites.

In this review, we focus on novel insights into mitochondria, the ER, and ER–PM contacts in mouse hypoglossal and sciatic motor neurons following peripheral nerve transection, a model of recoverable peripheral nerve injury, using FIB/SEM (Tamada et al., 2017; Tamada et al., 2021; Elgendy et al., 2024). vEM analysis of organelles overcomes the limitations of conventional transmission EM (TEM) analysis (Puhka et al., 2012; Terasaki et al., 2013; Zamponi et al., 2022), which provides only two-dimensional ultrastructural information. Consequently, vEM enables quantitative assessment of geometrical parameters, such as organelle size, number, and spatial distribution. Based on these three-dimensional findings, we discuss new perspectives on the mechanisms underlying neuroregeneration after injury.

2. vEM analysis of mitochondria in axon regeneration models

2.1. Mitochondrial distribution in the cell body and the AIS

As described in the previous section, mitochondria exhibit a variety of morphologies, including elongated, spherical, and branched forms, particularly within neuronal cell bodies. They also undergo dynamic morphological changes, termed fission and fusion, in response to cellular demands. Mitochondrial fragmentation following neuronal injury has been extensively studied in axons and is considered crucial for the efficient and rapid delivery of energy and Ca²⁺ buffering capacity to regenerating axon tips (Cho et al., 2009; Kiryu-Seo et al., 2010; Wang et al., 2021; Kiryu-Seo et al., 2016). Although pathologically enlarged mitochondria (Knott A. B. et al., 2008; Waterham et al., 2007; Kiryu-Seo et al., 2016) and sparsely distributed mitochondria in axons have been observed by live-cell light microscopy (Kiryu-Seo et al., 2016; Tosolini et al., 2024) and electron tomography (Hoffmann et al., 2021), the numerous mitochondria present in neuronal somata and those of normal size generally require EM for detailed visualization. Furthermore, accurate assessment of mitochondrial morphology and distribution, including fragmentation status and density, requires three-dimensional ultrastructural analysis using vEM.

The quantitative data from FIB/SEM analysis revealed no significant differences in mitochondrial morphology or size within neuronal cell bodies following injury when intact and axon-injured neurons were compared (Tamada et al., 2017). Because these findings differed from the expectation that mitochondrial fragmentation would be prominent in injured neurons, attention shifted to another region, the axon initial segment (AIS), which is morphologically defined as the region between the axon hillock and the first myelinated segment. The AIS is a highly specialized domain characterized by distinct cytoskeletal organization and the expression of specific membrane channels and receptors that function in action potential initiation and as a boundary between the cell body and axonal compartments (Rasband, 2010; Kuba et al., 2010; Ogawa and Rasband, 2008; Garrido et al., 2003; Pan et al., 2006; Zhou et al., 1998; Leterrier, 2018; Nelson and Jenkins, 2017; Garrido, 2026). Although some TEM studies have reported the presence of mitochondria within the AIS (Dimova and Markov, 1976; Li et al., 2004; Sasaki et al., 2005), single ultrathin sections are insufficient to accurately determine their distribution because the AIS extends for approximately 25 μm. FIB/SEM analysis enables visualization of entire mitochondrial structures and their distribution throughout the full length of the AIS (Tamada et al., 2021).

Surprisingly, FIB/SEM demonstrated that healthy motor neurons contain almost no mitochondria within the AIS, despite the abundance of mitochondria in the axon hillock, similar to that observed in the soma, and their reappearance in the first myelinated region (Figure 1A) (Tamada et al., 2021). In contrast, following axonal injury, large numbers of mitochondria were present within the AIS at densities comparable to those observed in other neuronal regions (Figure 1B). In parallel, FIB/SEM analysis revealed direct attachment of microglia to the AIS membrane without the interposition of other cellular elements (Figure 1C) (Tamada et al., 2021). These findings suggest that the membrane environment of the AIS changes following injury and becomes more similar to that of the neuronal cell body, as the specialized membrane organization of the healthy AIS, including its unique channels and scaffolding proteins, may normally restrict cellular adhesion.

