Super-resolution microscopy for analyzing neuromuscular junctions and synapses

Abstract

Super-resolution microscopy techniques offer subdiffraction limited resolution that is two- to ten-fold improved compared to that offered by conventional confocal microscopy. This breakthrough in resolution for light microscopy has contributed to new findings in neuroscience and synapse biology. This review will focus on the Structured Illumination Microscopy (SIM), Stimulated emission depletion (STED) microscopy, and Stochastic optical reconstruction microscopy (STORM) / Single molecule localization microscopy (SMLM) techniques and compare them for the better understanding of their differences and their suitability for the analysis of synapse biology. In addition, we will discuss a few practical aspects of these microscopic techniques, including resolution, image acquisition speed, multicolor capability, and other advantages and disadvantages. Tips for the improvement of microscopy will be introduced; for example, information resources for recommended dyes, the limitations of multicolor analysis, and capabilities for live imaging. In addition, we will summarize how super-resolution microscopy has been used for analyses of neuromuscular junctions and synapses.

Introduction

Super-resolution microscopy techniques overcome the resolution limit of light microscopy that is defined by Abbe’s diffraction limit of approximately 200 nm [2, 94]. For the recognition of this breakthrough, the 2014 Nobel Prize in Chemistry was awarded to Drs. Stefan Hell, Eric Betzig, and William E. Moerner for the development of the super-resolved fluorescence microscopy [36]. Several super-resolution microscopy techniques are commercially available, including Structured Illumination Microscopy (SIM) [45–47, 83], STimulated Emission Depletion (STED) microscopy [48, 62], and STochastic Optical Reconstruction Microscopy (STORM) / Photo Activated Localization Microscopy (PALM) [11, 50, 100, 108]. STORM and PALM are types of Single Molecule Localization Microscopy (SMLM). This review will compare the advantages and weaknesses of these super-resolution microscopy techniques and summarize how they have been applied to analyses of synapses. Neuromuscular junctions (NMJs) are large synapses formed between motor neurons and muscle fibers. Due to their large sizes and accessibility, these synapses have been studied extensively using histological and physiological methods [33, 43, 53, 72, 96, 115]. Even considering the long history of studying these synapses, recent applications of super-resolution microscopy to the examination of these synapses have revealed new findings.

Resolution

Super-resolution microscopes offer subdiffraction limited resolution, which is better than the resolution of confocal microscopes of approximately 200 nm. In neurobiology, confocal microscopes can image axons and nerves, dendritic arbor morphology, mitochondrial morphology, nuclei, synapses and NMJs. However, superresolution microscopes have been useful for imaging the detailed structures of spines, protein distribution patterns in nodes of Ranvier, nuclear pore proteins, the subsynaptic localization of pre- and postsynaptic proteins, and synaptic vesicles. An en face view of NMJs is large, with a size ranging from approximately 10 – 60 μm (human, mouse) [22, 56]; however, some of the key features of this synapse have a size that is below the diffraction limited resolution of 200 nm. For example, synaptic vesicles are 55 nm [64, 82], active zones are 50 – 100 nm [73, 82, 86], and the synaptic cleft is 50 – 100 nm [72]; in addition, the junctional fold crest width is 207 nm and the opening width is 55 nm [139] (Figure 1), the laminin protein length is 77 nm [37], and the Lrp4 protein length is approximately 60 nm [72]. Thus, the improved resolution of super-resolution microscopy is necessary for analyzing these features, and examples of such studies will be described later. In this review, we will focus on the three commercially available super-resolution microscopy techniques: SIM, STED, and STORM. The differences in their resolution will be discussed along with a brief explanation of each imaging mechanism.

Figure 1.

Abbe’s diffraction limit, the sizes of synaptic structures and tools used in super-resolution microscopy. The X-axis indicates the size in micrometers to nanometers. The sizes of the synaptic structures and objects shown in the figure are as follows: NMJs in the en face view are 10 – 60 μm (human, mouse) [22, 56], the central nervous system (CNS) synapse is 1 – 2 μm (mouse) [59], the synaptic cleft of the NMJ is 50 – 100 nm [72], the junctional fold crest width is 207 nm and the opening width is 55 nm [139], the active zone is 50 – 100 nm [73, 82, 86], the synaptic vesicle (SV) is 40 – 50 nm [72, 102], the antibody is 12 × 14 nm [68, 80], the nanobody (variable domain, VD) is 2.5 × 4 nm [60], and green fluorescent protein (GFP) is 2.4 × 4.2 nm [51].

