Immunoelectron microscopy: a comprehensive guide from sample preparation to high-resolution imaging

Abstract

Immunoelectron Microscopy (IEM) is a technique that combines specific immunolabeling with high-resolution electron microscopic imaging to achieve precise spatial localization of biomolecules at the subcellular scale (< 10 nm) by using high-electron-density markers such as colloidal gold and quantum dots. As a core tool for analyzing the distribution of proteins, organelle interactions, and localization of disease pathology markers, it has irreplaceable value, especially in synapse research, pathogen-host interaction mechanism, and tumor microenvironment analysis. According to the differences in labeling sequence and sample processing, the IEM technology system can be divided into two categories: the first is pre-embedding labeling, which optimizes the labeling efficiency through the pre-exposure of antigenic epitopes and is especially suitable for the detection of low-abundance and sensitive antigens; the second is post-embedding labeling, which relies on the low-temperature resin embedding (e.g., LR White, Lowicryl) or the Tokuyasu frozen ultrathin sectioning technology, which can improve the deep-end labeling while maintaining the ultrastructural integrity of the tissue. The accessibility of deep antigens is enhanced while maintaining ultrastructural integrity. The two techniques have significant complementarities: the former has high labeling efficiency but limited cellular structure preservation, while the latter has better tissue structure preservation but needs to balance the problems of resin penetration and antigenic epitope masking. This article provides a systematic analysis of the entire IEM workflow, focusing on the synergistic strategies for fixation and dehydration, experimental method selection, and specific application cases. It also introduces a quantitative analysis framework based on systematic random sampling (SUR) and deep learning algorithms (such as Gold Digger), including FIB-SEM 3D reconstruction (with isotropic resolution reaching 5 nm) and correlative light and electron microscopy (CLEM) multimodal integration strategies for functional-structural co-localization. Through technological innovation and cross-platform integration, IEM is driving the advancement of ultrastructural pathology diagnostics and precision nanomedicine to new heights.

Introduction

Immunoelectron Microscopy (IEM) combines the ultrastructural imaging capability of electron microscopy with the molecular specificity of immunolabeling. By conjugating electron-dense markers—such as colloidal gold nanoparticles—to antibodies, IEM enables precise localization of proteins, pathogens, or macromolecular complexes at the nanoscale, with resolutions reaching up to 0.5 nm. This technique bridges the gap between molecular functional studies and ultrastructural analysis, providing spatial information that surpasses the diffraction limit of light microscopy (~ 200 nm) and even that of super-resolution fluorescence methods, such as STED and PALM/STORM (typically ~ 20–50 nm).

Currently, IEM strategies are mainly categorized into two approaches based on the timing of immunolabeling relative to sample processing. Pre-embedding immunolabeling involves antibody incubation before resin embedding, allowing for efficient labeling of membrane-associated antigens; however, permeabilization can compromise the ultrastructure. Post-embedding immunolabeling is performed on ultrathin sections after embedding, which better preserves cellular morphology but may suffer from antigen epitope masking due to fixation and embedding.

Regardless of the method, a central challenge in IEM is to balance ultrastructural preservation with antigenicity retention. To overcome this, researchers are integrating cryo-electron microscopy techniques—such as high-pressure freezing coupled with freeze substitution—to maintain native antigen conformation and combining IEM with correlative light and electron microscopy (CLEM) to link dynamic live-cell imaging with static ultrastructural information. Furthermore, advances in novel nanoprobes (e.g., DNA-guided assembly of gold particles) and AI-assisted ultrastructural quantification are driving IEM toward greater sensitivity and automation. These developments continue to expand their applications in frontier fields such as viral replication mechanisms and nanoscale mapping of neuronal synapses.

Development history of immunoelectron microscopy

In 1931, German scientists Knoll and Ruska developed the first transmission electron microscope; the use of electron beam imaging opened a new era of exploration of the microscopic world. However, the traditional electron microscopy technique made it difficult to achieve specific identification of biomolecules, and this technical dilemma was ushered in by a breakthrough in 1961—Singer’s team successfully coupled ferritin with antibodies (Fig. 1), which enabled the precise localization of tobacco mosaic virus antigens for the first time, marking the birth of the IEM technique [1]. In the process of developing new labeling systems, horseradish peroxidase (HRP) was introduced into IEM studies in 1966, and the high electron density properties of its enzymatic deposition reaction products significantly enhanced detection sensitivity [2]. By 1971, Faulk and Taylor pioneered the colloidal gold immunolabeling technique [3], and the probe rapidly became a mainstream marker for immunoelectron microscopy by virtue of its tunable nanoscale particle size (5–30 nm), high electron-density properties, and good chemical stability. At the same time, sample preparation technology has also been revolutionized: the Epon resin embedding solution developed by Roth’s team [4] achieved excellent ultrastructural preservation by high-temperature polymerization at 60 °C but compromised the antigenic epitopes; this contradiction was mitigated with the development of a low-temperature embedding resin by Carlemalm’s team (Lowicryl K4M/HM20), which has a − 35 °C temperature range. This paradox was alleviated with the development of a low-temperature embedding resin (Lowicryl K4M/HM20) by Carlemalm’s team, whose − 35 °C UV polymerization process balanced ultrastructural integrity with antigenic activity retention [5]. The revolution in advanced techniques began with breakthroughs in physical fixation methods. Moor’s high-pressure freezing technique vitrifies biological samples within milliseconds under 2000 bar pressure [6], preserving native molecular conformations to the greatest extent. Combined with the optimized frozen ultrathin sectioning technique by Japanese researcher Tokuyasu [7], it avoids the dual damage of chemical fixation and resin embedding, significantly improving the preservation of antigenic activity. In the twenty-first century, the maturation of focused ion beam-scanning electron microscopy (FIB-SEM) has successfully expanded the analytical dimension of immunolabeling technology to three-dimensional space [8, 9], which has provided powerful technological support for the revelation of the spatial distribution network of biomolecules. These technological breakthroughs have built up the technological map of modern immunoelectron microscopy to analyze the complex systems of life accurately.

Fig. 1.

Key events in the development of IEM by biologists

Key steps in IEM experimental workflow

A central challenge in immunoelectron microscopy sample preparation lies in maintaining antigenicity while preserving ultrastructural integrity. Due to variability in sample types and research objectives, protocols must be optimized to meet specific experimental needs. Nevertheless, regardless of the sample type, preparation procedures generally consist of several critical steps: fixation, dehydration, embedding, and immunolabeling. The choices and optimization made at each step directly affect image quality and the reliability of the results.

Fixation

IEM sample fixation requires the dual goals of morphology preservation and antigenic activity retention, the core of which lies in the rapid termination of biological activities by physical or chemical means [10] while maximizing the maintenance of the cellular ultrastructure and the natural conformation of biomolecules. Since Palade’s establishment of classical chemical immobilization methods [11], the field has undergone a major revolution from traditional chemical cross-linking to modern physical freezing techniques. In this paper, we will systematically analyze the characteristics of the two technical systems of chemical fixation and cryo-fixation and their recent progress.

