Abstract
The biology of multicellular organisms is coordinated across multiple size scales, from the sub-nanoscale of molecules to the macroscale, tissue-wide interconnectivity of cell populations. Here we introduce a method for super-resolution imaging of the multiscale organization of intact tissues. The method, called magnified analysis of the proteome (MAP), linearly expands entire organs four-fold while preserving their overall architecture and three-dimensional proteome organization. MAP is based on the observation that preventing crosslinking within and between endogenous proteins during hydrogel-tissue hybridization allows for natural expansion upon protein denaturation and dissociation. The expanded tissue preserves its protein content, its fine subcellular details and its organ-scale intercellular connectivity. Off-the-shelf antibodies can be used for multiple rounds of immunolabeling and imaging of a tissue's magnified proteome, with our experiments demonstrating a success rate of 82% (100/122 antibodies tested). We show that specimen size can be reversibly modulated to image both inter-regional connections and fine synaptic architectures in the mouse brain.
Biological systems such as the mammalian brain consist of thousands of distinct cell types forming highly interconnected functional networks1–5. Understanding how these diverse cells interact to generate systems-level responses is essential for many fields of biology. Deciphering these complex interactions will benefit from tools that can simultaneously characterize the molecular identity and fine subcellular architectures of individual cells as well as their system-level connectivity, because these properties together determine many cell functions. To date, technologies only provide a subset of the required multilevel information.
Proteomic imaging techniques can potentially provide multilevel information in both animals and human samples6–10. The proteome is an ideal substrate for the integrated analysis of functional components owing to proteins' unmatched diversity, functional roles, and distinct subcellular localization at single-molecule precision. The existing large antibody libraries (close to 100,000 antibodies), once rigorously validated, could enable the detection of more than 70% of the human proteome and corresponding cellular architectures10.
For example, multiplexed proteomic imaging techniques (e.g., array tomography and SWITCH) can provide molecular details for individual cells and may allow reconstruction of their surrounding tissue environment11–13. Emerging intact tissue clearing approaches preserve proteins and the continuity of neural fibers, which may enable reconstruction of immunolabeled neural architectures14,15. Super-resolution imaging of immunolabeled thin tissue sections has been successfully used to characterize minute subcellular structures (e.g., synapses)16–19. In addition to conventional super-resolution approaches, Chen et al.20 have recently demonstrated that protease digestion of a hydrogel-tissue hybrid homogenizes its mechanical characteristics and allows approximately four-fold linear expansion of the hybrid. Using this approach, termed expansion microscopy (ExM), they demonstrated super-resolution imaging of thin tissue sections with custom-made antibody probes and diffraction-limited microscopes20.
Despite the success of these new technologies, integrated reconstruction of the fine subcellular architectures, molecular details, and intercellular connectivity of diverse cell types in large-scale biological systems remains an unmet goal in biology, as these methods can currently only provide a subset of the required multilevel information. For instance, the protease digestion step in ExM causes loss of proteins, which limits the number of protein structures that can be imaged in the same sample. In addition, inter-cellular connectivity is largely lost in ExM as it requires tissue sectioning to allow immunolabeling of an unprocessed tissue with limited permeability. By contrast, emerging whole-tissue clearing methods preserve neural connectivity and endogenous proteins, but accurate reconstruction of neural architectures remains challenging owing to the low resolution of diffraction limited microscopy.
Here we present MAP, a method that enables multiscale proteomic imaging of intact biological systems for combined extraction of the molecular identity, subcellular architectures, and intercellular connectivity of diverse cell types within a single tissue. MAP preserves both the three-dimensional (3D) proteome content and organization and organ-wide cellular connectivity within an intact tissue-hydrogel hybrid while rendering it reversibly size-adjustable up to four- to five-fold for multiresolution imaging.
Results
Organ-scale tissue expansion without loss of proteins
The key to MAP is to prevent intra- and inter-protein crosslinking during the hydrogel-tissue hybridization step21 and then to dissociate and denature proteins to allow natural expansion of the hybrid (Fig. 1a). We hypothesized that a high concentration of acrylamide (AA) monomers might effectively prevent protein crosslinking by quenching reactive methylols formed by the protein-formaldehyde reaction. If the AA monomer concentration is low, the reactive methylols would react with amide groups within the same protein or adjacent proteins to form methylene bridges (Fig. 1a, left column)22. Such intra- and inter-protein crosslinking would prevent dissociation of protein complexes and limit subsequent tissue expansion23. With increased AA concentration (Fig. 1a, right column), methylols might preferentially react with excess AA monomers, effectively reducing inter-protein crosslinking while maximally tethering individual proteins to an expandable hydrogel mesh. Dissociation and denaturation of protein complexes with heat and anionic surfactant then facilitates natural expansion of the hydrogel-tissue hybrid while preserving overall spatial organization of the proteome.