The XY-plane resolution of these techniques is as follows: SIM, 120 nm, STED and STORM, 20 – 30 nm (Table 1). Their resolution is significantly improved compared to that of confocal microscopy, which has a resolution of approximately 200 nm in the XY plane and 500 nm along the Z-axis (for a green fluorescent dye observed with an NA=1.4 lens) [114]. The Z-axis resolution of confocal microscopy is lower than the XY-resolution because of the broad focal region, as determined by Abbe’s diffraction limited resolution theory, which was formulated by Abbe [2] and Rayleigh [94]. The Z-axis resolution of the super-resolution microscopy techniques are as follows: SIM, 300 nm, STED, 75 – 100 nm (Easy 3D, STED 3×), and STORM, 50 nm (STORM 3D). In addition to the differences in resolution, these techniques differ in the mechanisms that improve the resolution. SIM and STORM reconstruct super resolution images based on the postprocessing of the raw data micrographs. In contrast, STED microscopy generates an image with optical data only without postprocessing the raw data. More specifically, SIM generates super-resolution images using a Fourier transform that is applied to the raw data micrographs. STED microscopy generates super-resolution images by scanning the laser beam over the specimen, similar to confocal microscopy. STORM generates superresolution images by collecting thousands of micrographs and reconstructing them using Gaussian distribution approximation and assigning dots, similar to pointillism. The mechanisms of each method will be discussed next.

Table 1:

Comparison of super resolution microscopy specifications

Parameter	SIM	STED	STORM
Resolution XY-plane	115 – 120 nm [85, 111]	20 – 30 nm [3, 44, 71]	20 – 30 nm [85, 111]
Resolution Z-axis	300 nm	75 – 100 nm (Easy 3D, STED 3×)	50 nm (STORM 3D)
Multi-color capability	Two channels can be acquired simultaneously and four to six colors without simultaneous acquisition [91, 140].	Dual-color [12, 26, 86, 125], triple-color [7, 18, 110, 121, 141], and quadruple-color STED [136] have been reported.	Dual-color detection is routinely used [1, 6, 20, 25, 52, 57, 65, 66]. Triple-color STORM has been reported in combination with spectral demixing [69].
Speed	Commercial model from Nikon offers a rate of 15 FPS. Zeiss Elyra 7 system offers a frame rate of 255 FPS [85, 111].	Commercially available resonant scanners from Leica offer a rate of 50 FPS, or up to 428 FPS by restricting the region of interest [15, 16].	Limited temporal resolution with an imaging speed of less than 1 FPS [85, 111].
Live imaging	Suitable	Suitable	Not preferred
Post-processing	Necessary, using Fourier transform of 9 to 15 raw data micrographs [45, 47, 83].	Not necessary, a raw data micrograph is a super resolution image [48].	Necessary, using Gaussian fitting of tens of thousands of raw data micrographs [5, 10, 11, 50, 100, 108].

FPS= frames per second; SIM= structured illumination microscopy; STED= stimulated emission depletion; STORM= stochastic optical reconstruction microscopy

SIM uses a wide-field microscopy setup and rotating grid pattern illumination to generate Moiré fringes and to collect nine to 15 micrographs by changing the mask orientation, and the final super resolution image is reconstructed using a Fourier transform of the acquired micrographs [45, 47, 83]. SIM has been further developed so that this method can be applied in three dimensions to improve the resolution in the lateral and axial directions, which is called 3D-SIM [46]. The newer version of SIM uses lattice spot pattern illumination instead of grid pattern illumination to generate higher contrast images and reduce the illumination. Various SIM methods have been developed to increase spatial resolution, temporal resolution, and imaging depth by combining a custom microlens array, total internal reflection (TIRF) microscopy, laser-based point-scanning microscopy, two-photon excitation microscopy, and a lattice light sheet. These different types of SIM have been described in detail in an excellent review article [137]. These SIM techniques achieve an approximately two-fold improvement in resolution in the XY plane (lateral, 115–120 nm) and the Z-axis (axial, 300 nm) compared to confocal microscopy. SIM microscopy instruments are commercially available from various suppliers, including GE Healthcare Life Sciences, Nikon, Oxford Nanoimaging, and Zeiss.

STED microscopy achieves subdiffraction limited resolution by combining a confocal microscope with a depletion laser. The depletion laser has a longer wavelength than the excitation laser and illuminates the sample in a doughnut-like ring pattern and in a coaxial manner with the excitation laser. The STED depletion laser returns the excited fluorophores to the ground state by stimulated emission [48]. The emission generated from this process has the same wavelength as the STED laser and can be separated from the emission light wavelength, i.e., the fluorescent signal. Thus, the internal diameter of the doughnut-like ring pattern of the depletion laser determines the area of fluorophore excitation and therefore the resolution of an image, which is 20 – 30 nm in the XY plane [3, 44, 71]. The two lasers scan the sample in a coaxial manner and generate an image in a manner similar to confocal microscopy. Therefore, a postprocessing of the raw data is not necessary to generate super resolution images. The depletion laser illumination can be split in the axial direction to form a Z-axis doughnut ring pattern for improving the Z-axis resolution, which can reach 100 nm or below. The resolution of STED microscopy is also improved using a technique called time-gating, which restricts the time window for collecting fluorescent signals after the initiation of excitation. The combination of time-gated detection with a pulsed excitation laser contributes to improving the resolution based on the differences between the fluorescence lifetimes of the fluorophores that are depleted by the STED laser [78, 130]. The time-gating detection also contributes to decreasing the power of the STED depletion laser and maintaining super resolution. The use of fluorescent dyes with longer emission wavelengths is disadvantageous for improving resolution in confocal microscopy. However, this is not the case for STED microscopy because the resolution is defined by the size of the doughnut-like ring pattern of the depletion laser. Thus, the use of red to far-red fluorescent dyes with a pulsed 775 nm STED laser is beneficial for expanding the fluorescent dye repertoire and improving live imaging, which will be described in a later section. STED microscopy instruments are commercially available from various suppliers, including Abberior Instruments and Leica Microsystems.

