Among macro- to meso-scale imaging of nervous tissue, magnetic resonance imaging (MRI) remains the gold standard for non-invasive imaging in both basic research and the pathological evaluation of neurodegenerative diseases (van der Weijden et al., 2021; Paquola and Hong, 2023). To assess axon and myelin integrity, diffusion tensor imaging (DTI) and myelin water imaging/fraction (MWI/MWF) are often employed (Mori and Zhang, 2006; Lee et al., 2021). DTI parameters such as fractional anisotropy (FA) and radial diffusivity (RD) measure water molecule diffusion, providing insight into tissue structural characteristics. Although myelin restricts water diffusion perpendicular to the axon, intact myelinated axons tend to exhibit higher FA and lower RD, whereas demyelinated regions are characterized by reduced FA and increased RD (Basser et al., 1994; Conturo et al., 1999; Takagi et al., 2009; Fujiyoshi et al., 2016). A validation study employing the cuprizone-induced demyelination and remyelination model demonstrated a correlation between DTI parameters and myelin content, with RD emerging as the most sensitive myelin status indicator (Yano et al., 2017). These methods are valuable for myelin imaging owing to their sensitivity to microstructural changes; however, these metrics are not exclusively myelin-specific, as several factors influence their detection, including axonal density, fiber orientation dispersion, and extracellular water content (Friedrich et al., 2020). Conversely, myelin water imaging directly quantifies MWF, estimating the proportion of water trapped between the myelin bilayers relative to the total water signal (Mackay et al., 1994; Whittall et al., 1997), isolating the signal specifically associated with myelin water based on its distinct T2 relaxation time, offering a more myelin-specific measurement than DTI. However, its sensitivity to noise, longer scan times, and complex data processing remain challenging (Uddin et al., 2019; Paquola and Hong, 2023; Xu et al., 2025). Combining DTI’s microstructural sensitivity with MWF’s direct myelin detection provides complementary information. Active development of high-resolution in vivo MRI aims to achieve ex vivo-level resolutions, improving clinical studies and tractography (Mohammadi and Callaghan, 2021; Baadsvik et al., 2024; Bosticardo et al., 2024; Thompson et al., 2025). Track-density imaging (TDI), a novel MRI technique, produces high-resolution white matter images using diffusion MRI fiber-tracking data (Calamante et al., 2010; Figure 3). This super-resolution method utilizes long-range information from fiber tracks to generate images with enhanced spatial resolution and anatomical contrast beyond the acquired voxel size (Calamante et al., 2011). A voxel—or a “volume element”—represents a discrete unit of data on a three-dimensional (3D) grid, extending the two-dimensional (2D) pixel concept into 3D space (Thaler et al., 1978). Studies comparing TDI maps of ex vivo mouse brains with histological staining have shown good agreement, confirming that TDI provides valid anatomical information (Calamante et al., 2012).

Although macro-scale MRI has traditionally been CNS-centric, it is increasingly being applied to the PNS to provide morphological and functional insights (Chen et al., 2019). Advanced MRI metrics, particularly diffusion-weighted imaging and DTI, are applicable to the PNS, offering quantitative biomarkers for nerve fiber organization, axonal flow, and myelin integrity (Martín Noguerol et al., 2017). For instance, diffusion tensor tractography has been used to visualize the degeneration and regeneration of rat sciatic nerves, enabling the precise tracking of fiber recovery (Takagi et al., 2009). However, PNS-specific imaging faces distinct challenges compared to CNS-specific imaging. The small size and complex branching of peripheral nerves require higher spatial resolution; the presence of surrounding adipose tissue necessitates robust fat-suppression techniques. The varying orientation of nerves relative to the main magnetic field (B₀) can induce “magic angle” effects (Chappell et al., 2004), possibly confounding quantitative metrics—a limitation less prevalent in the more uniformly oriented white matter tracts of the CNS.