Figure 1. Magnified and accessible 3D proteome library of whole intact organs.
(a) Comparison of hydrogel-tissue hybridization chemistry, subsequent protein denaturation, and hybrid expansion between low (left column) and high (right column) concentrations of acrylamide (AA; blue) infusion. High-concentration AA prevents intra- and inter-protein crosslinking during the hydrogel-tissue hybridization step by quenching reactive methylols formed from amine residues (green) reacting with formaldehyde (red). Dissociation and denaturation of non-crosslinked protein complexes allows natural expansion of the hybrid. In low-concentration AA (left column), the reactive methylols react with available amine or amide groups within the same protein or adjacent proteins and form methylene bridges (red line). Such intra- and inter-protein crosslinking prevents complete denaturation, dissociation of the proteins and thereby limits subsequent tissue expansion. (b) Bovine serum albumin (BSA) tissue phantom post-fixed in higher AA concentrations showed higher degrees of expansion after denaturation. A albumin-containing tissue phantom made of 4% BSA, 4% paraformaldehyde (PFA), and 4% AA was cut into equal size disks, washed overnight, and post-fixed with 2% glutaraldehyde (GA) or various concentrations of AA and 4% PFA. Tissue phantom were denatured at 95°C in 200 mM sodium dodecyl sulfate (SDS) solution for 1 h and subsequently incubated in deionized (DI) water. Representative images of tissue phantoms before and after expansion are shown. One-way ANOVA was performed (***P< 0.001). Error bars show SD (n = 3). (c) Expansion of 1-mm-thick coronal mouse brain slices increased with AA concentration. Mice were perfused with various amounts of AA and sodium acrylate (SA) at a constant AA/SA ratio with 4% PFA in phosphate-buffered saline (PBS). After sectioning, hydrogels were formed, followed by denaturation and DI water incubation. One-way ANOVA was performed (P < 0.001). The V0 represents the volume before expansion, and V is the volume after expansion. Error bars show mean ± SD (n = 6). (d) Average diameter of 1-mm-thick slices relative to the initial diameter after expansion and subsequent shrinkage. Slices were allowed 24 h for both expansion and shrinkage (n = 6). The L0 is the mean length of original brain, and L is the mean length after expansion and shrinkage. (e) Representative photos showing expansion and shrinkage of a 1-mm-thick coronal block. Starting from the top left, the original brain section, the expanded state, and the shrunken state are shown. The length increased about 4.2-fold after expansion and decreased to 1.3-fold the original size after shrinkage. (f,g) MAP applied to a whole mouse brain and other organs. The entire process from perfusion to full expansion took 7 days. Compared with the original organ size (pictures at right bottom corner), the final expansion of the brain showed a more than 4-fold increase in length. (h) MAP applied to cerebral organoid. Top, dark field images; bottom, bright field image. Scale bars, 10 mm.
To test our hypothesis, we first measured the effect of AA concentration on tissue expansion. We prepared albumin-containing tissue phantoms and post-fixed them in 4% para formaldehyde (PFA) with different concentrations (0–20%) of AA. We incubated the tissue phantoms in detergent solution at 95°C for 1 h to denature and disrupt protein aggregates. As expected, phantoms fixed in higher concentrations of AA showed higher degrees of expansion in water (Fig. 1b). We observed a similar trend in mouse brain tissues that were perfused with different concentrations of AA, polymerized, denatured, and expanded (Fig. 1c). Thus, we used high concentrations of AA in order to maximize expansion during MAP in all subsequent experiments.
Using this approach, we achieved a four-fold linear expansion of a whole mouse brain within seven days without protease treatment (Fig. 1f). Tissue expansion was reversible and tunable using buffers with different salt concentrations and osmolarities (Fig. 1d,e). This method is applicable to other organs including heart, lung, spinal cord, liver, intestine, and kidney and also cerebral organoids (Fig. 1g,h and Supplementary Fig. 1).