Chemical fixation

Chemical fixation stabilizes tissue architecture by forming covalent crosslinks between fixatives and biomolecules such as proteins, lipids, and nucleic acids. An ideal fixative for IEM should meet three essential criteria: rapid tissue penetration, moderate crosslinking strength, and minimal antigen masking [10, 12]. Commonly used fixatives include aldehydes, OsO₄, tannic acid [13, 14], and glycol aldehydes, with a detailed comparison presented in Table 1

Table 1.

Comparison of Common Fixatives

Fixative	Penetration ability	Fixative components	Tissue swelling	Storage time	EM staining	Effect on immunolabeling
Glutaraldehyde	Strong	Enzyme activity, Glycogen, Protein, Nucleic acid	No	≤ 7 days	None	Masks antigen epitopes; high concentration causes tissue shrinkage, affects reagent penetration
Formaldehyde	Stronger	Enzyme activity, Protein, Nucleic acid	No	≤ 7 days	None	Good preservation of antigen activity
Osmium tetroxide	Mild	Best fixative for lipid preservation	Yes, increases sectioning difficulty	≤ 4 h	Yes	Severely destroys antigen activity
tannic acid	Mild	Proteins, sugars (enhances aldehyde fixation effect)	Yes, increases sectioning difficulty	–	Staining agent; enhances heavy metal staining effect	Masks antigen epitopes and increases background signal
Glyoxal	Strong	Membrane proteins, cytoskeletal and matrix proteins	Yes, increases sectioning difficulty	≤ 24 h	None	Low pH may enhance epitope exposure differences

Among these, aldehyde-based fixatives are widely used due to their unique crosslinking mechanisms. Both formaldehyde and glutaraldehyde form three-dimensional crosslinked networks through reactions with amino groups, but they differ significantly in molecular properties and practical outcomes. Formaldehyde, with a smaller molecular weight (30.03 Da), exhibits superior tissue penetration, capable of infiltrating 1 mm³ of biological tissue within minutes [15, 16]. To maximize antigen preservation, paraformaldehyde is commonly used in IEM protocols [17–19]. However, paraformaldehyde alone may result in suboptimal structural preservation. To address this, low concentrations of glutaraldehyde (0.01–0.05%) [20] are often combined with paraformaldehyde [21] in a mixed-fixation strategy during the early stages of immunolabeling [10, 22]. This approach effectively minimizes antigen masking while optimizing the balance between ultrastructural preservation and antigenicity.

OsO₄ is frequently used for post-fixation due to its strong lipid crosslinking ability, which stabilizes membrane structures and enhances contrast in biological specimens [23, 24]. However, OsO₄’s strong oxidative properties can inactivate biomolecules [25, 26], and residual OsO₄ can interfere with immunolabeling techniques such as silver enhancement [27].

To address these limitations, uranyl acetate and tannic acid are often used in combination as alternative fixatives [28, 29], and lead citrate staining is applied to enhance contrast. Glyoxal, a newer fixative, has demonstrated unique advantages in neurobiological applications, offering improved antigen preservation under milder fixation conditions [30]. Despite ongoing efforts to refine chemical fixation protocols, the technique remains subject to notable limitations [16]. Strong crosslinking reduces membrane permeability, thereby impeding the diffusion of immunoreagents and increasing the likelihood of antigen epitope masking [31]. Aldehyde-based fixatives in particular [32, 33] can significantly alter organelle morphology and volume [34].

As a result, cryofixation methods have been explored to preserve ultrastructural integrity [6, 35, 36], especially in studies focused on membrane architecture [34, 37–39], where they provide clear advantages over conventional chemical fixation.

Cryofixation

Cryo-electron microscopy (cryo-EM) entered a phase of rapid development in the 1970s. In 1968, Moor and Riehle first proposed the technique of pressure freezing which achieved biological sample fixation based on physical principles rather than chemical crosslinking, establishing a new paradigm in structural biology research. The first commercially available high-pressure freezer, the BALZERS HPM 010, was released in 1987 [6], marking the transition of this technology into practical application.

The core principle of high-pressure freezing (HPF) is to reduce the freezing point of water to − 90 °C by applying approximately 2000 bar of pressure, allowing intracellular water to undergo rapid cooling from room temperature to − 196 °C within milliseconds. This process vitrifies the water into an amorphous glass-like state, preventing ice crystal formation and thereby avoiding structural damage to the sample [40, 41]. As a physical fixation method, HPF preserves the native architecture of biological tissue, avoiding morphological artifacts associated with chemical fixation [42], and enables the capture of dynamic cellular events such as membrane fusion [35, 43].

To prevent structural damage from mechanical compression during HPF, cryoprotectants are commonly used. Typical agents include sucrose, fish gelatin, dextran, and DMSO [44, 45]. Modern HPF systems such as the Leica EM ICE series feature fully automated temperature control and optimized cryoprotectant formulations (e.g., 20% sucrose, 15% fish gelatin, 5% dextran), which can improve the effective fixation depth to approximately 400 μm [46].

In the context of immunoelectron microscopy, HPF is frequently combined with freeze substitution (FS) (Fig. 2) [37, 38, 47, 48], cryo-ultramicrotomy, and modified Tokuyasu protocols to form a complete sample preparation workflow [49–51]. These procedures require dedicated instrumentation: FS requires a freeze-substitution device, and cryo-ultrathin sectioning (including Tokuyasu cryo-sectioning) requires an ultramicrotome equipped for cryogenic operation. For laboratories with limited access to HPF equipment, atmospheric-pressure freezing using liquid ethane (− 189 °C) can achieve vitrification to a depth of approximately 20 μm. However, to effectively inhibit ice crystal formation, the sample must be pre-cooled to below − 150 °C within 30 s [52].

Fig. 2.

Structural preservation of the wool follicle (∼200 μm in diameter) is markedly improved by HPF-FS compared to conventional chemical fixation (e.g., glutaraldehyde and OsO₄ at room temperature). HPF-FS enables clear visualization of mitochondrial cristae, ribosomes, and interactions between keratin intermediate filaments (KIFs) and trichohyalin granules (TGs). During follicle development, HPF-FS also reveals KIF-mediated cell–cell contacts through desmosomes. Partial figure adapted from Harland D. et al., Journal of Microscopy, 2020, 278(1), by permission of John Wiley & Sons

Dehydration

Dehydration is a critical step in electron microscopy sample preparation. Its primary goal is to replace water in biological specimens with organic solvents, creating a suitable physical environment for resin infiltration and ultrathin sectioning. Common room-temperature dehydrating agents include ethanol, acetone, and propylene oxide. Ethanol is widely used for graded dehydration (30%–50%–70%–90%–100%, v/v) due to its high penetration rate, but its limited miscibility with most epoxy resins necessitates the use of acetone or propylene oxide as transitional solvents for final dehydration [53–55]. Acetone possesses strong lipid solubility [56] and can be used independently for graded dehydration [57, 58] or as a substitution solvent in FS protocols [59, 60]. Propylene oxide reduces surface tension during resin infiltration and is commonly used as a bridging solvent to facilitate mixing between the specimen and resin [53, 54]. In addition to conventional solvents, Yang et al. replaced ethanol with tetrahydrofuran (THF) during the development of a neutral embedding resin [61], thereby slowing the dehydration rate and improving tissue preservation [62].

