1.

Introduction

The unique combination of molecule-specific labeling, minimal perturbation of biological complexes, and three-dimensional (3D) imaging makes fluorescence microscopy the most popular imaging modality in cell biology. To understand how a cell works and which mechanisms occur in the case of a dysfunction or disease, it is essential to understand how biomolecules are organized into complexes and the dynamics of this organization. This becomes especially apparent when we aim to visualize complex neuronal networks in brains connected by synapses. Synaptic transmission, the biological process by which a neuron communicates with a target cell across a synapse, relies on spatially and temporally coordinated multistep processes that require the precise organization of proteins. In the case of chemical synapses, transmission involves the release of a neurotransmitter from the pre-synaptic neuron, and neurotransmitter binding to specific post-synaptic receptors.^1,2 However, despite their relevance for efficient transmission, knowledge about the molecular organization of proteins at synapses and their interplay remains limited. This is mainly due to the fact that major components of synapses are too small to be resolved by standard fluorescence microscopy, which is restricted to $\sim 200\mathrm{nm}$ by the diffraction barrier. For example, synaptic vesicles, key elements for transmitter cargo, exhibit diameters of 40 to 50 nm and release their transmitters into the synaptic cleft with a width of 20 to 30 nm where they finally bind to postsynaptic receptors.^3–5 Therefore, electron microscopy was for decades the only method capable of resolving the nanoscale characteristics of synapses.^6,7

In the last decade, super-resolution (SR) microscopy has evolved into a powerful method for sub-diffraction resolution fluorescence imaging of cells and structural investigations of cellular organelles.⁸ SR microscopy methods can now provide a spatial resolution that is well below the diffraction limit of light microscopy, enabling invaluable insights into the spatial organization of proteins at synapses.⁵ However, current SR measurements become error-prone below 20 nm.^9–12 Even though the combination of sequential structured illumination with single-molecule localization microscopy enabled fluorescence imaging in cells approaching 10 nm resolution,^13,14 these methods are highly complex requiring expert knowledge and sophisticated microscopes. In addition, they are limited to a few fluorophores that exhibit suited blinking performance in special photoswitching buffers or can be photoactivated by irradiation with ultraviolet light, which does not allow fluorescence imaging with more than 2 colors with 20 to 30 nm spatial resolution. In addition, the large number of proteins involved in synaptic signaling located within the small size of synaptic compartments creates a crowded situation that impedes fluorescence labeling at high densities required to resolve the molecular organization of synapses.^15,16

Hence, the choice of labeling strategy is critical as it affects the achievable resolution via the linkage error (the displacement of the fluorophore from the target molecule) and the labeling density. Standard immunolabeling with primary and secondary antibodies, which is the method of choice for labeling of synaptic proteins, causes linkage errors between 10 and 20 nm.^11,17 Furthermore, antibodies have to compete for epitope accessibility of densely packed synaptic proteins. Fab fragments and nanobodies are smaller than antibodies and therefore are good alternatives but efficient nanobodies are not available for every synaptic protein and even a size of only a few nanometers hampers high density labeling required to achieve a structural resolution of 20 to 30 nm.^15,16 Here, we show how post-expansion immunolabeling of approximately eightfold expanded hippocampal neurons enables high density labeling of synaptic proteins and in combination with Airyscan and structured illumination microscopy (SIM) 3D fluorescence imaging of up to 4 target molecules with 20 to 30 nm lateral resolution.

2.

Materials and Methods

3.