Preservation of multiscale architectures
We next asked whether MAP retains multiscale structural information and enables super-resolution imaging with diffraction-limited microscopes. To estimate the amount of distortion incurred from expansion, we imaged gel-embedded cultured cells before and after MAP processing (Fig. 2a). At the subcellular scale, MAP expansion improved visualization of microtubules and allowed imaging of single tubular structures (Fig. 2b,c). The estimated distortion error (root-mean-square error, RMSE) was less than 3% of measured length at both the subcellular scale (Fig. 2d) and the multicellular scale (Fig. 2g). The degree of gel expansion was not a function of cell density, as indicated by the consistent local expansion within differently populated cell clusters (Fig. 2e,f).
Figure 2. MAP enables isotropic expansion with minimal distortion.
(a) HeLa cells stained for alpha-tubulin and imaged before and after MAP processing. (b) Fine tubulin structures imaged before and after expansion, respectively. (c) Intensity profiles along the paths indicated in b based on the pre-expansion size. (d) Root-mean-squared measurement error (RMSE) of tubulin images before and after MAP (black line, mean; filled area, standard deviation; n = 4 samples). (e) Morphological image processing of gel-embedded HeLa cells before and after MAP processing. (f) Local pairwise expansion ratios versus the fraction of the connecting line segment inside of nuclei (black line, overall expansion ratio [3.82]; red dashed line, standard linear regression [m = 0.04 ± 0.02, b = 3.83 ± 0.01]). (g) RMSE of TOTO-1 images before and after MAP same as in d (n = 4 samples). (h) 100-μm-thick mouse brain slice showing blood vessels stained with lectin before and after MAP processing. (i) RMSE of lectin images same as in d. Errors measured after MAP processing (n = 6 samples) are compared with errors measured between two rounds of embedding and imaging of the sample before MAP (n = 5 samples). (j–l) 100-μm-thick mouse brain slices stained with anti-GFP antibodies and three other antibodies before and after MAP processing. Magnified 3D-rendered images in middle columns show the same structures imaged with the same microscopic resolution near optical sampling limit. Magnified 2D images showing parvalbumin (PV) ultrastructure in middle panels in k were taken from other samples. NF-H, neurofilament heavy unit. (m) Intensity plot (left) along the lines in zoom-in single sections (right) of the insets in j. Dimension in the plot is of before MAP. Images were obtained using a 10×, 0.3 NA dry objective (e) and a 63×, 1.30 NA glycerol-immersion objective (a and b) with single-photon (1p) excitation by 488, 568, and 594 nm, and a 20×, 0.95 NA water-immersion objective (h and j–m) at the same resolution below optical sampling limit with either 1p excitation by 488 and 594 nm or 2p excitation by 780 nm. Scale bars, 1 mm (white; e and h), 100 μm (gray; inset in e, and j–l), 10 μm (yellow; a, and insets in j–l), and 1 μm (magenta; b and m).
To test whether MAP preserves multiscale tissue architectures, we imaged a 100-μm-thick mouse brain block labeled with lectin before and after MAP (Fig. 2h). The distortion analysis showed less than 4% error (Fig. 2i), which was at most a two-fold increase, compared to the inevitable distortion from sample mounting for imaging (Fig. 2i, remounting). Both the intra-vascular space with low protein concentration and the extra-vascular space with high protein concentration showed similar degrees of expansion (Fig. 2h, insets), suggesting that the effect of protein concentration and cell density on gel expansion is minimal. Diffraction-limited microscopy revealed fine 3D details of immunolabeled cells such as cytoskeletal filament structures (Fig. 2j,k) and better-resolved axonal fibers (Fig. 2l) after the MAP process. These results together indicate that MAP preserves multiscale structural information of biological samples and enable super-resolution imaging with diffraction-limited microscopes, achieving approximately 60-nm lateral resolution.
Labeling proteome with conventional antibodies
We next asked if MAP preserves the 3D proteome composition and organization and enables super-resolution imaging of fine subcellular architectures using commercially available antibodies. Antibody targets were selected from a wide range of antigens including membrane proteins, cytoplasmic proteins, nuclear proteins, neurofilament (NF) proteins, and synaptic proteins to survey the overall proteomic landscape (Fig. 3 and Supplementary Table 1). To validate the antibodies, we tested whether antibody staining of MAP-processed tissues shows comparable structures to PFA-fixed tissues, and antibodies from different host species targeting the same protein, if available, provide concordant images (Supplementary Figs 2 and 3).We found that 100 of 122 antibodies were compatible with MAP-processed samples, and 43 of 51 target molecules were successfully visualized (Fig. 3, Supplementary Table 1, and Supplementary Figs 2 and 3). This high success rate indicates that loss of epitopes by protein denaturation was minimal. Moreover, some antibodies (e.g., calbindin [CB] and calretinin [CR]) showed negative staining when different antibodies targeting the same endogenous protein showed strong signal (Fig. 3b and Supplementary Table 1). This result demonstrates that MAP preserves the tissue proteome with minimal loss of specific epitopes and that the library can be effectively probed using off-the-shelf antibodies without any modification.