Figure 1.

(A) Characteristic mitochondrial distribution around the axon initial segment (AIS) in healthy motor neurons. Mitochondria were observed in the soma, axon hillock, and myelinated regions (green). In contrast, mitochondria were largely absent from the AIS. (B) Altered mitochondrial distribution in the AIS following injury. In injured motor neurons, numerous mitochondria were distributed throughout the AIS. Their distribution pattern was similar to that observed in the soma, axon hillock, and myelinated regions. (C) Microglial association with the AIS following injury, corresponding to the same region shown in panel B. The C1 and C2 panels show representative FIB/SEM images demonstrating the absence of intervening cellular elements between the AIS and microglia. The C3 panel shows a three-dimensional reconstruction of the soma, AIS, microglia, and surrounding cells. Yellow: microglia. Scale bar 5 μm. A, B, C1 and C2 are modified from Tamada et al. (2021).

2.2. Interpretation of the findings

Using FIB/SEM analysis of the soma in peripherally injured motor neurons, a predominance of fragmented mitochondria could not be detected, which was one of the notable findings of this study. To date, the critical position of mitochondrial fragmentation during axonal repair has not been fully elucidated. Recent studies have suggested that the balance between mitochondrial fusion and fission, rather than fragmentation alone, is important for effective mitochondrial transport within axons (Au et al., 2022).

The marked alterations in mitochondrial distribution, from near exclusive in the healthy AIS to substantial accumulation following injury, are particularly noteworthy. One possible explanation for this influx of mitochondria is the disruption of the specialized molecular organization of the AIS, as indicated by microglial reactions and depletion of the cytoskeletal protein ankyrin G (AnkG) following injury (Tamada et al., 2021). Injury-induced degradation of AnkG has been associated with the loss of neuronal polarity and the expression of microtubule-associated protein (MAP2), a dendritic marker (Schafer et al., 2009). Under these conditions, somatodendritic proteins and organelles can enter the axon because of impaired polarity maintenance (Hedstrom et al., 2008; Teliska et al., 2022). Recently, similar phenomena, including mitochondrial exclusion from the healthy AIS and mitochondrial accumulation within the AIS under conditions of AIS disruption, have been reported in both Drosophila neurons (Wodrich et al., 2024) and human iPSC-derived neurons (Tjiang and Zempel, 2022). Although the mechanisms responsible for mitochondrial exclusion from the healthy AIS remain unclear (Yang et al., 2023), the accumulation of mitochondria under pathological conditions may reflect a protective cellular response.

However, mitochondria are continuously transported between the cell body and axon even under healthy conditions. Therefore, the differences observed between healthy and injured neurons may reflect changes not in mitochondrial “motility” itself but in the proportion of “stationary” mitochondria within the AIS. If the mitochondria accumulated in the injured AIS are predominantly stationary, they may contribute to buffering cytosolic Ca²⁺ levels, which are likely elevated following injury. The relationship between increased intracellular Ca²⁺ levels and AnkG degradation has been well documented and involves calpain, a calcium-dependent cysteine protease (Ma, 2013; Schafer et al., 2009; Del Puerto et al., 2015; Varadarajan et al., 2022). Furthermore, Benusa et al. (2017) described a mechanism by which microglia surrounding the AIS induce Ca²⁺ upregulation and activity, accompanied by calpain-dependent AnkG degradation. The presence of microglia around the AIS, as revealed by FIB/SEM analysis, is consistent with this hypothesis. Additional studies have reported that increased microglial activity influences the AIS under injury and inflammatory conditions, including those observed in axotomy models (Clark et al., 2016; Kato et al., 2016; Li et al., 2012). Although the functional significance of microglial contacts with the AIS remains controversial, differences in microglial subtypes and the roles of AIS-generated action potentials in regulating neuronal function may contribute to these varying observations (Baalman et al., 2015; Gallo et al., 2022; Que et al., 2024; Jenkins and Bender, 2025).