Positron emission tomography (PET) is another in vivo, molecular imaging modality that uses radiolabeled tracers targeting myelin-associated proteins, making it effective for evaluating CNS diseases such as multiple sclerosis. Several promising tracers have demonstrated myelin sensitivity (de Paula Faria et al., 2014; Auvity et al., 2020): stilbene derivatives, including [¹¹C]N-methyl-4,4’-diaminostilbene (MeDAS), targeting myelin basic protein (MBP) (Wu et al., 2010; van der Weijden et al., 2022, 2025), ¹⁸F-florbetaben (Matías-Guiu et al., 2015; Pytel et al., 2020), and ¹⁸F-florbetapir (Carotenuto et al., 2020; Zhang et al., 2021), and thioflavin-T derivatives, including [¹¹C]Pittsburgh Compound B (PiB), targeting the beta-sheet of MBP (Stankoff et al., 2011) and its fluorinated derivative, ¹⁸F-flutemetamol, originally developed as amyloid radioligands but now investigated as potential myelin tracer (Nelissen et al., 2009; Zeydan et al., 2022). PET offers high molecular sensitivity for quantifying myelin integrity but has lower spatial resolution than MRI and involves ionizing radiation, limiting longitudinal studies. Many researchers combine MRI’s anatomical detail with PET’s molecular specificity for a comprehensive assessment of myelin integrity and pathology (Carotenuto et al., 2020; van der Weijden et al., 2023). Regarding PNS application, although several molecular targets such as MBP are conserved, PET imaging is hindered by the small size of peripheral nerves, leading to significant partial volume effects. Additionally, non-specific uptake in surrounding tissues, including muscle and fat tissue, compromises contrast, limiting PET’s utility in peripheral neuropathies compared to that in the CNS. Recently emerging multimodal approaches, including hybrid PET/MRI using amyloid tracers, have demonstrated potential in visualizing peripheral-nerve pathology (Shouman et al., 2021), suggesting that anatomical guidance mitigates these spatial limitations.

Recent advancements in X-ray computed tomography (CT) enable non-destructive, micron-scale 3D myelin imaging, overcoming conventional contrast limitations for soft tissue. Small-angle X-ray scattering-CT provides label-free visualization by detecting scattering signals generated by the ultrastructural periodicity of myelin (De Felici et al., 2008; Jensen et al., 2011). This allows the simultaneous quantification of myelin integrity and orientation, but imaging large volumes such as entire human brain remains challenging (Georgiadis et al., 2021). X-ray phase-contrast tomography detects phase shifts for high-contrast whole-brain imaging (Chourrout et al., 2022). This method relies on simple ethanol dehydration rather than on staining to generate contrast, although high-resolution imaging requires limited-access synchrotron sources. Diffusible iodine-based contrast-enhanced CT improves contrast by using iodine to bind tissue lipids, thereby increasing the radiodensity of myelinated regions for high-throughput, 3D visualization of nerve architectures (Gignac et al., 2016; Gignac and Kley, 2018; Balcaen et al., 2023). Limitations of this method include tissue shrinkage, prolonged diffusion times for large specimens, and a lack of myelin specificity owing to non-specific lipid binding (Gignac and Kley, 2018).

For myelin imaging with LMs, toluidine blue staining of resin-embedded sections (Figures 4A,B) is a common method in research and pathology for neurodegenerative diseases, including amyotrophic lateral sclerosis and multiple sclerosis (Carriel et al., 2017; Ghnenis et al., 2018). In the standard protocol for resin embedding, the myelin sheath structure is well-preserved because glutaraldehyde strongly fixes proteins while osmium tetroxide stabilizes lipids. Osmium tetroxide effectively fixes lipids and stains myelin a dark brown to black (Figure 4C). It can be integrated into standard paraffin-embedding procedures and subsequent staining techniques, including connective tissue staining and immunohistochemistry (IHC) (Scipio et al., 2008). Conversely, conventional hematoxylin and eosin staining for the paraffin-embedded sections may not yield accurate results for myelin morphology observation because the myelin sheath is often swollen or shrunk (inset in Figure 4D) owing to the lipid loss during dehydration, as they are poorly preserved by common formaldehyde-based fixatives like formalin. For paraffin-embedded sections, myelin-specific stains such as Luxol Fast Blue staining (Figure 4E), Sudan Black B staining (Stilwell, 1957; Ineichen et al., 2017), and Klüver–Barrera (a combination of Luxol Fast Blue and Nissl staining) (Klüver and Barrera, 1953), are more frequently employed. Additionally, the haloaurophosphate complex (black gold) (Schmued and Slikker, 1999; Savaskan et al., 2009) and modified Gallyas silver staining (Gallyas, 1979; Pistorio et al., 2006) are suitable for high-resolution imaging of myelinated axons.