Figure 3. MAP preserves the 3D proteome library and subcellular details of an intact tissue and enables its super-resolution imaging.
(a) A total of 122 antibodies were tested on MAP-processed mouse brain sections and PFA-fixed control tissues (see Supplementary Table 1). One hundred of 122 antibodies were compatible with MAP, and 43 of 51 target molecules were successfully labeled. (b) Representative images showing morphological details of PV, calbindin (CB), and calretinin (CR) positive cortical neurons. (c,d) 3D rendering of fine cytoskeletal structures visualized by neurofilament protein staining: NF-H in a cortical neuron (c), and SMI-312, a pan-axonal marker, and tyrosine hydroxylase (TH) in a subcortical neuron (d). See Supplementary Videos 1 and 2. (e) Structural relationship between astrocytes and capillaries visualized with glial fibrillary acidic protein (GFAP), lectin, and histone H3 in a subcortical region. See Supplementary Video 3. (f) 3D rendering of a cortical sample stained with neurofilament medium unit (NF-M) and GABAB receptor subunit-1 antibodies. The inset shows synaptic neurofilaments co-localized with the gabaergic post-synaptic proteins. See Supplementary Video 4. (g) Maximum intensity projection (MIP) image showing dendritic spines visualized with a cell type-specific marker, CB. (h) Synaptic structure resolved with a pre-synaptic marker, bassoon, and a post-synaptic marker, homer1, in a cortical region. Yellow box shows a GFP-positive neuron whose dendrite was imaged. The inset (white box) highlights the elliptical structures of pre- and post-synaptic proteins distributed at a synaptic junction. (i) Intensity plot along the axis of the synapse in h (rectangular region, expanded) showing the separation between pre- and post-synaptic distribution profiles. The lines represent the Gaussian curve fitting. (j,l,m) Additional synaptic markers successfully visualizing molecular synapses in MAP-processed tissues. (j) A glutamatergic pre-synaptic marker (VGluT2) co-localized with bassoon in a cortical region (inset). (k) Comparison of the number of synapses between bassoon and VGluT2 in j (n = 3 images). A two-tailed paired t test was performed (*P < 0.05). Error bars show mean ± SEM. (l) The distribution of glutamatergic pre- (VGluT1) and post-synaptic (PSD95) markers in a cortical region. (m) 3D rendering showing the distribution of homer1 clusters in a subcortical region. See Supplementary Video 5. The images in f, j and m were obtained from samples stained in a second round. The images were obtained with a 25×, 0.95 NA water-immersion objective (b), a 40×, 1.25 NA oil-immersion objective (e), and a 63×, 1.30 NA glycerol-immersion objective (c, d, f–h, j, l and m) with 1p excitation (488, 568, 594, and 647 nm). Scale bars, 50 μm (white; b–f and low magnification images in h and l), 10 μm (gray; inset in e, and h, j, l and m), and 1 μm (yellow; insets in h, j, l and m).
Super-resolution imaging of 3D subcellular architectures
Preservation of both nanoscopic structures and the proteome in MAP enables super-resolution imaging of a broad range of fine subcellular architectures. Antibodies targeting filament proteins successfully visualized details of the cytoskeletal networks of various cell types (Fig. 3c–e and Supplementary Videos 1–4). No apparent filament fragmentations were observed in the magnified tissues. Labeling of glial fibrillary acidic protein (GFAP) visualized the fine foot process structures surrounding blood vessels (Fig. 3e and Supplementary Video 3)24,25. NF medium subunit (NF-M) visualized both processes and synapses26, a subset of which co-localized with the postsynaptic marker, GABABR1 (the C-terminal domain of the GABAB receptor subunit-1) (Fig. 3f and Supplementary Video 4). Labeling of calcium-binding proteins (parvalbumin [PV], CB, and CR) robustly visualized morphological details of positive cells (Fig. 3b). A small subset of CB-positive fibers showed expression of CB in their dendritic spines (Fig. 3g). Many synaptic protein markers were compatible with MAP (Fig. 3f,h,j,l,m and Supplementary Table 1). Staining of synaptic proteins clearly visualized distinct and well separated elliptical disk-shaped clusters of pre- and post-synaptic proteins (Fig. 3h,l,m and Supplementary Video 5), enabling their quantitative analysis (Fig. 3k).