With advances in cryo-preparation technologies, FS has become a focus of interest due to its superior preservation of ultrastructure. First introduced by Fernández-Morán in 1960 [63], FS employs a programmable temperature control system to achieve stepwise dehydration by warming the sample gradually from − 90 °C to 0 °C. Traditional FS protocols often take more than 24 h to complete. McDonald and colleagues shortened the dehydration duration to under 3 h through equipment and workflow optimization [64], although this approach remains less effective and reliable for large samples [65].

To improve protocol performance, Sobol et al. developed a segmented gradient substitution method. The specimen is first dehydrated in acetone at − 90 °C for 44 h, followed by substitution with an acetone solution containing 0.5% glutaraldehyde and 1.5% water during warming to − 40 °C over 24 h. Finally, low temperature embedding in LR White resin is completed at 0 °C. This stepwise control of chemical crosslinking and dehydration rates significantly improves the preservation of delicate structures such as cell membranes [66].

To preserve cryo-fixed structures during FS, fixatives are commonly added to the dehydrating agents. In conventional EM preparation, low concentrations of OsO₄ or uranyl acetate are typically used [67]. These agents not only maintain the fine structure produced by cryo-fixation but also enhance image contrast through heavy metal staining [38]. However, in immunoelectron microscopy, OsO₄ is generally avoided due to its oxidative damage to antigenic epitopes. Alternatives include paraformaldehyde and glutaraldehyde [68], or uranyl acetate combined with Lowicryl resin embedding [69]. Potassium permanganate has shown excellent preservation of membrane morphology [70, 71], and its combination with uranyl acetate [72] has been proposed as a partial substitute for OsO₄. This approach helps maintain structural integrity while supporting antigen–antibody interactions, though concentration must be carefully controlled to avoid artifact formation [67, 72].

Although HPF combined with FS yields excellent results, it has not fully replaced conventional chemical fixation and dehydration methods [58, 73]. The high pressure involved in HPF can mechanically damage specimens, leading to sample rupture, and ice crystal damage remains a common challenge [74, 75].

Embedding

In electron microscopy sample preparation, embedding media function by filling the voids within biological specimens with a three-dimensional polymer network, thereby providing both ultrastructural stabilization and mechanical support. The physicochemical properties of the embedding resin directly influence the integrity of ultrathin sections and the resolution of the final imaging. Currently, the most used embedding media fall into two major categories: epoxy resins and acrylic resins, as summarized in Fig. 3.

Fig. 3.

Classification of resins used for immunoelectron microscopy embedding

Epon resins, as classic thermosetting epoxy polymers (typically polymerized at 60–70 °C), are known to present certain limitations in immunoelectron microscopy, including antigen epitope masking [76, 77] and autofluorescence in the 500–550 nm range [78]. However, their excellent mechanical strength significantly reduces the occurrence of surface cracking in frozen sections compared to other resins [79], and their resistance to electron beam damage [80] remains advantageous for IEM applications [81].

Acrylic resins are widely used in IEM [5, 82] due to their low-temperature polymerization (ranging from − 50 °C to room temperature) and hydrophilic properties [83]. Hydrophilic Lowicryl resins such as K4M and K11M promote better antibody diffusion, thereby improving immunolabeling efficiency [84]. However, their high moisture absorption can result in surface wetting of resin blocks, making sectioning more difficult [85]. The UV polymerization at low temperatures also helps preserve fine ultrastructure [86].

LR White and LR Gold are synthetic resins composed of polymerized acrylic acid or its ester monomers. These resins offer good transparency, minimal shrinkage [87, 88], and are effective in studies involving chemically sensitive antigens [89, 90] and cell cycle–related processes [68]. Because LR White can restrict RNA probes to the resin surface, it has also been applied in single-molecule fluorescence in situ hybridization (smFISH) studies [91]. A new acetone-soluble acrylic resin developed by Perez-Garza et al. has shown improved performance in immunofluorescence (IF) and fluorescence in situ hybridization (FISH) [92].

LR Gold, polymerized under UV light at low temperature, has lower hardness and viscosity, offering better infiltration into freeze-substituted samples [93]. Its hydrophilic nature provides a higher signal-to-noise ratio in synaptic studies compared to HM20 resin [94].

Over time, embedded samples may undergo changes in hardness, and reagent proportions may need seasonal adjustment. In multiple immunolabeling protocols, extending resin infiltration time can increase sample hardness, facilitating sectioning [95, 96].

Immunolabeling

Immunoelectron microscopy antibody incubation is one of the key steps in the immunoelectron microscopy technique, and its optimization is crucial for signal intensity and quality.

Primary antibody selection

The selection and optimization of primary antibodies in the immunoelectron microscopy experiment is the core link that determines the success or failure of the experiment. Ideal antibodies should possess dual characteristics: high specificity to accurately recognize target antigens and avoid false-positive interference, and high affinity to maintain stable binding throughout the complex sample processing. Since EM samples undergo rigorous fixation, organic solvent dehydration, and embedding steps, antigen epitopes may become masked or structurally altered, significantly increasing the risk of false negatives or false positives [97]. Although monoclonal antibodies exhibit high specificity, they can lead to false negatives if the target epitopes are destroyed by aldehyde fixation or resin embedding. Polyclonal antibodies, recognizing multiple epitopes, effectively mitigate epitope masking but require careful attention to potential nonspecific cross-reactivity [98–100]. Therefore, it is recommended to validate antibody specificity by immunoblotting and to perform pre-assessment of antibody binding efficiency on fixed samples using light microscopy, providing crucial verification before subsequent EM experiments [95, 101, 102].

In sample processing, low concentrations of glutaraldehyde (0.1%–0.5%) combined with paraformaldehyde (2–4%) are commonly used for fixation. Hussain et al. demonstrated that glutaraldehyde pre-fixation of antibodies allows improved labeling efficiency even within a single glutaraldehyde fixation system [103]. Post-fixation quenching with glycine neutralizes free aldehyde groups to reduce nonspecific binding [104]. Regarding antibody penetration limitations caused by resin embedding, conventional IgG antibodies (~ 150 kDa) vary slightly in size depending on species, subtype, and glycosylation. In contrast, nanobodies (~ 15 kDa) exhibit superior penetration and localization capabilities in epoxy resin-embedded samples [105, 106]. Hydrophilic resins such as LR White can also enhance antibody permeability [107, 108]. Notably, nanobodies not only excel in tissue penetration, but their small molecular size improves labeling precision and spatial resolution [109, 110]. Permeabilization treatments using detergents such as Triton X-100, saponin, or Tween-20 may further enhance antibody accessibility, though their impact on ultrastructure integrity requires careful evaluation [111, 112].