The teased-fiber method mechanical dissociates nerve fascicles into individual myelinated axons, enabling observation of continuous longitudinal profiles without sectioning (Figures 4F–F”). While clinically indispensable for differentiating segmental demyelination from axonal degeneration, this method is labor-intensive owing to manual preparation (Bolon et al., 2020). Combined with confocal microscopy, this method facilitates the high-resolution imaging of axon-glia interfaces and cytoskeletal organization at the nodes of Ranvier (D’Este et al., 2017; Sell et al., 2024).

Fluorescence microscopy offers higher contrast and can preserve physiological information (Figures 4G–J). Confocal laser scanning microscopy has demonstrated broad utility for in vivo and/or finer imaging and quantitative analysis of biological specimens, including myelin (Reynolds et al., 1994; Beirowski et al., 2004; Schain et al., 2014; Kwon et al., 2017). Conventionally, target structures are fluorescently labeled by IHC or transgenic techniques. Conversely, reflectance imaging, such as spectral confocal reflectance microscopy, exploits optical properties of the multi-layered myelin sheath to generate label-free visualizations based on thin-film interference (Schain et al., 2014; Kwon et al., 2017). This signal generation relies on the tightly periodic structure of compact myelin. The distinct optical signature allows for the spatial differentiation of compact myelin domains from other structural features, specifically the nodes of Ranvier and Schmidt-Lanterman incisures. Because the reflectance signal is highly sensitive to lamellar organization, this modality is particularly effective at detecting pathological changes, such as myelin decompaction or swelling, where the loss of structural integrity results in signal alteration or attenuation (Gonsalvez et al., 2019; Craig et al., 2024). Other advanced LM modalities have been utilized for non-invasive, high-resolution, label-free myelin imaging. These include coherent anti-stokes Raman scattering microscopy, providing contrast by detecting the specific molecular vibrations of C-H bonds abundant in the lipid-rich myelin sheath (Wang et al., 2005; Huff and Cheng, 2007; Freudiger et al., 2008; Poulen et al., 2021); birefringence microscopy that detects the intrinsic birefringence, an optical anisotropy arising from the highly organized lipid bilayer structure of myelin and reflecting its structural integrity and function (Blanke et al., 2021, 2023; Morgan et al., 2021); non-linear LMs such as two-photon excitation microscopy that capture intrinsic tissue autofluorescence (Canty et al., 2013; Costantini et al., 2021; Wu et al., 2021); and second- or third-harmonic generation microscopy, generating signals based on non-centrosymmetric structures or optical heterogeneities at lipid-aqueous interfaces, respectively (Farrar et al., 2011; Lim et al., 2014; Rishøj et al., 2022). Although conventional resolution is limited by diffraction to approximately 0.2 μm, super-resolution imaging systems achieve resolutions below 0.1 μm, enabling more accurate analysis with high-contrast imaging (Figure 4H; D’Este et al., 2017; Coto Hernández et al., 2022b; Chen et al., 2024). Recent advances further complement conventional 2D analysis, enabling higher resolution and larger volume 3D imaging with detailed quantitative evaluation (Velasco et al., 2021; Wu et al., 2022; Tsutsumi et al., 2023). This progress integrates these advanced microscopes with innovative sample preparations, such as tissue clearing, and sophisticated computational processing.

Several LM imaging approaches specifically target myelin. Fluoromyelin™ is a widely used and rapid dye for labeling myelin lipid components (Figures 4I,J; Monsma and Brown, 2012; Wang et al., 2019). Several fluorescent compounds that cross the blood–brain barrier have been developed for in vivo optical imaging, many with potential applications in PET imaging (Wu et al., 2006; Wang et al., 2009, 2010, 2011). Among them, near-infrared probes (Xiang et al., 2005; Wang et al., 2011) enable deeper tissue penetration and higher-resolution in vivo imaging with advanced LMs (Wu et al., 2021). The fluorescent probe Nile red can detect polarity changes in myelin lipids (considered early indicators of pathology before demyelination) when combined with spectroscopy (Luchicchi et al., 2021; Teo et al., 2021, 2025). Microscopy with an ultraviolet surface-excitation system utilizes deep ultraviolet light excitation (approximately 280 nm), penetrating tissue only superficially and effectively provides optical sectioning without the need for thin physical slicing (Fereidouni et al., 2017). This system offers rapid, label-free myelin imaging with high-intrinsic contrast, supporting 2D and 3D analyses (Kolluru et al., 2022).