Highly multiplexed immunolabeling
Another potential advantage of the MAP technology is that it enables highly multiplexed labeling and imaging of the magnified 3D proteome within a single tissue. We successfully performed seven rounds of immunostaining of a MAP-processed 100-μm-thick mouse brain tissue with no obvious signs of tissue damage (Fig. 4). Outstanding mechanical stability of the MAP-processed sample enabled repeated manual handling of the tissue with tools that are widely used in biology labs. The images after destaining of the imaged anti-GFP antibody (first round) and after solely adding secondary antibodies targeting the eluted anti-GFP antibody (second round) showed little signal, suggesting that the destaining process effectively eliminates antibodies. For the following five rounds of labeling, we continued to use anti-GFP as a landmark in addition to two other antibodies for each round. The consistent GFP signals show that the same antigen can be repeatedly labeled without loss of antigenicity. Successful visualization of ten other targets demonstrates that MAP facilitates exploring diverse proteins, structures, and cell types within a single tissue.
Figure 4. Multiplexed staining of MAP-processed tissue.
Repeated staining and destaining of a MAP-processed tissue. A 100-μm-thick mouse brain slice including an anterior hippocampal region was used. The tissue in a PBS solution with 0.1% Triton X-100 was imaged. 64-μm-thick MIP images are displayed. Host species of primary antibodies are shown in the parentheses. Break-lines indicate destaining. GFP was stained and imaged at 635-nm excitation channel to show consistent labeling of the same antigen. 488- and 559-nm channels were used to show multiplexed labeling of diverse antigens and complete elution of imaged antibodies used in previous rounds. The same imaging and post-processing conditions were used for 635-nm channel images throughout the first and second rounds. MBP, myelin basic protein; TUJ1, neuron-specific class III beta-tubulin. Scale bars, 200 μm.
MAPing neural projections
We next explored the potential utility of MAP for mapping inter-areal neural connectivity. If MAP could preserve continuity of neural fibers within a magnified intact brain tissue along with its 3D proteome, highly specific antibodies would enable visualization of a subset of neural projections connecting different brain regions. The sparse labeling and enhanced spatial resolution offered by MAP may allow more accurate reconstruction of the labeled projections. In addition, highly multiplexed proteomic imaging may enable reconstruction of many different cell types with integrated molecular and fine morphological details in a single tissue.
To explore this possibility, we first asked if the continuity of neural processes is preserved within a magnified sample. We expanded a 0.5-mm-thick Thy1-eGFP mouse brain coronal block and then imaged GFP-labeled neurons and their projections27. As shown in Figure 5a–f, MAP preserves the continuity of the GFP-expressing projections spanning a large tissue volume as well as their fine morphological details (e.g., dendritic spines) (Fig. 5g). Immunolabeling of various cytoplasmic proteins including NF proteins (e.g., NF-H and SMI-32), calcium-binding proteins, and metabolic enzymes (e.g., tyrosine hydroxylase [TH]) confirmed that MAP preserves the continuity of neural fibers of various cell types (Fig. 5i–n and Supplementary Videos 6–8).
Figure 5. MAP preserves inter-cellular connectivity and enables its reconstruction at single fiber resolution.