Crosslinking, Gelation, and Denaturation

Pre-stained neurons or untreated fixed neurons were crosslinked with 4% formaldehyde (Sigma, F8775-25ML) and 30% acrylamide (Sigma, A4058) for 18 h at 37°C. For the gelation, 10-fold robust expansion microscopy (TREx) monomer solution¹⁸ consisting of 1.1 M sodium acrylate (Sigma, 408220), 2.0 M acrylamide, 0.009% N,N′-methylenebisacrylamide (Sigma, M1533), and 1× PBS was prepared, aliquoted, and stored at $-20°\mathrm{C}$. Freshly thawed monomer solution was mixed with 0.15% N,N,N′,N′-tetramethylethylenediamine (Sigma, T7024) and 0.15% ammonium persulfate (Sigma, A3678) shortly before gelation. A drop of $60\mu \mathrm{l}$ gel solution was pipetted on a piece of parafilm inside a gelation box on ice, and a coverslip was flipped onto it with the cell side facing down. The gels polymerized in a humidified chamber for 3 h at RT. To homogenize the solidified gels, denaturation buffer consisting of 200 mM sodium dodecyl sulfate (Invitrogen, AM9820), 300 mM NaCl (Sigma, S5886), and 50 mM tris-(hydroxymethyl)-aminomethan (Sigma, 1.08382) was prepared and adjusted to pH 9. After the buffer was pre-warmed to 100°C, gels were denatured for 1 h at 100°C. For post expansion, immunostaining gels were washed 3 times for 30 min in 14 cm petri dishes with 1× PBS.

4.

Results

Currently, immunolabeling represents the method of choice for visualization of synaptic proteins in pre- and post-synaptic compartments but the size of antibodies prevents dense labeling with minimal linkage error required for reliable multicolor fluorescence imaging with 20 to 30 nm nanometer spatial resolution. This labeling barrier thus determines the achievable structural resolution independent of the resolution capacity of the used SR microscopy method. Fortunately, there is a solution to this problem that enables improved labeling of synaptic proteins in dense compartments with reduced linkage error and multicolor super-resolved fluorescence imaging on standard microscopes. The approach, termed expansion microscopy, uses the physical expansion of the cellular structure by linking proteins into a dense, cross-linked network of a swellable polyelectrolyte hydrogel that expands after enzymatic digestion or denaturation in water.

Since its first introduction by Chen et al.,²⁰ expansion microscopy has shown impressive results, including the magnified visualization of proteins, lipids, and nucleic acids with fluorescent proteins, antibodies, functionalized sphingolipids, and oligonucleotides, respectively, in cells, tissues, and human clinical specimen.^21,22 It is possible to expand samples in series enabling gel expansion factors of 20× and higher²³ but such high expansion factors reduce the labeling density and consequently the signal-to-noise ratio thus requiring highly sensitive imaging methods. Airyscan confocal and SIM provide only moderate resolution improvement factors of $\sim 1.4$²⁴ and $\sim 2.0$⁸ translating into lateral resolutions of up to $\sim 140$ and 100 nm, respectively, but they enable 4-color fluorescence imaging with various fluorophores spanning the entire visible spectrum. Hence, expansion factors of 8 to 10× allow us to visualize the distribution of up to 4 proteins simultaneously in synapses at lateral resolutions of at least 20 to 30 nm.

To compare the labeling densities of pre- and post-expansion immunostaining, we first labeled the pre-synaptic proteins VGluT1 and Munc13-1 in hippocampal neurons with identical primary and secondary antibodies before and after expansion (Fig. 1).^25–28 After the first pre-expansion immunolabeling step, we used TREx²⁹ applying formaldehyde and acrylamide as an anchoring molecule for proteins followed by high-temperature denaturation after gelation. Proteinase K digestion was omitted to preserve protein epitopes. After initial approximately threefold expansion in PBS, synaptic proteins were immunostained once again with the same primary and secondary antibodies labeled with a different fluorophore overnight. After washing, the samples were placed in ${\mathrm{ddH}}_2\mathrm{O}$ to fully expand approximately eightfold as determined from the comparison of the hydrogel size before and after expansion and the cross-sectional profiles of pre-expansion immunolabeled neurofilament L¹⁸ in primary neurons before and after expansion using the same protocol (Fig. S1 in the Supplementary Material). The expansion factor is slightly lower than in the original TREx protocol most likely due to the protease-free treatment. Furthermore, it has to be considered that expansion factors may not be perfectly homogenous throughout the sample and can vary for different cellular organelles.³⁰ The resulting Airyscan microscopy and SIM images clearly demonstrate that our protocol exhibits excellent protein retention, i.e., protein epitopes survive the denaturation, gelation, and expansion step in the absence of proteinase K, i.e., synaptic proteins can be efficiently immunostained post-expansion (Fig. 1).