Immunohistochemistry is widely used to visualize myelin morphology and profile by localizing molecular markers with tagged antibodies (Figure 4G). This section summarizes several key molecular markers for myelin and myelin-forming glial cells at the LM-level IHC.

The protein constituents of the myelin sheath serve as robust IHC markers. MBP (Omlin, 1982; Barbarese et al., 1988; Werner et al., 2013), 2’3’-cyclic nucleotide 3’phosphodiesterase (CNPase) (Gravel et al., 1997), and myelin-associated glycoprotein (Quarles, 2007) are conserved components of CNS and PNS myelin. In the CNS, the major structural proteins include proteolipid protein (PLP) (Griffiths et al., 1998; Greenfield et al., 2006), myelin oligodendrocyte glycoprotein (MOG) (Scolding et al., 1989), and Claudin-11 (Bronstein et al., 1997; Tiwari-Woodruff et al., 2001), whereas PNS myelin is enriched in myelin protein zero (P0/MPZ) (Eylar et al., 1979). The epitope-specific serology enables discrimination between compact and uncompacted myelin structures. Matsuo et al. (1997) demonstrated that a specific peptide sequence of MBP (residues 82–88 in humans) is structurally masked within healthy compact myelin but becomes exposed during structural loosening or degeneration. Antibodies targeting this specific epitope, which are often referred to as degraded MBP (dMBP), selectively label uncompacted or damaged myelin and differ from conventional anti-MBP antibodies that recognize the total myelin content. This approach provides a sensitive metric for assessing myelin integrity and white matter pathology (Ihara et al., 2010).

Myelinating glia markers are equally important. The oligodendrocyte lineage is defined by stage-specific molecular signatures: pan-lineage identity is established by the transcription factors Olig2 (Takebayashi et al., 2000; Zhou et al., 2000) and Sox10 (Stolt et al., 2002; Sock and Wegner, 2021); oligodendrocyte precursors identified by PDGFRα, the surface antigen NG2 (Nishiyama et al., 1996), and a transcription factor Nkx2.2, regulating early differentiation and is downregulated in the mature state (Qi et al., 2001; Cai et al., 2010); differentiating states are characterized by galactosphingolipid (GalC and O4) expression (Sommer and Schachner, 1981); and mature, myelinating states are characterized by CNPase and CC1 (Bin et al., 2016). The Schwann cell lineage is also well-characterized by state-dependent markers. Sox10 (Kuhlbrodt et al., 1998; Inoue et al., 1999; Britsch et al., 2001) and S100β (Stefansson et al., 1982) serve as canonical markers, while the myelination program is directed by the transcriptional regulator Egr2/Krox20 (Topilko et al., 1994; Decker et al., 2006), driving the expression of integral myelin proteins such as P0 and PMP22 (Snipes et al., 1992; Pareek et al., 1993). Conversely, the non-myelinating phenotype is characterized by high levels of glial fibrillary acidic protein (GFAP) and P75NTR (Yu et al., 2009; Jung et al., 2011). Following nerve injury, cells transform into a repair phenotype, under the control of c-Jun, re-expressing GFAP, P75NTR, and the progenitor-associated factor Sox2 (Jessen and Arthur-Farraj, 2019; Jessen and Mirsky, 2022). Clinically, the immunohistochemical detection of these markers is utilized for analyzing histopathological features in demyelinating diseases such as multiple sclerosis (Greer and Pender, 2008; Stork et al., 2020) and Charcot–Marie–Tooth syndrome (Nadol et al., 2018). Additionally, in situ hybridization for specific mRNAs, such as MBP, PLP, and MOG, is used to investigate the state of myelin-forming glial cells (Breitschopf et al., 1992; Michel et al., 2015).