(a–g) Connectivity and morphology of cells are preserved in the MAP-processed tissue. (a) A 500-μm-thick slice of a mouse brain hemisphere stained with anti-GFP antibody and imaged after expansion and shrinkage. (b) The entire neuronal territory and axon fiber of a cortical pyramidal neuron included within the yellow-boxed region in the left panel. The volume was 3D reconstructed with background removal. (c–e) Images of the same neuron in b obtained with a 10×, 0.3 NA objective (c and d) and a 10×, 0.6 NA objective (e), showing its morphological details before MAP, after MAP expansion, and after MAP shrinkage. (f,g) Fine subcellular structures preserved and resolved in the expansion states: dendritic network (f) and dendritic spines (g). (h–l) Neuronal fibers and cell body (asterisks) morphology visualized using various markers (CR, SMI-32, NF-H, SMI-312, and TH) in different cortical (i–k) and subcortical (l) regions of a 1-mm-thick hemisphere slice (h). See Supplementary Video 6. The inset in i shows the isolated neuronal cell body and its axon with background removal. (m) Long-range tracing (white line, longer than 3 mm) of an inter-regional TH neurofilament within the magenta box in h. See Supplementary Video 7. (n) Subvolume within a square region in h (green box) containing dense SMI-312–positive fibers. The red line shows a tracing result across the volume. See Supplementary Video 8. (o) One of the densest subregions in the left inset in n visualizing all individual fibers resolved and traced. (p) A dense region in the right inset in n showing closely crossing fibers. Depth-color coding of the fibers clearly shows good separation along the z-axis. (q) The intensities of the two fibers closely located in a similar plane in p are plotted and fitted to Gaussian curves to show their clear separation. (r,s) Comparison of manual tracing results of a 100-μm-thick hippocampal sample stained with SMI-312 before (r) and after (s) MAP processing. Apparently detectable fibers were traced by two individuals. Discordant traces are selectively displayed for before MAP image (r), and the same traces, which resulted in concordant tracing after expansion, are shown for after MAP image (s). (t,u) Comparison of discordance ratios between manual tracing results before and after MAP processing. Two markers (SMI-312 and NF-H) were used with four regions of interest for each marker. Two-tailed paired t tests were performed (**P < 0.01, *P < 0.05). Error bars show mean ± SEM. Images were obtained using a 10×, 0.6 NA CLARITY-optimized objective(a–c and e), a 10×, 0.3 NA water-immersion objective(d), a 20×, 0.95 NA water-immersion objective (m), a 63×, 1.30 NA glycerol-immersion objective(l and n–p), and a 25×, 0.95 NA water-immersion objective (f, g, i–k, r and s) with 1p excitation by 488, 568, and 594 nm. Scale bars, 1 mm (gray; a), 200 μm (white; b–f, i–k, m, n, r and s), and 10 μm (yellow; g, l, o and p).
To test if MAP enables more accurate tracing of densely packed fibers, we imaged NF-stained samples before and after MAP processing using a high NA (0.95) water-immersion objective. Two individuals not involved in image acquisition traced fibers within the sample volume (Fig. 5r,s). When we compared the concordance between the two tracing results, the discordant rate was significantly lower after MAP (Fig. 5t,u) even though MAP enabled the tracers to detect more fibers (total numbers of traced fibers were 160 before MAP and 214 after MAP). This result demonstrates that MAP indeed enables more accurate reconstruction of immunolabeled neural fibers.
Immunolabeling and imaging of mm-thick tissues
Reconstruction of individual neurons requires labeling and imaging of thick brain tissues because nerve fibers can extend across a large volume. To test whether MAP is applicable to large-scale brain tissues, we expanded a 1-mm-thick mouse brain block (5-mm-thick after expansion, Fig. 6b) and passively stained it with anti-GFP antibody (Fig. 6a). We then imaged the sample using both a high NA (0.95), short working distance (WD) (2 mm) water-immersion objective and a low NA (0.6), long WD (8 mm) CLARITY objective. Both objectives showed fine dendritic spines of GFP-expressing neurons throughout the entire volume (Fig. 6e). Imaging of the sample did not require depth-dependent modulation of the laser power, indicating that staining was uniform and that signal attenuation by light scattering was negligible (Fig. 6c,d). MAP is also compatible with stochastic electrotransport28, a method that enables rapid tissue labeling. Using stochastic electrotransport, we were able to label an 8-mm-thick expanded tissue uniformly within only two days (Fig. 6f–j). The sample was highly transparent. We were able to image the 8-mm-thick sample successfully with the same laser power up to the WD of the objective (Fig. 6h).
Figure 6. Immunolabeling and imaging of thick MAP-processed tissues.
(a–e) Passive staining of a 1-mm-thick (5 mm after expansion) dorsolateral mouse brain slice with anti-GFP antibodies. (a) The tissue was incubated for 3 days with primary antibody and 3 days with secondary antibody, and expanded in DI water. (f–j) stochastic electrotransport28 staining of a 2-mm-thick (8 mm after expansion) mouse hemisphere slice with anti-GFP antibodies. (f) The tissue was stained for 1 day with both primary and secondary antibodies. (b,g) Lateral (top panels) and top (bottom panels) views of the sample. White boxes indicate the regions of interest for z-stack images. The cortical (b) and hippocampal (g) regions were chosen at least 2.5- and 4-mm far from the lateral tissue boundary, respectively, to exclude the chance of dominant staining by lateral diffusion at a middle section. (c,h) Laser power used along depth to obtain a z-stack image. (d,i) Longitudinal tile-scan images along z-axis with 1p excitation and a 10×, 0.6 NA, 8-mm WD objective. (e,j) Sectional images at different depths either with a 10×, 0.6 NA, 8-mm WD (magenta) or a 20×, 0.95 NA, 2-mm WD (blue) objectives. The same image acquisition parameters were used for sectional images with each objective. Insets in e are zoom-in images showing dendritic spines. Scale bars, 5 mm (gray; b and g), 500 μm (white; d and i), 100 μm (yellow; e and j), and 10 μm (blue; insets in e).