STORM uses a wide-field microscopy setup, activates few fluorophores stochastically with low intensity light, collects tens of thousands of frames (micrographs), and reconstructs one super resolution image by Gaussian fitting of each fluorophore in the XY plane according to the collected images [5, 10, 11, 50, 100, 108]. The stochastic activation of fluorophores is a random event that allows each fluorophore in the XY plane to be separated temporally. This temporal separation is spread over tens of thousands of frames (micrographs) collected during one imaging session. Many frames are necessary for the Gaussian fitting of the collected signals to determine the location of single molecules; therefore, STORM is also called Single Molecule Localization Microscopy. A super resolution image is reconstructed by collecting these single molecule data in one XY plane. STORM images are generated by placing each fluorophore at a calculated location in a single plane by pointillism. Thus, the final image is composed of dots. This postprocessing calculation process improves the lateral resolution beyond the diffraction limit. The lateral resolution of STORM is up to 20 nm, which is approximately 10 times higher than that of confocal microscopy. The Z-axis resolution is improved using a different principle. Fluorescent signals that are not exactly at the focal plane are slightly deformed due to the astigmatism-induced stretch in the X or Y direction. Based on the astigmatism-induced deformation of the signal from a single fluorophore, the axial location of the single fluorophore can be calculated to estimate the distance from the focal plane and the direction in terms of the Z-axis. This method improves the axial resolution to approximately 50 nm, which is also approximately 10 times higher than that of confocal microscopy. Super-resolution microscopy techniques similar to STORM that use the stochastic activation of fluorescent signals and SMLM techniques include PALM. Single molecule localization microscopy instruments are commercially available from various suppliers, including GE Healthcare Life Sciences, Nikon, Oxford Nanoimaging Limited, and Zeiss.

Recently, a variety of other techniques have been developed to improve the resolution of imaging. Some of these techniques will be briefly introduced here. The resolution of confocal microscopy has been improved by the postprocessing optical image data using deconvolution software, which is called confocal super resolution or super resolution imaging in contrast to super resolution microscopy. This type of approach is offered by several confocal microscopy suppliers, including Leica, Nikon, and Olympus. Meanwhile, Zeiss has improved the resolution by developing an array of detectors to acquire the Airy disk pattern of a signal and use this information to reconstruct an image. These new techniques have improved the resolution by approximately two-fold compared to that of confocal microscopy, which approaches the resolution of SIM. On the other hand, lattice light sheet microscopy has been developed to image live three-dimensional structures and dynamic events at sub-second intervals [19, 74, 138]. The resolution of this technique exceeds that of 3D images generated by high numerical aperture confocal or spinning disk microscopy; however, it does not exceed the resolution of widefield SIM [19]. Another technique, expansion microscopy, has been reviewed previously [123] and has also been combined with lattice light sheet microcopy [41]. In this review, we will focus on SIM, STED, and STORM super-resolution microscopy techniques.

Number of colors

Similar to conventional fluorescence microscopy techniques, super-resolution microscopy can be performed with multiple colors. The image construction of SIM depends on the postprocessing of nine to 15 acquired images with Moiré fringes; therefore, multicolor imaging is possible if the excitation and emission wavelengths are separated between the fluorophores [91, 140]. Commercially available SIM equipment from Nikon and Zeiss acquires two channels simultaneously to meet the fast imaging needs of live imaging and four to six different colors for multicolor requirements without simultaneous acquisition.