Although static analyses with IHC are valuable for observing myelin in research and pathological diagnosis, they are often insufficient for elucidating the dynamic regulation of myelination. Transgenic mouse strains expressing fluorescent reporters (such as green fluorescent protein) under cell-specific promoters are powerful tools for visualizing cell morphology and tracking cellular dynamics in real time, significantly advancing our understanding of nervous system function, including myelin biology.

Examples include PLP-EGFP mice for visualizing oligodendrocyte lineage differentiation and myelination (Mallon et al., 2002), CNPase-EGFP mice for detailed myelination observation in the CNS and PNS (Yuan et al., 2002; Deng et al., 2014). Additionally, Sox10-Venus mice exhibit a robust Venus signal, enabling direct imaging of oligodendrocytes and Schwann cells even in live or unfixed tissue (Figures 4H–J; Shibata et al., 2010; Mohan et al., 2019). Furthermore, crossing Cre-driver lines such as Wnt1-Cre (Danielian et al., 1998), P0-Cre (Yamauchi et al., 1999), MBP-Cre (Kolb and Siddell, 1996), PLP-Cre (Doerflinger et al., 2003), and Sox10-Cre (Stine et al., 2009) with Cre-dependent reporter strains (Kawamoto et al., 2000) provides powerful tools for tracing the dynamics of myelinating glial cells.

Electron microscopies are indispensable tools in neuroscience research for observing the ultrastructures of myelin, neurons, and synapses, offering nanoscale resolution beyond the diffraction limit of LMs. The two main EM modalities are scanning EM (SEM) and transmission EM (TEM). SEM generally provides high-resolution 3D surface views by detecting electrons scattered from a sample’s surface. Techniques such as the modified KOH-collagenase (Miller et al., 1982; Ushiki and Ide, 1986, 1987, 1988) and osmium maceration methods (Koga and Ushiki, 2006; Figure 5A) enable unique views of the ultrastructural anatomy of nervous tissue components, including the myelinated and unmyelinated axons, nodes of Ranvier, and surrounding cellular components with their structural associations (Nomura et al., 2013; Bando et al., 2015). TEM is generally employed for 2D imaging by detecting electrons transmitted through ultrathin-sliced (usually 50–80 nm in thickness) samples. The samples are usually embedded within a resin and sectioned using a diamond knife. TEM offers superior resolution and contrast, facilitating detailed visualization of fine intracellular microstructures (Figure 5B). For instance, the periodicity of the multilamellar myelin sheath can be clearly visualized using TEM at magnifications > 50,000x (Figure 5B right). However, TEM imaging is limited to a field of view of a few square millimeters, as the sections are mounted on metal grids with a maximum diameter of 3 mm.

Given the high-resolution capacity of EM with a high degree of precision, appropriate sample preparation, particularly rapid fixation, is crucial to avoid myelin-related artifacts (Wang et al., 2022). In the conventional specimen preparation of nervous tissues, 2.5% glutaraldehyde is typically used for primary fixation, followed by 1% osmium tetroxide for post-fixation. As mentioned earlier, osmium tetroxide effectively fixes the lipid in myelin. Although 0.1 M cacodylate buffer (pH 7.4) is frequently used, phosphate buffer reportedly yields comparable or, sometimes, superior outcomes (Liang et al., 2021). Conventional chemical fixation relies on the diffusion of fixatives, which can induce osmotic changes and shrinkage and thus potentially distort the myelin ultrastructure. High-pressure freezing (HPF) followed by freeze-substitution has emerged as alternative method (Möbius et al., 2016). By physically immobilizing tissue within milliseconds, HPF prevents ice crystal formation, transforming cellular water into an amorphous, glass-like state in a process known as vitrification. This eliminates osmotic artifacts, faithfully retaining axonal circularity and the native compaction of myelin lamellae. A major limitation of this method is that effective vitrification is physically restricted to a depth of approximately 200 μm from the surface. This technique is unsuitable for large tissue volumes because it requires immediate, precise dissection or slicing prior to freezing, which can introduce mechanical damage if not performed with extreme care.