Fig. 1

Post-expansion immunostaining enables efficient labeling of proteins in crowded synaptic compartments. (a) Airyscan image (maximum projection of $z$-stack) of approximately eightfold expanded hippocampal neurons stained pre- and post-gelation with the same primary VGluT1 antibody. The secondary antibodies were labeled with CF568 (pre-gelation, magenta) and AF488 (post-gelation, cyan). Scale bar $10\mu \mathrm{m}$. (a1) Magnified images (single $z$-stack slice) of the marked pre-synaptic vesicle bouton from panel (a) show that post-expansion immunolabeling resolves more synaptic details. Scale bars $2\mu \mathrm{m}$. (b) $\mathrm{SI}{\mathrm{M}}^2$ image (maximum projection of $z$-stack) of approximately eightfold expanded hippocampal neurons stained pre- and post-gelation with the same primary Munc13-1 antibody. The secondary antibodies were labeled with ATTO643 (pre-gelation, magenta) and CF568 (post-gelation, cyan). Scale bar $10\mu \mathrm{m}$. (b1) Magnified images (single $z$-stack slice) of the marked active zone show that post-expansion immunolabeling resolves more synaptic details and exhibits a smaller linkage error. Scale bars $2\mu \mathrm{m}$. Scale bars of expanded hydrogels are shown.

The slightly higher labeling densities measured for post-expansion immunolabeling corroborate that additional protein epitopes become accessible for antibody labeling after threefold expansion [compare pre- and post-expansion labeled images in Figs. 1(a1) and 1(b1)]. An additional benefit of post-expansion labeling is a reduction of the linkage error proportional to the expansion factor compared to pre-expansion labeling. In our experiments, the immunolabeling linkage error of 17.5 nm defined by the primary and secondary antibodies³¹ decreases to only $\sim 2\mathrm{nm}$ in approximately eightfold expanded samples. Ultimately, the improved epitope accessibility and reduced linkage error of post-expansion immunolabeling facilitate the visualization of individual pre-synaptic vesicles (Fig. S2 in the Supplementary Material).

Next, we performed multicolor post-expansion immunolabeling of the postsynaptic density protein Homer1³² (ATTO643) and RIM1/2 (CF568), a key component of the pre-synapse that regulates vesicle docking and transmitter release efficiency at the active zone.^33,34 In addition, we immunolabeled Synaptobrevin 2 (AF488), the major SNARE protein (soluble N-ethylmaleimide sensitive factor attachment protein receptor) with a high copy number on synaptic vesicles required for fast calcium-triggered synaptic-vesicle exocytosis and endocytosis.^35–38 Finally, we labeled lysine residues of all synaptic proteins with CF405M-NHS to visualize local protein densities.³⁹ Similar to correlative light and electron microscopy, this approach enables the localization of immunolabeled proteins in the cellular context provided by protein density. Airyscan microscopy and SIM images of eightfold expanded and post-expansion labeled hippocampal neurons clearly demonstrate that CF405M-NHS labeling enables identification of synapses because of the higher protein density (Fig. 2 and Fig. S3 in the Supplementary Material). Furthermore, Airyscan microscopy and SIM side view images of approximately eightfold expanded synapses show that synaptic clefts can be easily resolved confirming that the method achieves a lateral resolution of at least 20 to 30 nm (Fig. 2 and Fig. S3 in the Supplementary Material).