In STED microscopy, images are generated similarly to confocal microscopy by scanning the specimen with the excitation laser and acquiring the emission wavelength of each fluorophore. The difference from confocal microscopy is the need to match the emission wavelengths of the fluorophores with the wavelength of the STED laser for the depletion of the excited fluorophores to the ground state. Specifically, the long wavelength end of the emission spectrum needs to overlap with the wavelength of the STED depletion laser. Many commercially available fluorophores have emissions at long wavelengths and can be successfully combined with STED depletion laser wavelengths (592 nm, 660 nm, 775 nm). Therefore, multicolor STED microscopy is routinely performed. For example, dual-color STED analysis has been performed to study mouse NMJs [86], synaptic protein localization in central nervous system synapses [12], neuronal and synaptic morphology of live synapses [125] and nodes of Ranvier [26]. Triple-color STED was utilized to study the spine synapses of the central nervous system [54] and other biological samples [7, 18, 110, 121, 141]. Furthermore, quadruple-color STED has been used for live imaging and for imaging of fixed samples with only one excitation laser wavelength and one depletion laser wavelength [136]. The use of one depletion laser wavelength is recommended for the alignment of images with different fluorescent wavelengths in multicolor STED microscopy. Tables with the recommended fluorescent dyes and dye combinations for multicolor analysis can be found on the microscope manufacturers’ websites [4, 71]. In addition, bright SiR dyes have been successfully used for STED live imaging [92].

STORM requires fluorophores that switch between a non-emissive state and a ground state under high intensity light and a specialized buffer environment for these switching fluorophores. The fluorophore relaxes stochastically to the ground state and is then excited to emit fluorescent light. Fluorophores with these characteristics are called photoswitchable (photoblinking) dyes (for STORM) and photoswitchable fluorescent proteins (for PALM). Tables of the available colors can be found on the microscope manufacturers’ websites [5, 10], and new dyes are being developed [23]. The direct STORM (dSTORM or activator-free STORM) method allows the use of commercially available photoswitchable dyes conjugated to secondary antibodies without the need for activator-reporter pairs of dyes, which is beneficial for multicolor analysis. For multicolor STORM, dual color detection methods have been used for many publications [1, 6, 20, 25, 52, 57, 65, 66]. Commercially available STORM systems from Nikon and Zeiss allow the detection of two different fluorescent labels. Triple-color STORM analysis has been performed by combining STORM with a spectral demixing (spectral unmixing) method to resolve fluorophores that have overlapping emission spectra [69]. However, detecting three different colors in STORM is challenging, which is explained in the later section titled “Points of consideration for Super-Resolution Microscopy analyses”.

Imaging speed and live imaging

The temporal resolution of commercially available SIM microscopes is similar to STED microscopes and better than STORM. The temporal resolution of SIM is better than that of STORM because the number of raw micrographs needed to reconstruct a super-resolution image is only nine to 15 for SIM compared to tens of thousands for STORM. In addition, SIM requires a lower light intensity for imaging compared to STED and STORM; therefore, SIM is favored for live imaging despite the limited improvement in resolution. Several SIM methods have been developed to increase the temporal resolution by combining SIM with a custom microlens array, point scanning microscopy, or a lattice light sheet [137]. A commercial SIM model from Nikon offers an imaging frame rate of 15 frames per second (FPS). The recent development of SIM using lattice light and the Zeiss Elyra 7 system improved the frame rate to a very high speed of 255 FPS. SIM live imaging has been used to analyze various nervous system structures, including growth cones [39], growth cone cytoskeletons and vesicles of cultured neurons [87], dendritic spines of cultured hippocampal neurons [58], and dendrites and dendritic spines in brains of live zebrafish larvae or live mice [127]. Furthermore, live imaging based on SIM was successfully performed in the International Space Station to analyze cultured primary human macrophages exposed to microgravity [122].

STED microscopy is similar to confocal microscopy because both are scanning microscopes. A point scanning microscope allows the imaging of a smaller region of interest or a line to increase temporal resolution rather than scanning the entire field of view. STED microscopy has achieved video-rate fast live imaging; for example, 38.5 FPS was achieved for analyzing vesicle fusion and endocytosis [107], and 28 FPS was achieved for analyzing synaptic proteins in neurons [67]. A commercially available resonant scanner offers a very high scanning speed of 12,000 Hz and obtains an image of 512 lines in approximately 20 milliseconds (50 FPS), or can capture up to 428 FPS by restricting the region of interest [15, 16]. Furthermore, a custom-built STED microscope in the laboratory of Dr. Stefan W Hell reached >1,000 FPS with the use of an electro-optical deflector [103]. Leica offers two objective lens with a long working distance of 300 μm for live imaging STED microscopy: (1) a glycerol immersion lens with motorized correction collar, HC PL APO 93x/NA 1.3 Glycerol motCORR STED White lens and (2) a water immersion lens with motorized correction collar, HC PL APO 86x/NA 1.2 Water motCORR STED White lens [71]. Other examples of STED microscopy being successfully applied for live imaging include analyses of cytoskeletons in neurons [27, 28], interactions of mitochondria and endoplasmic reticulum [17], mitochondrial proteins [55], and many others [28, 54, 67, 81, 131, 132]. The use of a pulsed STED laser with a far-red 775 nm wavelength reduces phototoxicity compared to the use of a continuous wave STED lasers with shorter wavelengths. In addition, silicon-rhodamine (SiR) dyes have been used successfully for live imaging STED microscopy because these dyes have properties suitable for live imaging STED microscopy, as they are bright, photostable, without phototoxicity, and have a far-red spectrum of emission [92]. Cell-permeable SiR dyes are available commercially from Spirochrome for staining actin, tubulin, lysosome, and DNA. For live labeling of SNAP-tag or Halo-tag fusion proteins, cell permeable and fluorescently labeled SNAP substrates and Halo ligands are available from New England Biolabs and Promega [17, 38].