Although SEM is traditionally used for surface imaging, it is now frequently applied for the observation of ultrathin sections collected on conductive flat materials such as silicon wafer and glass slide (Figure 5C). Although the resolution of SEM remains lower than that of TEM, this approach offers distinct advantages, including the ability to collect a much larger number of serial section collections, a wider field of view, and greater section durability for repeated observation.

This capacity for large-area imaging is dramatically enhanced by multibeam SEM, which simultaneously utilizes multiple electron beams (61 or 91 beams) to acquire centimeter-scale images at nanometer-scale resolution with unprecedented speed (Eberle et al., 2015). By stitching thousands of high-resolution images, this method provides a seamless view from the tissue level down to the ultrastructural level (Figure 5D). This high-throughput, nanometer-scale imaging capability substantially reduces the time required for large-area acquisition, facilitating brain connectomics research across unprecedented volumes (Shapson-Coe et al., 2024; Wu et al., 2024).

3D-reconstruction volume analysis from serial EM images is a powerful approach for understanding ultrastructural anatomy. In the traditional array-tomography method, manually collected serial ultrathin sections are observed by general TEM or SEM. This method does not require specialized equipment beyond a conventional ultrathin sectioning device and EMs; however, significant technical expertise is required, and the risk of losing sections due to human error remains unavoidable. The automated tape-collecting ultramicrotome (ATUM)-SEM method overcomes this by automatically collecting serial sections onto a conductive tape, greatly simplifying the creation of large section libraries (Schalek et al., 2012; Hayworth et al., 2014). The conductive tape facilitates high-resolution and wide-area imaging by preventing charging. The sections can also be post-stained and re-imaged. This system can reliably collect thousands of serial sections, even at thicknesses of ≤50 nm. This process would be challenging, even for a skilled technician. A disadvantage of this method is that precise computational image alignment of the image series before 3D reconstruction is time-consuming.

Two other techniques, serial block-face (SBF)-SEM and focused ion beam (FIB)-SEM, generate inherently aligned image stacks by iteratively imaging sample block surfaces and subsequently removing a thin layer to expose a new surface. In SBF-SEM, a diamond knife removes layers within the specimen chamber, enabling larger-volume reconstructions exceeding hundreds of micrometers (Denk and Horstmann, 2004). In FIB-SEM, an ion beam mills the surface, offering superior z-resolution in a few nanometers, roughly one-tenth the thickness of a physical ultrathin section, enabling the isotropic analysis (Heymann et al., 2006; Knott et al., 2008; Xu et al., 2017). However, the field of analysis is smaller than that of other methods, practically limited to an area of approximately 100 μm square. A key advantage of SBF- and FIB-SEM is that they can acquire inherently aligned serial z-stack images without loss, greatly simplifying the 3D reconstruction process. However, a common drawback is that tissues are consumed with each imaging cycle. Moreover, preparing samples with optimal conductivity by additional procedures, such as en bloc conductive staining, is essential. Recent 3D analysis approaches combining ATUM and FIB efficiently revealed ultrastructures in the targeted area at isotropic resolution (Kislinger et al., 2020). Notably, a region of interest is first identified within the extensive ATUM section library and subsequently targeted for high-resolution, isotropic 3D imaging using FIB-SEM.

Volume EM strategies have been extensively employed in CNS research to resolve complex 3D neural circuits and neural-glial interactions. In the PNS, although the architecture is generally simpler, Schadt et al. (2025) highlighted that 3D volumetric analysis is fundamentally required to resolve the true ultrastructural integrity of the axon-myelin unit, uncovering pathologies such as myelin outfoldings that are ambiguous in 2D cross-sections. In contexts such as nerve regeneration, Leckenby et al. (2019) demonstrated long-distance tracking to capture complex axonal behaviors and trajectories over large spatial ranges. The application of volume EM in the PNS is specifically driven by the need to visualize longitudinal complexities, including pathological ultrastructures and extended regenerative pathways.