3. vEM analysis of ER and ER–PM contacts in axon regeneration models

3.1. Morphological findings of ER and ER–PM contact alterations after injury

The complex organization of the ER makes it difficult for light microscopy to resolve detailed membrane structures, particularly within neuronal cell bodies. Most current knowledge regarding pathological alterations in ER morphology and ER–PM contacts has been derived from isolated cultured mammalian cells, C. elegans, yeast, and other model systems, in which extensive ER stacking is rarely observed (Calì and Brini, 2021; Nakatsu and Tsukiji, 2023; Gamuyao and Chang, 2024). As described below, FIB/SEM analysis has provided novel three-dimensional insights into ER architecture and distribution while preserving native cell–cell interactions (Bharathan et al., 2023).

In the cell bodies of healthy motor neurons, FIB/SEM revealed an uneven ER distribution between the perinuclear region and the cell periphery near the plasma membrane (Figure 2A) (Elgendy et al., 2024). The ER surrounding the nucleus exhibited a lamellar structure, which is thought to correspond to Nissl bodies (tigroid substance). In contrast, the peripheral ER displayed a mesh-like organization, with a significantly smaller volume fraction than that observed in the perinuclear region. Following injury, these lamellar structures collapsed and were replaced by a more uniformly distributed mesh-like network (Figure 2B) (Elgendy et al., 2024). These qualitative changes were further supported by quantitative analyses of branching points associated with mesh formation.

Figure 2.

(A) Representative three-dimensional reconstruction of a motor neuron cell body. ER sheets (arrowheads) were abundant in the perinuclear region, whereas sparse ER tubules were observed in the peripheral region. Plasma membrane (PM) is indicated by blue dotted lines. N, nucleus. (B) Representative three-dimensional reconstruction of an injured motor neuron. The ER sheet structure was less distinct in the perinuclear region, and the ER was distributed more uniformly regardless of its proximity to the nucleus or PM (blue dotted lines). N, nucleus. (C) Three-dimensional reconstruction of ER–PM contacts in the motor neuron cell body. Red areas indicate individual ER–PM contact sites. The ER was attached to the PM (blue) in a patch-like manner. (D) ER–PM contacts in an injured motor neuron. Red sheets indicate ER–PM contact sites. Their surface areas were larger than those observed in healthy motor neurons (panel C). PM, blue. Scale bar 2 μm. These images are modified from Elgendy et al. (2024).

Simultaneously, vEM analyses clearly visualized ER–PM contacts as sheet-like structures, enabling quantitative measurement of their surface area (Wu et al., 2017). ER–PM junctions vary substantially among cell types and physiological states with respect to abundance, molecular composition, and the extent of membrane apposition to the PM (Chang et al., 2017; Chung et al., 2022; Deardorff et al., 2014; Wu et al., 2017). Analysis of the hypoglossal axotomy model demonstrated that ER–PM contacts increased significantly, from approximately 3% in healthy motor neurons to 7% 1 week after injury (Figures 2C,D), indicating a marked enhancement of ER–PM interactions during the recovery process (Elgendy et al., 2024).

3.2. Interpretation of the findings

ER stress occurs under various forms of cellular stress, including axonal injury, and numerous studies have reported a close relationship between ER stress and alterations in ER morphology (Walter and Ron, 2011; Hetz and Saxena, 2017; Oñate et al., 2016; Wang and Kaufman, 2016; Öztürk et al., 2020). In the context of regenerative responses following axotomy, ER remodelling may play important roles not only at sites of axonal elongation but also within neuronal cell bodies (Lee et al., 2019; Yang et al., 2020; Carvalhais et al., 2026). Morphological alteration from a sheet-based to a tube-based structure may function as a resilience factor in hazardous situations by facilitating the transfer of intracellular molecules (Elgendy et al., 2024).