The STORM (PALM, SMLM) imaging speed is approximately 0.1 FPS, which is a lower temporal resolution than that of SIM and STED. The limited temporal resolution of STORM stems from the requirement for tens of thousands of frames (micrographs) to reconstruct a super resolution image [9, 57, 106, 135]. Multiple algorithms have been developed to increase the temporal resolution, including a multi-emitter fitting algorithm that identifies the locations of molecules that may have overlapping point spread functions. However, the quality of STORM image reconstruction depends significantly on the analysis software package used [101, 116, 118, 120], which will be discussed in the next section.

Points of consideration for Super-Resolution Microscopy analyses

Various super-resolution microscopy techniques are commercially available; however, each technique has advantages and disadvantages that are worth consideration prior to choosing the method used for a project. For example, if the project requires the highest resolution, STORM and STED offer the best resolution currently available as a commercially available super-resolution microscope (lateral resolution of 20 to 30 nm). The preference might be affected by the use of a reconstructed image by the former versus that of an optical image by the latter. If the project requires information of protein numbers, STORM has been demonstrated to offer quantitative information [35].

Sample preparation should follow the manufacture provided guides, which describes the required embedding/mounting condition [71, 79]. A suitable sample thickness depends on the microscopy type. SIM is suitable for tissue thickness up to 10 to 15 μm and is not suitable for thick tissue due to out of focus light interfering with the computational removal of these noise [46]. STED has been shown to image 150 μm into tissue at subdiffraction limited resolution by combining optical clearing of the specimen and the objective lens with motorized correction collar, HC PL APO 93x/NA 1.3 Glycerol motCORR STED White lens. Leica provides a protocol for this type of deep tissue imaging [70, 129]. STORM has been combined with TIRF microscopy and thus has a shallow imaging depth below one micrometer. Improved protocols have been reported to image few μm to 30 μm thick sections using STORM [42, 49].

At this level of resolution, sizes of antibodies used for immunohistochemistry become relevant. An immunoglobulin has a size of 12 to 14 nm; therefore, nanobodies or single domain antibodies are preferred due to the smaller size of 2.8 to 4.4 nm [117]. The green fluorescent protein (GFP) used for PALM is similar to nanobodies and has a size of 2.4 to 4.2 nm [51]. Also, resolution is improved by directly labeling the protein of interest using bioorthogonal click chemistry combined with small organic dyes [84].

If the study requires multicolor analysis, SIM and STED microscopy offer multicolor analysis with a larger number of colors in an image. Three-color analyses are easily achievable for STED microscopy, and four-color analyses are possible [136]. However, three-color analyses are difficult for STORM due to the specific buffer requirement for each photoswitchable fluorophore [5]. In addition to the limitation of color number, STORM requires the use of a higher concentration of primary antibodies to fulfill the requirement to saturate the signal of the specimen [5, 10]. Therefore, the staining conditions may differ from those used in the usual immunohistochemistry protocol, and the signal-to-noise ratio needs to be considered carefully.

If acquisition speed is a priority to capture fast, dynamic events by live imaging, SIM and STED offer a higher frame rate than STORM. For live imaging, the intensity of excitation light and photobleaching are significant factors for consideration, and these are common problems for any live imaging technique. SIM uses lower intensity light and is considered suitable for live imaging. Recent STED microscopes are equipped with pulsed lasers instead of continuous wave lasers and highly sensitive hybrid detectors to reduce the excitation light intensity. In addition, photobleaching becomes relevant when the project requires 3D analysis and multiple scanning, such as when generating a z-stack. Newer generation fluorophores that are bright and stable/bleach resistant have been developed recently and should be considered for improving super-resolution imaging, including Alexa Fluor & Alexa Fluor Plus (Thermo Fisher), ATTO (Sigma-Aldrich), CF (Biotium), SiR (Spirochrome, New England Biolabs), and STAR (Abberior) dyes [4, 71, 92].

The data file size of the image differs between these techniques. STED microscopy generates file sizes similar to those of confocal microscopy because STED generates an optical image raw data file as the final super resolution image. On the other hand, STORM requires tens of thousands of frames (micrographs) to reconstruct a super-resolution image. Thus, the typical data file size of STORM is much larger than that of STED and can exceed 5 gigabytes for one final super-resolution image [109]. Therefore, STORM requires significant capacity for data transfer and storage and computing power for processing the data.