Fig. 2

Synaptic protein organization visualized by four-color imaging of approximately eightfold expanded synapses. (a) Representative Airyscan overview image of hippocampal neurons using NHS-staining (cyan, CF405M) and immunostaining for Synaptobrevin2 (green, AF488), RIM1/2 (orange, CF568), and Homer1 (magenta, ATTO643). Scale bar $10\mu \mathrm{m}$. (a1) Magnified side view images of a single synapse marked by the white rectangle in panel (a). Scale bars $2\mu \mathrm{m}$. (a2, a3) Additional side view and slightly tilted (a3) synapse images. Scale bars $2\mu \mathrm{m}$. (b) Representative SIM² overview image with whole proteome NHS-staining (cyan, CF405M) and immunostaining for Synaptobrevin2 (green, AF488), RIM1/2 (orange, CF568), and Homer1 (magenta, ATTO643). Scale bar $10\mu \mathrm{m}$. (b1, b2) Magnified images of single synapses marked by the white rectangles in panel (b) show a side and frontal view of a synapse. Scale bars $2\mu \mathrm{m}$. (b3) Additional synapse side-view image from another overview image. Scale bars $2\mu \mathrm{m}$. Scale bars of expanded hydrogels are shown (Video 1, MP4, 13.1 MB [URL: https://doi.org/10.1117/1.NPh.10.4.044412.s1]).

Closer inspection of the Homer1 and RIM1/2 signals reveals that the pre-synaptic as well as postsynaptic protein densities are not homogeneously distributed along the synaptic cleft (Fig. 2, Fig. S3 in the Supplementary Material, and Video 1). In particular, RIM1/2 appears to be compartmentalized in sub-synaptic domains (SSDs).^28,40 Recently, it has been shown that postsynaptic SSDs at excitatory synapses are aligned with pre-synaptic SSDs, forming so-called trans-synaptic nanocolumns.⁴¹ Another SR microscopy study recently demonstrated the existence of trans-synaptic nanocolumns comprising SSDs of RIM1/2 at the pre-synapse and gephyrin, glycine receptors, and gamma-aminobutyric acid type A receptors at the postsynapse showing that this concept is conserved at inhibitory synapses.⁴² This led to the proposal that SSDs are functional units corroborating efficacy of neurotransmission and synaptic plasticity.^40,41 Closer inspection of our expanded 2D and 3D four-color fluorescence images shows that synaptic vesicles close to the pre-synaptic membrane overlay with the RIM1/2 signals (Fig. 2, Fig. S3 in the Supplementary Material, and Video 1).

5.

Discussion and Conclusions

Our data demonstrate that approximately eightfold expansion of neurons in combination with post-expansion immunolabeling enables 3D 4-color SR fluorescence imaging with 20 to 30 nm lateral resolution and minimal linkage error (Figs. 1 and 2, Figs. S2 and S3 in the Supplementary Material, and Video 1). The fact that we can resolve individual immunolabeled neurofilaments with a diameter of 10 nm¹⁸ in densely packed neurofilament bundles in neurons confirms that our method actually achieves at least 20 nm lateral resolution (Fig. S4 in the Supplementary Material). Such high resolutions and the ability to visualize three synaptic proteins simultaneously in context to the overall protein density outperforms approaches with higher expansion factors because they come along with higher dilution factors and consequently lower signal-to-noise ratios impeding high quality fluorescence imaging by confocal microscopy and SIM. Certainly, imaging of expanded neurons by single-molecule localization microscopy facilitates higher spatial resolutions but requires re-embedding of the expanded samples in a neutral polyacrylamide gel and remains limited to two-color imaging.^27,43,44 Using immunolabeling of Synaptobrevin 2 as an efficient marker of the vesicle pool^45,46 and four-color SR fluorescence imaging allowed us to identify individual synaptic vesicles colocalized with fusion sites (immunolabeling of RIM1/2) (Fig. 2, Fig. S3 in the Supplementary Material, and Video 1).^41,47,48 Therefore, we expect that multicolor SR expansion microscopy in combination with post-expansion immunolabeling is ideally suited to address important open questions about the 3D molecular organization of synapses.