Discussion
As a step toward organ-scale reconstruction of diverse cell types and their surrounding environment and to advance the study of complex system-level interactions, we developed a simple and scalable method that enables preservation, reversible expansion, and imaging of the 3D proteome organization within an intact tissue. We discovered that the protein content of a whole organ can be preserved and magnified by preventing intra- and inter-protein crosslinking, denaturing, and dissociating the protein complexes to allow natural expansion of hydrogel-tissue hybrids. The expanded hybrids secure both fine subcellular architectures and organ-scale cellular connectivity. These multiscale properties can be directly imaged by using off-the-shelf antibodies to label the structures' constituent proteins. The high success rate that we achieved with commercial antibodies might be because synthetic peptides or denatured protein fragments are commonly used for conventional antibody production29. An existing large antibody library, once validated, can be used without any modification. This technique is easy to implement, and it does not require any special equipment or chemicals.
MAP enables repeated interrogation of the same expanded specimen. This is possible because the preserved epitopes, which have already reached complete denaturation by harsh treatment (95°C, 200 mM sodium dodecyl sulfate [SDS]) for expansion, do not undergo any substantial modification during the milder antibody elution step (70°C, 200 mM SDS). We demonstrated seven rounds of staining of the same tissue without any signs of tissue damage. However, the practical limit of this approach must be carefully examined.
MAP has the potential to enable scalable super-resolution imaging of large-scale samples. Unlike other methods, thick tissues can be labeled on a practical time scale because antibody-labeling occurs after complete lipid removal and tissue permeabilization. The great transparency of the MAP-processed samples allows high-resolution imaging of the physically expanded tissue with minimal loss of resolution. Currently, 1.0 NA, 2.5-mm WD water-immersion objectives are best suited for MAP. Although the 1.0 NA, 8-mm WD CLARITY objective has the longest WD, it does not provide high quality images because it is optimized for immersion media with a high refractive index (RI). Further development of high NA, long WD water objectives or the development of high RI immersion media that are compatible with MAP would extend the utility of MAP. Another challenge in the MAP approach is the dilution of fluorescent signals that accompanies physical volumetric expansion. Four-fold linear expansion decreases signal density by 64-fold. Therefore, significantly higher laser power is required, which in turn causes photobleaching. Future studies will need to explore the compatibility of signal amplification techniques with MAP.
The reversible modulation of physical sample size that is attained with MAP enables multi-scale proteomic imaging of a single tissue to capture both system-scale properties and fine local details in a practical way. For example, following sample shrinkage, global projection patterns of labeled neurons may be imaged using high-speed microscopy techniques and long WD objectives (e.g., 25×, 1.0 NA, 8-mm WD objective, currently available)30. After imaging, the same tissue can be expanded for super-resolution imaging of regions of interest. The expanded tissue may need to be sliced prior to imaging to meet the objective's WD limit. A microscope with a built-in vibrating-blade microtome could also be used for whole-mount imaging to avoid loss of connectivity information31–33. This approach may allow us to drastically reduce the imaging time and costs associated with data storage and handling of expanded samples.
The advantages of MAP (Supplementary Table 2) may allow combined extraction of rich molecular details, minute subcellular architectures, and cellular connectivity from diverse cell types within a single tissue. Together with its simplicity and broad applicability, MAP may complement existing methods and enable new approaches in the study of complex biological systems.