Finally, the STORM image reconstruction quality depends significantly on the analysis software used, and analysis software packages were compared recently by using a common data set [101, 116]. There seems to be a considerable difference in the outcomes, which may affect the conclusion drawn from the reconstructed images. This was the case for the 3D analysis more than the 2D analysis. Validation is essential for any research technique, but it is essential for STORM to validate the use of different approaches because a reference raw data image does not exist for this technique. In contrast, STED microscopy generates optical images as raw data, which reduces concerns about postprocessing and analysis software differences. In addition, STED images can be compared against confocal microscopy images obtained using the same equipment in a sequential manner.

SIM for NMJ analysis

Three-dimensional structured illumination microscopy, 3D-SIM [46], has been used to analyze the role of the large adaptor protein Ankyrin2-L in the stabilization of Drosophila larval NMJs through the regulation of presynaptic microtubules and cell adhesion molecules [90]. 3D-SIM revealed a repetitive lattice structure with approximately 200 nm periodicity comprised of Ankyrin2-L and β-spectrin in axons but a less organized structure in presynaptic boutons. The function of Ankyrin2-L, which is to crosslink synaptic membrane proteins and spectrins to the microtubule cytoskeleton, is required to stabilize the presynaptic terminals. In addition, 3D-SIM was also used to analyze the role of the noncanonical BMP signaling pathway in the maturation of Drosophila larval NMJs [119]. Furthermore, 3D-SIM has been used to detect immunohistochemical signals and has been combined with confocal microscopy for detecting fluorescent in situ hybridization signals in Drosophila larval NMJs [124]. This publication describes the detailed protocol of the experimental procedure with the methods to check the quality of the 3D-SIM raw data and reconstruction with an open source ImageJ plugin called SIMcheck [8].

By using 3D-SIM in mouse NMJs, the distribution pattern of synaptic proteins, including the presynaptic active zone protein Piccolo and the postsynaptic proteins rapsyn, voltage-gated sodium channel, and integrin α7, was analyzed relative to that of the postsynaptic junctional folds [139]. Consistent with previous reports, these proteins were observed at the expected locations of NMJs.

STED microscopy for NMJ analysis

The first application of super-resolution microscopy for the analysis of NMJs was the imaging of the active zones of Drosophila NMJs using STED microscopy (Table 2). Active zones are synaptic vesicle release sites at the presynaptic membrane [24]. Active zones appear in electron microscopy images as T-bars in Drosophila NMJs or as small aggregates in vertebrate NMJs. In Drosophila larval NMJs, these active zones can be labeled with the nc82 antibody (DSHB), which stains the active zone-specific protein Bruchpilot. This protein was previously shown to be distributed in a punctate pattern at motor nerve terminals by confocal microscopy without any obvious structure inside each punctum. The subdiffraction-limited resolution of STED microscopy revealed a structure within each punctum for the first time and elucidated a ring-like pattern of Bruchpilot protein in each of these puncta [61]. The distribution pattern of the Bruchpilot protein was further analyzed using antibodies recognizing the N-terminus and the C-terminus (nc82) of this protein. STED microscopy revealed that the N-terminus of Bruchpilot protein resides inside the ring-like pattern of C-terminus of Bruchpilot protein, which was detected by the nc82 antibody in the en face view of the NMJs (top-down view looking at the active zones from the cytosolic side) [40]. Based on the transverse section view of the NMJ (cross-section, perpendicular view of the synapse), the N-terminus of Bruchpilot protein resided next to the voltage-gated calcium channels (VGCCs, Cacophony labeled with GFP) in the presynaptic membrane, whereas the C-terminus of Bruchpilot protein was distributed away from the presynaptic membrane. These results were consistent with the immunoelectron microscopy analysis of the T-bars at Drosophila NMJs using the same set of antibodies that recognized the N- and C-termini of Bruchpilot [40].

Table 2:

Super resolution microscopy techniques used for NMJ analyses

Technique	Species	Analysis
SIM	Drosophila	The role of Ankyrin2-L in the stabilization of larval NMJs [90]. The roles of noncanonical BMP signaling pathway for the maturation of larval NMJs [119].
		Combined SIM to detect immunohistochemical signals and confocal microscopy to detect fluorescent in situ hybridization signals at larval NMJs [124].
	Mouse	The distribution pattern of the presynaptic active zone protein Piccolo and the postsynaptic proteins rapsyn, voltage-gated sodium channel, integrin α7, and postsynaptic junctional folds [139].
	Human	N/A
STED	Drosophila	The distribution pattern of Bruchpilot protein. The N-terminus of Bruchpilot resided next to the vGcCs in the presynaptic membrane, whereas the C-terminus of Bruchpilot protein distributed away from the presynaptic membrane [40, 61]. The distribution pattern of Abp1, DCSP3, Futsch [63], Liprin-α and glutamate receptors [40], RIM-BP [73], Unc13A and Unc13B [13], and Unc18 and Syntaxin-1A [95, 128]. Active zone proteins including Bruchpilot, Rim-BP, and Unc13A show puncta number between 2 to 6 per one active zone unit structure of NMJs. [43].
	Mouse	The distribution patterns of Bassoon, Piccolo, and PQ-VGCC in adult and aged NMJs. This study revealed the selective degeneration of active zone proteins in aged NMJs [86].
	Human	N/A
STORM	Drosophila	Quantification of Bruchpilot protein number per active zone in NMJs of wild-type and Rab3 mutant larvae [35].
		The roles of Bruchpilot and synaptotagmin for rapid glutamate release [89].
		The roles of Bruchpilot and complexin for tethering synaptic vesicles to active zones [104].
	Mouse	The distribution pattern of acetylcholine receptor in junctional folds [139] and SNAP25 in adult NMJs [56].
	Human	The distribution pattern, puncta size, density, and signal intensity of SNAP25 in adult NMJs [56].