Immunoelectron microscopy is a powerful technique for elucidating the precise subcellular localization of target molecules within their ultrastructural context, thereby highlighting their specific functions (Nordengen et al., 2015; Weil et al., 2016). The widely applied immuno-gold method visualizes antibody-bound targets as electron-dense nanometer-scale colloidal gold particles, enabling simultaneous observation of the labeled molecule and the surrounding cellular architecture. A major challenge in iEM is the need to preserve both antigenicity and ultrastructures, leading to the development of several protocols. In the pre-embedding protocol, target antigenicity is well-preserved because the immunoreaction is processed before the typical EM protocol, including osmium fixation, dehydration, and embedding that may denature antigens (Polishchuk and Polishchuk, 2019; Tao-Cheng et al., 2021). Ultrastructures are fairly well-preserved if permeabilization is adequately mild, while excessive treatment easily causes the artificial collapse of many structures, such as plasma membranes comprising myelin lamellae. The post-embedding protocol is excellent for preserving ultrastructure, as samples undergo full EM preparation for hydrophilic resin embedding before immunolabeling. Loss or masking of antigens during the harsh embedding process often greatly reduces labeling efficiency. Samples for these two methods are typically prepared using standard EM protocols with modifications for IHC. Conversely, the cryosection iEM protocol such as the Tokuyasu method avoids resin embedding, effectively resolving the trade-off between ultrastructural preservation and antigenicity (Tokuyasu, 1973; Möbius and Posthuma, 2019). Additionally, by using high-pressure freezing to vitrify samples, this approach minimizes chemical cross-linking and dehydration and thereby preserves antigenicity (van Donselaar et al., 2007). Furthermore, this approach allows for efficient labeling within a native-like myelin architecture free from shrinkage artifacts. However, this technique requires specialized equipment for sample preparation, including a high-pressure freezer and cryo-ultramicrotome. Furthermore, different from pre- and post-embedding protocols, the cryo-approach necessitates a dedicated workflow starting from fresh tissue; thus, it cannot be applied retrospectively to conventionally fixed samples.

Correlative light and electron microscopy combines the advantages of LMs and EMs, allowing the mapping of molecular or functional information with ultrastructural context. While iEM visualizes target molecules directly with nanometer precision, it is essentially EM restricted to small fields of view. The workflow leverages the primary strength of LM, which is the ability to identify specific, fluorescently-labeled structures across large fields of view even under physiological conditions, to guide subsequent EM imaging of the same region of interest. Although iEM and CLEM are distinct approaches, the use of fluorescence-colloidal gold dual-conjugated antibodies can effectively bridge these approaches. This specific method allows researchers to identify regions of interest via fluorescence in the microscale field of LM and to subsequently visualize the direct localization of target molecules via gold particles in EM, thereby utilizing iEM and CLEM in a complementary manner.

A distinct advantage of CLEM, particularly when using genetically encoded fluorescent markers to label specific cells or organelles, is the superior preservation of ultrastructure. Unlike pre-embedding immunolabeling, this approach eliminates the need for antibody reactions, rendering harsh treatments such as antigen retrieval or permeabilization unnecessary.

During CLEM process, the fluorescence quenching that occurs during the preparation of the electron microscopy (EM) sample, including osmium fixation and resin embedding, poses a persistent challenge, particularly for conventional antibody-based staining. In order to address this challenge, the utilization of osmium-resistant fluorescent proteins (Tanida et al., 2020) and proximity labeling with biotin ligases (Sanada et al., 2022) enables “in-resin CLEM,” a technique designed to retain fluorescence signals even after full EM processing. These advancements also suggest a potential solution for a technical challenge in conventional CLEM: the accurate correlation of images due to the discrepancy in z-axial resolution between LM and EM. While simple 2D overlays are often insufficient for identifying small, sub-micron structures, and precise correlation typically requires complex 3D correlation strategies, in-resin CLEM offers a compelling alternative strategy. By detecting fluorescence directly from the ultrathin sections for EM, this method theoretically eliminates the z-axis mismatch, facilitating precise superposition.

The EM component of a CLEM experiment can employ several optional methods depending on the aims of the investigation. The multibeam SEM is a method that facilitates CLEM workflows across areas spanning several square centimeters (Shibata et al., 2019). 3D-CLEM extends this approach to volume imaging, enabling the analysis of complex, fluorescently-labeled neurons and axons (Luckner et al., 2018; Booth et al., 2019). This approach enables powerful quantitative analyses by correlating functional or molecular signals from LM with high-resolution 3D morphological data from EM.