To block nonspecific binding, 5% fetal bovine serum, serum from the secondary antibody host species, or fish gelatin is commonly used. Practical protocols often combine carboxymethylated BSA with fish gelatin for enhanced blocking efficacy [106] or employ glycine, Triton X-100, Tween-20, and hydrogen peroxide as antibody signal enhancers to further reduce nonspecific background [96].

Secondary antibody selection

Among the commonly used secondary antibodies, protein A-colloidal gold is the preferred choice because of its unique physicochemical properties, small molecular weight (∼ 42 KDa), good penetration in ultrathin sections, and efficient labeling ability [4]. It can significantly improve the signal-to-noise ratio in complex tissues [107, 113], is chemically stable [114], and can widely bind the Fc segments of IgG from many species [115], combining universality and reliability. For multiple immunolabeling, targets can be distinguished by different particle sizes, and IgG binds to a wide range of markers, and its molecular structure (Fab and Fc segments) is not prone to degeneration [116], so it is also commonly used as a secondary antibody [117, 118]. However, the high molecular weight and the 8–10 nm size of the antibody itself can cause a significant decrease in labeling efficiency due to spatial site resistance. Antibody compatibility needs to be systematically verified during experiments, including false negatives due to epitope masking or cross-reactivity ruled out by sequential reversal incubation and negative controls [119]. On this basis, the spatial distribution of the secondary antibody to the antigen needs to be matched so that resolution and signal intensity can be balanced [107]. To balance resolution and sensitivity, small molecule probes [120] such as nanoantibodies or Fab fragments [121] can be introduced, which have a molecular weight of only 1/10th that of conventional IgG, have a good stability penetration ability, and can significantly reduce spatial site resistance to enhance the precision of multiple labeling [122, 123].

Marker selection

In immunoelectron microscopy, conjugating markers to antibodies is a critical step to enhance signal and enable visual detection. The selection of markers requires careful consideration of multiple factors, including electron density, size, specificity, stability, multiplexing capability, biocompatibility, and signal intensity. Like immunolabeling in light microscopy, dual or multiplex labeling strategies can also be applied in immunoelectron microscopy—either by using the same type of marker in different particle sizes or by combining different types of markers. However, due to steric hindrance and nonspecific interference, the efficiency of multiplex labeling is generally lower than that of single labeling and requires careful optimization through rigorous control experiments. Currently, a wide variety of markers are available for immunoelectron microscopy, including enzyme markers, ferritin, colloidal gold, nanogold, and quantum dots.

Enzyme markers

Horseradish peroxidase (HRP), a classical enzyme marker, has been important in immunoelectron microscopy since the 1960s. Its small molecular weight (~ 40 kDa) gives it excellent tissue penetration ability, which makes it particularly suitable for the precise localization of intracellular antigens. In 1966, Graham and Karnovsky first used HRP to study the permeability of glomerular basement membranes, revealing its potential in biomarkers [124]; in the same year, Nakane’s team catalyzed a diaminopentane-containing enzyme by HRP, which was used as an immunomarker. Nakane’s team established a fundamental model for enzyme-labeled electron microscopy imaging by HRP-catalyzed oxidation of diaminobenzidine (DAB) to generate electron-dense precipitates [125]. However, conventional HRP labeling suffers from signal diffusion, i.e., the precipitate generated by the enzymatic reaction tends to diffuse from the active site, resulting in limited localization accuracy [126–128].

This shortcoming has led to technological innovation: in 2012, Martell et al. achieved a revolutionary breakthrough with the development of genetically encoded ascorbate peroxidase (APEX) [129], which can be directly expressed in fusion with target proteins and catalyze DAB polymerization to form a high-contrast, localized precipitate in situ in living cells, which avoids the disruption of cellular structures caused by chemical fixation and permeabilization, and also avoids the disruption of cellular structures caused by chemical immobilization and permeabilization. This not only avoids chemical fixation and permeabilization but also significantly improves signal resolution [130]. In 2019, Fazal’s team further optimized APEX probes to be compatible with frozen sections and resin-embedded samples and even to label ultrastructures, such as mitochondrial cristae, at the nanoscale, advancing enzyme labeling into the realm of single-cell organelle research [131].

Ferritin

Ferritin played a key role as the first electron-dense marker in the early development of immunoelectron microscopy. In 1959, Singer S.J. successfully labeled bacterial flagellar antigens by covalently coupling ferritin to antibodies for the first time through a toluene diisocyanate cross-linking strategy [132]. This groundbreaking study was made possible by the unique structure of ferritin, which consists of hydroxyl iron oxide crystals of about 4,500 iron atoms (~ 7 nm in diameter), which produces significant electron contrast under electron microscopy, making it an ideal signal source for early antigen localization. However, the large spatial site resistance of ferritin, with a molecular weight of up to 450 kDa, makes it difficult to penetrate dense biological tissues and is only suitable for cell surface antigen labeling [133, 134], a limitation that has prompted researchers to seek out smaller markers. In 1971, inspired by ferritin, Faulk and Taylor developed a colloidal gold technique using sodium citrate reduction [3].

Gold particles

Since the first introduction of colloidal gold to immunoelectron microscopy by Faulk and Taylor in 1971 [3], this electron-dense marker has become a central tool in immunolocalization studies. Classical colloidal gold is formed by sodium citrate reduction into spherical particles with a particle size of 3–50 nm, and its binding to antibodies relies on the electrostatic attraction of the negatively charged surface of the gold particles to the positively charged regions of the antibody. This facile coupling method drove early widespread use, and high-resolution localization of antigens was achieved by optimizing the colloidal gold preparation process for different particle sizes [115, 119]. Commonly used colloidal gold particle sizes range from 3 to 50 nm, with smaller particles providing higher resolution but weaker signals, even while larger particles have stronger signals but poorer permeability, resulting in some loss of resolution [135, 136]. To allow colloidal gold couplings to penetrate cells more easily, even smaller sizes of colloidal gold, e.g., around 1 nm, have been successfully used to immunolabel intracellular structures [137, 138]. Colloidal gold can be directly conjugated to primary antibodies for immunolabeling [139, 140], which improves the spatial resolution of immunogold labeling, but antibody availability is limited, and labeling efficiency is low. Indirect labeling can also be used, with high signal amplification, versatility, and friendliness to low-expression antigens [141–144]. However, either method connects the antibody by physical adsorption, which is prone to dislodgement during sample processing and seriously affects labeling stability [145].