Abp1= actin-binding protein 1; BMP= bone morphogenetic protein; DCSP3= Drosophila cysteine string protein 3; NMJs= neuromuscular junctions; Rab3= Ras-related protein 3; SIM= structured illumination microscopy; SNAP25= synaptosome associated protein 25; STED= stimulated emission depletion; STORM= stochastic; optical reconstruction microscopy; RIM-BP= Rab3-interacting molecule-binding protein; Unc= uncoordinated

Further analyses of Drosophila NMJs using STED microscopy revealed the nanostructure distribution patterns of synaptic proteins, including actin-binding protein 1 (Abp1), Drosophila cysteine string protein 3 (DCSP3, marker for synaptic vesicles), and Futsch (microtubule-enriched area) [63], Liprin-α and glutamate receptors [40], Rab3-interacting molecule–binding protein (RIM-BP) [73], Unc13A and Unc13B [13], and Unc18 and Syntaxin-1A [95, 128]. Based on some these of studies, a model of the molecular nanostructure of Drosophila NMJs has been proposed [43]. Interestingly, many of these active zone proteins distribute in punctate clusters and do not show diffuse distribution at Drosophila NMJs. When these active zone proteins show increases in the protein quantity, they show the increases in the number of puncta or the quantity of proteins in each puncta [14]. Active zone proteins, including Bruchpilot, Rim-BP, and Unc13A, show puncta numbers between 2 and 6 per active zone unit structure in Drosophila NMJs. Unc13A is essential for both short-term plasticity that occurs over minutes and for long-term plasticity that occurs over hours to days [14]. Importantly, Bruchpilot protein is aberrantly organized at larval NMJs in a Drosophila Amyotrophic lateral sclerosis (ALS) model, which expresses human FUS with ALS-causing mutations in motor neurons [105]. These data suggest that active zone proteins organize a dynamic structure that plays a role in synaptic plasticity and is degenerated in neurological diseases.

STED microscopy was also used to study the active zones of adult mouse NMJs. Adult NMJs rely on P/Q-type VGCCs for synaptic transmission [93, 98]. STED microscopy revealed that calcium channels cluster in puncta and are distributed at roughly evenly-spaced distances throughout the presynaptic membrane [86]. Most of these calcium channel puncta colocalized with the active zone-specific protein Bassoon, which also clustered in puncta [86]. The colocalization of P/Q-type VGCCs and Bassoon was consistent with biochemical analyses demonstrating the direct or indirect interaction of these two proteins [21, 30]. Bassoon has structural similarities with another active zone protein, Piccolo, and these two proteins are believed to colocalize at nerve terminals in the central nervous system. Unexpectedly, STED microscopy revealed that Bassoon and Piccolo reside in a side-by-side manner, in which Piccolo puncta form a “sandwich” around a Bassoon punctum [86]. Interestingly, this distribution pattern showed similarities with an active zone model based on electron microscope tomography of mouse NMJs [82]. Further study of aged NMJs using STED microscopy revealed that the levels of PQ-type VGCCs and Bassoon proteins were significantly decreased, while Piccolo protein was resistant to age-related changes and remained at a similar level as that found in NMJs in young adult mice [86]. For the first time, the selective degeneration of active zone proteins in NMJs of aged animals has been revealed owing to the subdiffraction limited resolution of STED microscopy.

STORM for NMJ analysis

Similar to STED microscopy, STORM was used initially for the analysis of Drosophila larval NMJs. Active zones of Drosophila larval NMJs were analyzed quantitatively using direct STORM (dSTORM). As a Single Molecule Localization Microscopy method, dSTORM was used to quantify endogenous proteins [35]. The number of active zone-specific Bruchpilot proteins has been estimated to be 137 proteins per active zone, and three quarters of these proteins are organized into approximately 15 different positions with seven Bruchpilot proteins per position. This quantitative dSTORM technique successfully showed an approximately 60% increase in the Bruchpilot protein number per active zone in NMJs of Rab3 mutant larvae, which were previously known to have altered synaptic transmission and enlarged Bruchpilot aggregates [35]. Furthermore, STORM was used to analyze the roles of Bruchpilot and synaptotagmin in rapid glutamate release [89] and the roles of Bruchpilot and complexin in tethering synaptic vesicles to active zones [104] in Drosophila larval NMJs. Excellent reviews have been published elsewhere describing the use of STORM for Drosophila larval NMJ studies [33, 34].