Several hypotheses may explain the increase in ER–PM contacts following injury. First, rapid and efficient lipid exchange between the ER and PM may be required to support membrane repair and replenishment (Guillén-Samander and De Camilli, 2023). In the hypoglossal axotomy model, expression of extended synaptotagmin-1 (E-Syt1) was significantly increased during the recovery phase (Elgendy et al., 2024). E-Syt1 is an important ER–PM tethering protein involved in lipid transfer between the ER and PM (Giordano et al., 2013; Jeyasimman and Saheki, 2020; Reinisch and De Camilli, 2016; Toulmay and Prinz, 2012; Saheki et al., 2016). In yeast, several stress conditions induce non-vesicular lipid transfer from the ER to the PM through regulation of tricalbin, the orthologue of E-Syt proteins (Thomas et al., 2022; Mu et al., 2025). Furthermore, previous studies suggest that the ER contributes membrane components required for axonal elongation (Rao et al., 2016; Zamponi et al., 2022), although information regarding similar mechanisms within neuronal cell bodies remains limited. A second possibility is that elevated intracellular Ca²⁺ levels contribute to the increase in ER–PM contacts. As discussed in the mitochondrial section of this review, intracellular Ca²⁺ concentrations may increase following injury. E-Syt1 is strongly regulated by Ca²⁺ because its C2 domains, which mediate interactions with the plasma membrane phosphatidylinositol 4,5-bisphosphate (PI(4,5)P₂), are activated by elevated cytosolic Ca²⁺ concentrations (Giordano et al., 2013; Jeyasimman and Saheki, 2020; Reinisch and De Camilli, 2016; Saheki et al., 2016).

From the perspective of Ca²⁺ regulation, quantitative geometric information regarding ER architecture, including the distance between the ER and PM, represents a particularly promising area for future vEM studies. Such parameters are likely to be important because Ca²⁺ signalling is highly dependent on the spatial relationships between Ca²⁺ sources, receptors, and Ca²⁺-activated ion channels (Tamada and Iino, 2025; Denizot et al., 2026; Tamada, 2026).

4. Future perspectives and current limitations

Although this review focused on mitochondria and the ER, numerous other factors contribute to the regulation of axonal structure and function (Smith et al., 2023). In particular, microtubules play essential roles in intracellular transport and organelle morphology. vEM has the potential to simultaneously visualize and analyse these diverse cellular components within their native tissue environment. However, for some targets, the effects of fixative artifacts resulting from conventional aldehyde-based fixation methods must be considered, as these procedures may alter ultrastructural features. As an alternative approach, cryofixation using high-pressure freezing can be employed (Tamada et al., 2020), although this method is not applicable to all sample types.

One of the current advances in vEM methodology is the development of correlative light and electron microscopy (CLEM) (Müller-Reichert and Verkade, 2012). This approach integrates information on protein localization and gene expression obtained by light microscopy with ultrastructural data acquired by EM, enabling a more comprehensive understanding of the relationships between cellular function and morphology (Ohta et al., 2021; Hayashi et al., 2023). Furthermore, rapid progress in deep-learning-based image analysis is expected to accelerate vEM research. Recent algorithms can generate near-isotropic image datasets from anisotropic data through denoising and super-resolution techniques (Lu et al., 2024). In addition, advances in automated image segmentation have the potential to reduce one of the major bottlenecks in vEM analysis, namely the burden of segmentation for three-dimensional structures from serial image datasets (Conrad and Narayan, 2023; Müller et al., 2024). Together, these technological developments are expected to generate large-scale datasets linking cellular ultrastructure with molecular and functional information, thereby advancing our understanding of neural regeneration mechanisms.

5. Conclusion

Recent advances in vEM have enabled three-dimensional ultrastructural analysis of tissue samples at high resolution. Using this approach, previously unrecognized patterns of mitochondrial and ER morphology and distribution have been identified in healthy motor neurons, as well as substantial alterations following injury in the context of surrounding cellular elements. Such observations are particularly important for understanding nerve recovery mechanisms because intracellular and extracellular factors interact in a highly coordinated manner during regeneration. These novel morphological insights complement conventional molecular and cellular approaches and provide new opportunities to advance the study of neuroregeneration.

Funding Statement

The author(s) declared that financial support was received for this work and/or its publication. This work was supported by the Takeda Science Foundation 2022055542.

Author contributions

HT: Writing – original draft, Writing – review & editing.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. Paperpal was used for spelling checks and grammar corrections.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.