This paradox has given rise to the breakthrough development of nanogold. Nanogold is ultra-small-sized (1–3 nm) gold particles functionalized by chemical modification [146], and its features, such as structural stability and chemically coupled antibodies, have enabled it to show unique advantages in pinpointing and studying specific proteins [145, 147, 148]. However, the new challenges posed by the ultra-small size are also significant. Whether it is colloidal gold or gold nanoparticles, the microscopic resolution decreases when the size is too small, such as around 1 nm, and silver enhancement techniques need to be used for signal amplification. For example, Hacker et al. (1996) used silver ions deposited on the surface of gold nanoparticles to form composite structures, which can enhance the signal intensity by tens of times [149, 150], but this method is susceptible to nonspecific deposition due to interference from endogenous metals in tissues, and the OsO₄ post-fixation step can dissolve the silver shells [137]. For this reason, the gold toning technique developed by Leitinger et al. (2000) successfully blocked the oxidative attack of OsO₄ acid by gold plating the surface of silver shells with gold chloride [151, 152], and overcome to a certain extent the problems of the silver enhancement method in terms of poor specificity, pH and electron beam sensitivity, etc. Wakana (2008) and Yamamoto’s team (2016) further improved the stability of silver-enhanced labeling by low-temperature resin embedding with a gold shell coating strategy [153, 154]. Furthermore, Metal-tagging TEM (METTEM) is an emerging technique that utilizes heavy metal clusters, such as uranium or platinum-based probes, to replace traditional gold nanoparticles. These metal tags provide strong electron contrast while being compatible with resin embedding and post-staining protocols [19, 155].

Meanwhile, silver nanoparticles have attracted attention due to their unique optical properties. La Spina et al. (2020) found that the surface plasmon resonance effect of 20 nm silver particles could generate strong scattering signals under light microscopy, realizing light microscopy-electron microscopy dual-mode correlation imaging [156] while Gangwar et al. (2021) developed a novel multiple localization scheme (La Spina et al. 2020) by modulating the hybrid silver/gold labeling system developed a novel multiple localization scheme [157]. However, the application of silver particles is limited by their chemical sensitivity—OsO₄ treatment triggers silver dissolution and releases interfering ions—which has prompted researchers to turn to gold enhancement techniques. The silver-free enhancement method proposed by Weipoltshammer et al. (2000) for the direct deposition of Au–Au particles on Au nanoparticles avoids silver-related defects and achieves homogeneous particle size amplification through atomic-level precision deposition control, opening a new path for ultra-high resolution immunoelectron microscopy [158].

Quantum dots

Quantum Dots (QDs), as new semiconductor nanomarkers, are revolutionizing immunoelectron microscopy with their unique physicochemical properties. The core advantages of quantum dots over conventional markers are their ultra-high electron density and chemical stability [159–163]. Quantum dots with diameters between 2 and 10 nm can withstand sampling processes such as aldehyde fixation, organic solvent dehydration, and high temperature embedding in epoxy resins and show clear and recognizable signals under transmission electron microscopy [164–166]. However, the biocompatibility of quantum dots remains a key challenge, and toxicity can be reduced by surface polyethylene glycol (PEG) modification or biodegradable shell layer design [167, 168]. In addition, the size-dependent fluorescence properties of quantum dots (emission wavelength can be precisely tuned by particle size) and resistance to photobleaching make them ideal probes for CLEM: the same quantum dots are capable of both rapid localization of the target region by fluorescence imaging and providing nanoscale structural information under electron microscopy [162, 169, 170]. This bimodal property shows great potential in complex biological systems, such as the labeling of the cardiomyocyte Sigmar1 protein by streptavidin-coupled quantum dots, which for the first time revealed the dynamic distribution of the protein at the mitochondria-endoplasmic reticulum membrane contact site at the subcellular level [171, 172].

Grid selection and processing

In IEM, while the selection of antibodies and labeling agents is undoubtedly critical to experimental success, the support grid—serving as both a physical scaffold and a chemical modification platform—also plays a vital role in labeling efficiency, epitope preservation, and imaging resolution.

First, given that IEM experiments involve multiple rounds of chemical treatments and extreme conditions such as cryosectioning and resin polymerization, the grids must exhibit high mechanical strength and chemical resistance. Compared with copper grids, nickel and gold grids offer greater resistance to bending and can better withstand the mechanical stress associated with ultrathin sectioning. In addition, they help prevent corrosion induced by sodium azide (NaN₃), a commonly used preservative in antibody solutions, thereby reducing the risk of grid damage and background deposition [173, 174].

Second, to prevent the problem of delamination in subsequent tests, optimization means such as chemical modification and glow discharge [96, 175–177] can be adopted. The chemical modification of the grid surface coating and pre-coating with polylysine or gelatin can enhance the biomolecular binding strength and reduce the risk of delamination. Functionalized silane treatment with the introduction of amino or epoxy groups improves antibody immobilization efficiency through covalent cross-linking [178]. Such methods are particularly critical in low-temperature resin embedding (e.g., Lowicryl) to reduce antigenic epitope masking. Then again, the grids were treated for 10 min in 1% periodic acid and 15 min in 9% sodium periodate [179].

IEM technology systems

In IEM, pre-embedding and post-embedding labeling represent two core strategies that differ mainly in the sequence of antibody labeling and resin embedding. Pre-embedding labeling refers to performing antibody incubation before resin embedding, which minimizes the risk of antigen inactivation or epitope masking during embedding. This approach is particularly suitable for cell surface antigens or low-abundance targets, though its limitations include restricted antibody penetration and potential ultrastructural damage. In contrast, post-embedding labeling is carried out after resin embedding and ultrathin sectioning, with antibody labeling applied directly to the section surface. This method preserves fine cellular architecture and allows precise localization of intracellular antigens, but some epitopes may be lost due to chemical treatments during embedding. The choice between these strategies should be guided by a balance among antigen properties, localization accuracy, and preservation of ultrastructure.

Pre-embedding labeling

In pre-embedding labeling, antibodies are applied prior to fixation, dehydration, and resin embedding [180, 181], effectively minimizing the risk of epitope masking or denaturation during the embedding process [140]. This approach offers the advantages of shorter diffusion distances for labeling reagents and reduced steric hindrance, which significantly enhances the detection sensitivity—particularly for cell surface antigens and low-abundance targets [182]. To accommodate differences in antigen subcellular localization, experimental strategies must be tailored accordingly. Labeling of membrane surface antigens generally does not require membrane permeabilization. Alternatively, protocols may optimize antibody access while maintaining membrane integrity. For example, fixation in hypotonic buffer combined with saponin treatment has been shown to increase the labeling efficiency in pre-embedding protocols. In contrast, labeling of intracellular or membrane-enclosed antigens relies heavily on the choice of permeabilizing agents and their compatibility with specific sample types.