At mouse NMJs, the acetylcholine receptor distribution pattern in junctional folds was compared between SIM images, STORM images, and electron microscopy ultrastructure images. Based on the STORM images, the authors proposed a novel acetylcholine receptor distribution pattern, in which the receptor concentration is high at the shoulders of the junctional fold mouth but low in the ridges [139]. This distribution pattern is slightly different from the classical understanding, in which acetylcholine receptors have been thought to be distributed at the top and shoulders of the junctional fold ridges and not at the bottom of the folds. It would be interesting to validate this new model of the acetylcholine receptor distribution pattern using a different technique, such as immunoelectron microscopy, or using STORM with transverse section images of the NMJs because the new STORM analysis was based on en face images.

Human and mouse NMJs were compared using STORM to analyze the distribution pattern of synaptosome associated protein 25 (SNAP25), one of the SNARE proteins essential for synaptic vesicle fusion. In mouse NMJs, SNAP25 was distributed in a punctate fashion at a density (15 per μm²) [56] similar to the puncta density of the active zone proteins Bassoon and Piccolo (10 per μm²) [86]. Interestingly, human NMJs showed greater values for SNAP25 compared to mouse NMJs in terms of puncta size, puncta density per NMJ area, and the signal intensity of immunohistochemistry [56]. Based on these findings, the authors suggested that there were differences between human and mouse NMJs.

Super-Resolution Microscopy for brain synapse analyses.

Synapses of the central nervous system have been analyzed extensively using super-resolution microscopy. To mention just few examples, the first synaptic protein analyzed using STED microscopy was the presynaptic calcium sensor synaptotagmin in cultured hippocampal neurons [134]. The movement of synaptic vesicles in cultured hippocampal neurons was recorded using STED microscopy at a video rate live imaging [133]. Pre- and postsynaptic proteins in the main olfactory bulb of mice were analyzed using 3D-STORM [29]. The distribution patterns of synaptic proteins within small central nervous synapses were analyzed using STORM with fixed samples and STED live imaging [32, 54]. For a comprehensive review of the analysis of central nervous system synapses using super-resolution microscopy, many excellent reviews have been published [31, 75–77, 88, 97, 99, 112, 113, 126].

Conclusion

Super-resolution microscopy studies of Drosophila NMJs have come a long way and revealed impressive nanostructures and the molecular identity of active zones previously known as T-bars in electron micrographs. The combination of good molecular markers (for example, Bruchpilot for active zones) and rich genetic mutant resources provided a wealth of opportunity to study the molecular mechanisms of the physiological modification of Drosophila NMJs using super-resolution microscopy. Super-resolution microscopy studies of mammalian NMJs are still in their infancy. To our knowledge, these super-resolution methods have not been used for live imaging of vertebrate NMJs. However, vertebrate NMJs have been studied using electron microscope tomography in frogs and mice, which revealed the ultrastructural details of the presynaptic active zones. Super-resolution microscopy provides the opportunity to reveal the molecular identities of these modeled structures for a better understanding of vertebrate synapses. The synapse is a small structure with a complex molecular mechanism that is necessary to accomplish synaptic transmission and plasticity, and the development of super-resolution microscopy has improved the understanding of the underpinning molecular mechanisms.

Highlights.

• Advantages and disadvantages of SIM, STED, and STORM super-resolution microscopy

• Suitability of super-resolution microscopy types for synapse analyses

• Molecular nanostructures of neuromuscular junctions revealed using super-resolution microscopy

Abbreviations

Abp1

actin-binding protein 1

ALS

Amyotrophic lateral sclerosis

BMP

bone morphogenetic protein

CNS

central nervous system

DCSP3

Drosophila cysteine string protein 3

DSHB

Developmental Studies Hybridoma Bank

FPS

frames per second

GFP

green fluorescent protein

NMJ

neuromuscular junctions

Rab3

Ras-related protein 3

SIM

structured illumination microscopy

SiR

silicon-rhodamine

STED

stimulated emission depletion

STORM

stochastic optical reconstruction microscopy

SMLM

Single molecule localization microscopy

Lrp4

low-density lipoprotein receptor-related protein

PALM

photoactivated localization microscopy

RIM-BP

Rab3-interacting molecule–binding protein

SNARE

soluble N-ethylmaleimide sensitive factor attachment protein receptor

SNAP25

synaptosome associated protein 25

TIRF

total internal reflection

UNC

uncoordinated