Permeabilization agents

In pre-embedding IEM, permeabilization agents such as Triton X-100, saponin, or digitonin have historically been explored to facilitate intracellular antigen labeling. Digitonin selectively targets cholesterol-rich membranes and can preserve other organelles to some extent. However, as shown in Fig. 4, detergents inevitably disrupt membrane continuity and compromise ultrastructural preservation [183]. Therefore, while permeabilization is useful in light microscopy, it is generally avoided in modern IEM protocols, where maintaining fine ultrastructure is critical.

Fig. 4.

Triton X-100 facilitates antibody penetration in neural tissue during rehydration of freeze-substituted brain samples. Compared to detergent-free conditions, treatment with Triton X-100 (especially at 0 °C) preserves ultrastructure and improves immunolabeling quality. Scale bar = 500 nm. Figure reproduced from Pérez-Garza J., Parrish-Mulliken E. Microscopy and Microanalysis, 2023, 29(5), by permission of Oxford University Press

Applications

Pre-embedding immunolabeling has demonstrated unique advantages in studies involving the spatial reconstruction of membrane protein complexes, owing to its superior preservation of antigenic epitopes [184, 185]. For transmembrane proteins such as the lysosome-associated membrane protein LAMP1, pre-embedding approaches can effectively circumvent epitope masking caused by resin infiltration. When combined with 0.05% saponin for gentle permeabilization, this method allows for precise localization of transmembrane domains, and has been shown to outperform post-embedding labeling techniques in such contexts [186–188]. In studies of membrane surface antigens, such as G protein-coupled receptors (GPCRs) [17], pre-embedding labeling can increase labeling density by 2–threefold [121].

In the field of neuroscience, pre-embedding immunolabeling has shown value. Post-embedding labeling is often ineffective for monolayer cultures of neurons, and therefore vibratome sections are commonly used for pre-embedding detection of antigens in brain tissue [18, 39, 189, 190]. Compared with conventional chemical fixation, which may cause ultrastructural artifacts such as widening of dendritic spine necks, high-pressure freezing combined with freeze substitution has been shown to significantly reduce such deformation [18, 191, 192]. Furthermore, rehydration of freeze-substituted samples—via gradual replacement of acetone with aqueous solutions—can improve detection sensitivity for labile endogenous molecules [192].

To reconcile the trade-off between antibody penetration and ultrastructural preservation in pre-embedding protocols, several optimization strategies have been proposed. For example, reducing glutaraldehyde concentration to 0.05% can lower protein crosslinking by approximately 30% [193, 194]. Additional improvements include the use of HPF [195–197], refined permeabilization protocols [112, 198], and antibody engineering [194, 199]. However, the overall efficacy of these strategies remains limited. Therefore, expanding the IEM toolkit to include post-embedding immunolabeling remains essential for specific experimental needs.

Post-embedding labeling

Post-embedding labeling is the method of immunolabeling after sample embedding and preparation of ultrathin sections [85, 200–202]. It can be divided into two methods: low-temperature aqueous resin embedding and Tokuyasu frozen ultrathin section immunolabeling technique.

Low-temperature aqueous resin embedding

In immunoelectron microscopy, resin selection is crucial for preserving antigenicity. Low-temperature, hydrophilic resins allow substantial retention of antigen activity when substituted and polymerized under cold conditions. Polymerization can be carried out using a low-temperature UV polymerization chamber or a freeze-substitution device.

In post-embedding immunolabeling, resin selection should be systematically optimized according to sample characteristics and antigen sensitivity. For studies of neuronal synapses, in the absence of high-pressure freezing equipment, dual fixation with tannic acid and uranyl acetate has been used as an alternative to osmium tetroxide (OsO₄). Combined with Epon 812 epoxy resin polymerized at 60 °C for 48 h, this approach enabled adequate labeling density of synaptic proteins in the mouse hippocampus, while maintaining the synaptic cleft width within physiological range [27, 29, 203]. In studies on the development of spinal cord motoneurons in rats, Durcupan ACM resin polymerized at 60 °C, along with post-fixation by OsO₄, was successfully applied to visualize the spatial distribution of GABA and glycine receptors at synaptic terminals [179, 204–206].

For aldehyde-sensitive antigens, Lowicryl HM20 resin, which undergoes dehydration by freeze substitution and UV polymerization at − 50 °C, offers greater practicality [5, 207]. In contrast, Lowicryl K4M, despite its higher hydrophilicity and better infiltration properties, tends to cause distortions in the spatial arrangement of synaptic vesicles due to a higher degree of shrinkage during polymerization [207].

The LR White resin system has shown advantages in double-labeling studies. Its rapid polymerization at room temperature allows infiltration and hardening to be completed within 2 h. Using a combination of 5 nm and 10 nm gold-conjugated antibodies, researchers demonstrated the localization of two neurotransmitters within the same axonal microdomain [208, 209]. Compared with Lowicryl, LR White provides a safer, faster, and more economical alternative for immunolabeling of sectioned samples [210], and it does not require the use of traditional aldehyde-based fixatives [77]. It is also noteworthy that LR Gold resin, when UV-polymerized at − 15 °C, can preserve the integrity of microtubule structures in mouse cortical neurons even without OsO₄ post-fixation [27, 211].

Tokuyasu frozen ultrathin section immunolabeling technique

Conventional chemical fixation followed by resin embedding often causes alterations in protein higher-order conformation, such as mitochondrial cristae expansion, and deformation of organelles, including increased vesicle diameter in the endoplasmic reticulum. These artifacts severely limit the in-situ analysis of ultrastructural organization [34, 212]. The Tokuyasu cryo-sectioning technique builds upon the antigen-preserving capabilities of low-temperature resin embedding and further overcomes the limitations associated with chemical fixation and resin infiltration [7, 213]. This method preserves over 90% of native membrane protein conformations and enhances labeling density, making it particularly well-suited for in situ detection in multilayer membrane systems such as nuclear pore complexes and endocytic vesicles.

The breakthrough of the Tokuyasu cryo-ultrathin sectioning technique lies in its ability to balance ice crystal suppression with antigen preservation through sucrose gradient infiltration. After mild fixation with 1–4% formaldehyde, samples are infiltrated with 2.3 M sucrose for gradient cryoprotection and then sectioned at 70–150 nm thickness in methylcellulose embedding medium cooled by liquid nitrogen [7, 213]. The modified protocol by Griffiths et al. (1984) introduced a 0.1–0.5% methylcellulose coating, which maintained water content above 80% during grid collection. This modification reduced cytoplasmic membrane collapse and significantly improved imaging quality for delicate structures such as nuclear pore complexes [214].

In studies on apoptotic mechanisms in prostate cancer cells, this technique successfully localized the PAK6–SIRT4–ANT2 complex on the mitochondrial outer membrane, achieving higher spatial resolution than conventional resin embedding methods [215]. In membrane structure analysis, where chemical fixation often induces distortion and artifacts, Van Donselaar’s team (2007) combined high-pressure freezing with the Tokuyasu method to develop a dual-modality probe, fBSA-Au—consisting of fluorescently labeled bovine serum albumin conjugated with gold particles. This probe enabled spatiotemporal tracking of vesicular transport in HeLa cells and facilitated precise correlation between fluorescence and electron microscopy images [50, 216].

Classical workflows and methodological variants in IEM

Although IEM workflows are conventionally divided into two principal strategies, a variety of methodological variants have been developed to address challenges such as antigen preservation, antibody penetration, and ultrastructural integrity. Figure 5 summarizes the classical workflows together with representative modifications, highlighting the critical steps from fixation through labeling.

Fig. 5.

Classical workflows and methodological variants in IEM. IEM approaches are broadly categorized into pre-embedding and post-embedding strategies, defined by the sequence of embedding and immunolabeling. Variants have been developed to address challenges of antigen preservation, antibody penetration, and ultrastructural integrity. While chemical fixation provides the basis for conventional pre-embedding and Tokuyasu cryo-sectioning, the use of high-pressure freezing and freeze substitution enables both pre-embedding and post-embedding labeling routes. Specific equipment requirements include high-pressure freezing systems for HPF–FS and a cryo-ultramicrotome for the Tokuyasu method

In pre-embedding labeling, the fixation strategy dictates the subsequent workflow. With an HPF–FS system, samples are loaded into freezing carriers, vitrified under high pressure in liquid nitrogen, and dehydrated at − 80 to − 90 °C by freeze substitution with organic solvents such as acetone. After FS, samples are rehydrated stepwise to permit immunolabeling, followed by a second dehydration and resin embedding, essentially paralleling procedures used for chemically fixed material [18, 189, 190]. Because prolonged exposure to acetone at low temperature can disrupt protein folding and antigenicity, gradient rehydration with aqueous solutions has been introduced to restore protein conformation, recover tissue porosity, and enhance antibody penetration, thereby improving labeling efficiency [51, 183, 217].

Post-embedding immunolabeling is often performed with low-temperature aqueous resins, in which ultrathin sections are prepared prior to immunolabeling [209, 218]. However, conventional resin polymerization at high temperature can damage antigenic epitopes. The Tokuyasu method [173] circumvents this issue by aldehyde fixation, sucrose cryoprotection, rapid freezing, and cryo-ultrathin sectioning, followed by buffer rinsing and immunolabeling [49, 219]. HPF–FS has further streamlined post-embedding workflows[77, 220]: in the low-temperature aqueous resin approach, samples can be embedded and sectioned directly after FS, avoiding additional rehydration steps while preserving antigenicity [50, 51]. Alternatively, cryo-sections from HPF samples must remain under liquid nitrogen throughout handling; specimens are mounted in a cryo-sample holder, sectioned at –120 °C, and stabilized by chemical fixation during thawing before immunolabeling [50].

Statistical analysis

As the gold standard for molecular localization studies at the subcellular level, the standardization of quantitative analysis of immunoelectron microscopy colloidal gold labeling technique is of great significance in improving the reliability of the results. Early studies have confirmed that quantitative statistics can effectively eliminate the subjective bias of observers and provide objective evidence for biological processes [221–223]. However, the distribution characteristics of colloidal gold particles are modulated by multiple factors: endogenous factors reflect the true biological localization, while exogenous factors involve technical variables such as antigenic epitope accessibility, antibody affinity, and fixation methods [224, 225]. Understanding these influences is a prerequisite for establishing a reliable quantitation strategy. The standardized procedure consists of three key components.

The first is sample preparation quality control, which requires a positive control group (known antigen expression samples) versus a negative control group (knockout model or isotype control antibody-treated samples) to assess labeling specificity [154]. Fixative selection needs to balance antigenic epitope preservation with ultrastructural integrity, and a combination of low-concentration paraformaldehyde (0.5–2%) and glutaraldehyde (0.1–0.25%) is recommended [225]. Next is the signal statistics scheme, where the use of systematic uniform random (SUR) is effective in avoiding region selection bias [226, 227]. For specific implementation, it is recommended to establish a systematic sampling grid at low magnification (2000–5000 ×), with interval distances adjusted according to tissue heterogeneity. For organelle localization studies, a hierarchical sampling method is recommended: first, randomly selected cells, then selected intracellular systems are sampled [228, 229]. Finally, the quantitative analysis system is established, and for two-dimensional marker density calculations, the number of gold particles/area of the analyzed area (particles/µm²), attention needs to be paid to measurement bias due to projection effects [230]. Three-dimensional reconstruction techniques, in combination with the continuous slice dissector method, are required to calculate bulk density accurately, and the Rotator method is suitable for volume estimation of irregular structures [230, 231]. For spatial distribution analysis, marker aggregation can be detected based on the nearest neighbor index (NNI) [225].

For standardized calibration, the standard curve method for antigen concentration established by Lucocq (1994) is still widely used, with the caveat that nonlinear relationships may exist in regions of high antigen density [107]. Automated analysis systems such as Gold Digger developed in recent years are based on convolutional neural networks [206], which significantly outperform traditional thresholding algorithms (e.g., the TAC algorithm) in terms of marker identification accuracy and processing speed [232] and are particularly advantageous in the identification of specificity in complex backgrounds [233, 234].

Technology integration

The masking effect of OsO₄ fixation on epitopes and heavy metal staining interference limits the immunolabeling resolution of conventional transmission electron microscopy (TEM). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) enables high-contrast imaging of subcellular structures without OsO₄ fixation by detecting inelastic scattered electron signals. This technique provides a 5–eightfold increase in sensitivity to heavy metal labeling compared to conventional TEM, detects ultra-small colloidal gold particles (< 3 nm), and significantly improves the precision of labeling localization of fine structures such as photoreceptor disk membranes [235]. Notably, the signal intensity of HAADF-STEM is proportional to the square of the atomic number of the sample requiring methods such as platinum labeling to enhance the signal-to-noise ratio [236].

The development of bimodal probes has driven the development of CLEM. Synchronized detection of photoelectric signals at the single-particle level is achieved by covalent coupling of thiolated antibodies to gold nanoparticles [220, 237]. However, the photobleaching effect leads to a fluorescence signal decay rate 3–5 times faster than that of quantum dots [238]. Fluorescent gold nanoparticles have been shown to dissociate under certain conditions in CLEM [238–240], leading to poor co-localization between the probe and the target [241]. In addition to gold particles, quantum dots [162, 165] are also commonly used to prepare bimodal probes.

Conventional FIB-SEM lacks biomolecular information, although it can provide ultra-high-resolution information on cellular ultrastructure. Optical microscopy techniques (e.g., SIM, PALM, STED, and STORM), while capable of super-resolution imaging, have limitations in in-depth imaging and multiple labeling. Gopal et al. (2019) was able to accurately map structural and functional information about cell-material interactions in three dimensions by combining antigen labeling with FIB-SEM imaging [8]. FIB-SEM combined with immunoelectron microscopy preserves the immunogold labeling signal while homogenizing the XYZ-axis resolution [9].

Application case of IEM

IEM, with its nanometer-scale spatial resolution and high antigen specificity, has become a powerful tool for elucidating fine cellular structures and mechanisms of disease. Its applications in biomedical research have grown increasingly extensive and in-depth, particularly in fields such as tumor biology, virology, and neuroscience, continuously pushing the boundaries of basic research and clinical translation.

Tumor biology and targeted therapy

In tumor microenvironment research, IEM serves as the “gold standard” for elucidating the spatial distribution and interactions of key immune checkpoints and oncogenes. For example, IEM using anti-lipoteichoic acid (LTA) immunogold labeling clearly confirmed the microbial origin of Bifidobacterium-derived extracellular vesicles (Bif.BEVs). Furthermore, in lung cancer cells (LL/2), IEM combined with specific inhibitor experiments showed a significant reduction in intracellular LTA-positive signals, directly demonstrating that the endocytosis of Bif.BEVs is dependent on dynamin-mediated pathways. This provides a reliable visual validation standard for studying the functions of BEVs derived from Gram-positive bacteria [242].

In addition, IEM plays a crucial role in the design and evaluation of novel nanomedicine delivery systems. Zhao et al. [243] developed an exosome-based targeted delivery platform for primary central nervous system lymphoma (CNSL) by engineering exosomes derived from human mesenchymal stem cells to display anti-CD19 antibodies on their surface. Colloidal gold IEM was successfully used to confirm the presence and precise localization of the anti-CD19 antibodies on the exosome surface. Both in vitro and in vivo experiments demonstrated that this delivery system significantly enhanced blood–brain barrier (BBB) penetration and achieved precise targeting of tumor tissues, providing robust ultrastructural evidence for exosome-based precision therapies in the central nervous system.

Virology

IEM plays an irreplaceable role in deciphering the ultrastructure of viral replication cycles, particularly in visualizing virus replication complexes (VRCs) or viral factories (VFs). Ricciardi et al. employed CLEM to first locate cells expressing SARS-CoV-2 nonstructural protein NSP6 via fluorescence microscopy, followed by targeted IEM labeling. Their study revealed that NSP6 specifically localizes to “zipper-like” membrane structures derived from the endoplasmic reticulum (ER). With FIB-SEM, they reconstructed the 3D architecture of these zipper formations, showing how they serve as molecular scaffolds mediating the formation and interconnection of double-membrane vesicles (DMVs)—key replication organelles of SARS-CoV-2. This work provided landmark structural insights into coronavirus replication [244].

Neuroscience

The high heterogeneity and complex microenvironment of the nervous system demand ultrastructural studies at subcellular resolution, where immunoelectron microscopy (IEM) demonstrates irreplaceable advantages. On one hand, IEM enables in-depth analysis of rare or hard-to-obtain clinical samples. For instance, Otubo et al. (2021) successfully applied IEM to long-term cryopreserved hypothalamo-pituitary axis samples from non-human primates (marmosets), achieving high-resolution localization of vasopressin within neurosecretory granules and detailed reconstruction of neuronal ultrastructure. This strongly validated that optimized IEM protocols can effectively “reactivate” ultrastructural information from archived precious specimens, opening new avenues for neuroendocrine research [245].

On the other hand, IEM is a key tool for deciphering intercellular communication mechanisms within complex neural circuits. Nakadate and Kawakami [246] used IEM with anti-MEGF8 antibodies conjugated to 1.4 nm gold particles and, for the first time, revealed a unique synapse–mitochondria dual localization pattern of MEGF8 protein in the mammalian central nervous system (CNS). This finding provides essential ultrastructural evidence for MEGF8’s role in regulating neuronal energy homeostasis and synaptic function, thereby establishing a direct link between the learning disability phenotype of Carpenter syndrome and specific subcellular dysfunction.

Summary and perspectives

Immunoelectron microscopy (IEM) techniques have made significant progress in recent years, especially in the shift from 2D localization to 3D reconstruction and from single labeling to multimodal integration. However, despite this, IEM still faces a few technical challenges that need to be addressed through continuous innovation and optimization. First, the issues of probe stability and compatibility of multiple labeling remain one of the bottlenecks in IEM technology. For example, fluorescent gold nanoparticles may dissociate during the labeling process, thus affecting the stability of the labeling effect. The solution to this problem requires the development of new probes, especially those that can be stably immobilized and have multiple labeling functions at the same time, such as probes based on the stabilization of disulfide bonds and degradable quantum dots. Second, how to maintain antigenic activity while ensuring the fidelity of ultrastructures, especially in the study of dynamic processes, remains an important challenge. Although the high-pressure cryo-fixation technique can effectively maintain the structural integrity of the sample, a better balance between antigenic activity and ultrastructural fidelity still needs to be found. This requires in-depth optimization in chemical fixation, cryoprocessing, and antibody selection. In addition, methodological improvements before and after embedding are crucial, such as optimizing labeling by precisely controlling the freezing process and the use of fixatives to ensure that the antigen is not damaged during processing. Then, for the data analysis aspect, with the development of IEM technology, the large number of electron microscopy images generated brings new challenges in analyzing efficiency and accuracy. Traditional manual image analysis methods can no longer meet the demands of high-throughput experiments, so more efficient and automated image analysis tools need to be developed. For example, combining artificial intelligence (AI)-driven structure prediction and data analysis methods can effectively improve the speed and accuracy of image analysis. In addition, cryo-electron tomography (cryo-ET) combined with IEM can provide higher-resolution 3D images, further enhancing the ability to resolve protein conformations in situ. Through interdisciplinary collaboration and continuous technology iteration, IEM technology will play an increasingly important role in the field of single-molecule pathology diagnosis. Especially in the early diagnosis and personalized treatment of tumors and neurodegenerative diseases, IEM technology will provide more accurate molecular-level information for clinical applications, thus accelerating the diagnosis and treatment of diseases.

Author contributions

Jinsai Wu led the writing of the manuscript. Bo Su, Leiyan Gu, and Jie Zhang assisted with literature research, formatting, and figure preparation. Qiuxiao Shi and Danrong Hu contributed to the revision and served as corresponding authors. All authors read and approved the final version of the manuscript.

Funding

This work was financially supported by the Higher Education Teaching Reform Project of Sichuan University (Project No. SCU11260). We also thank the National Natural Science Funds (NSFC 82572961 and 32001003), and the Natural Science Foundation of Sichuan Province (2025ZNSFSC0773 and 2025ZNSFSC0775).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interest

The authors declare no competing interests.