Online Methods
Bovine Serum Albumin (BSA) hydrogel denaturation and expansion
Stock solutions of 40% BSA, 40% AA, 32% para formaldehyde (PFA), and 1% VA-044 were made and kept on ice throughout the experiment. A 10-mL solution of 4% BSA, 4% AA, 4% PFA, 0.1% bis-acrylamide (BA), and 0.1% VA-044 was made in phosphate-buffered saline (PBS). The solution was polymerized under vacuum at 37°C for 2h, and the resulting albumin-containing tissue phantom was sectioned. Individual tissue phantom sections were washed in excess PBS with shaking for 12 h. After washing, 4 phantoms were placed in 10 mL of each of the following PBS solutions for 4 h at 37°C: 2% glutaraldehyde (GA), 4% PFA, 4% AA with 4% PFA, and 20% AA with 4% PFA. Tissue phantoms were washed similarly, massed, photographed, and then incubated in a 200-mM SDS solution with 50mM sodium sulfite for 1 h at 95°C.Phantoms were washed again and incubated in deionized (DI) water for 12 h. After expansion, phantoms were massed and photographed.
General MAP protocol
i. Perfusion and hydrogel embedding
Thy1-eGFP-M mice (6–8 weeks old, male or female) were housed in a reverse 12-h light/dark cycle with unrestricted access to food and water. All experimental protocols were approved by the MIT Institutional Animal Care and Use Committee and the Division of Comparative Medicine and were in accordance with guidelines from the National Institutes of Health. After anesthesia, the mice were first washed transcardially with a mixture of 2–5% AA, 0–0.05% BA, 0–0.8% sodium acrylate (SA), and PBS, followed by perfusion with a mixture of 4% PFA, 30% AA, 0.05–0.1% BA, 10% SA, 0.1% VA-044 or V-50, and PBS. The perfusion solution could be slightly turbid. Upper transparent solution was used after centrifugation during 3 min with 1,000 G. All solution was protected from light and kept on ice before perfusion. Control samples were perfused first with PBS and then with 4% PFA and PBS. The brain and other organs (heart, lung, liver, intestine, kidney, and spinal cord) were harvested and incubated in 20–40 mL of the same fixative solution at 4°C for 2–3 days and then for 1–3 days at room temperature (RT) with gentle shaking to ensure uniform chemical diffusion and reaction throughout the sample. Following the diffusion and fixation steps, hydrogel-tissue hybridization was performed in situ by incubating the tissues using Easy-Gel (LifeCanvas Technologies, Seoul, South Korea) with nitrogen gas at 45°C for 2 h.
ii. Tissue denaturation
Hydrogel-embedded tissues were incubated overnight in a solution of 200 mM SDS, 200 mM NaCl, and 50 mM Tris in DI water (pH titrated to 9.0) at 37°C with gentle shaking. The samples were then incubated at 70°C for 0–50 h and 95°C for 1–24 h depending on their size. Whole organs were incubated for 24–48 h at 70°C followed by 12–24 h at 95°C. 1-mm-thick brain slices were incubated for 5 h at 70°C and then at 95°C for 1 h.
iii. Expansion
Denatured tissues were incubated in 40–100 mL DI water at RT for 12–48 h with gentle shaking. During DI water incubation, the solution was changed every 3–5 h.
Expansion according to various AA and SA concentrations
After anesthesia, mice were washed transcardially with 4% AA, 0.05% BA, 0.8% SA, 0.1% VA-044, and PBS and then with 4% PFA, 0.05% BA, 0.1% VA-044, one of four AA and SA combinations (5% AA + 0.8% SA, 10% AA + 1.7 % SA, 20% AA + 3.3% SA, and 30% AA + 5% SA) in PBS. Tissues were incubated in 20 mL of the same fixative solution at 4°C for 2 days and 5 h at RT with gentle shaking. After hydrogel-tissue hybridization the samples were incubated in denaturation solution at 70°C for 5 h and 95°C for 1 h with gentle shaking. Denatured tissues were incubated in 40 mL DI water at RT for 12 h with gentle shaking. During DI water incubation, the solution was changed every 3–5 h. Fiji (National Institutes of Health) was used to measure the size of the expanded samples34.
Shrinkage and RI matching
A customized RI matching solution was made by dissolving 50 g diatrizoic acid, 40 g N-methyl-D-glucamine, and 55g iodixanol per 100 mL PBS28. The RI was targeted to 1.47. This solution was used for both shrinkage and imaging. Depending on sample size, samples were incubated in 1–10 mL solution at RT for 2–24 h with gentle shaking. The solution was changed every 1–12 h. For imaging tissues before MAP processing, samples were incubated in 1–10 mL of this solution without PBS at RT with gentle shaking for 2–5 h prior to imaging.
MAP processing of cerebral organoid
Cerebral organoids were made from stem cells following a previously described protocol35. Organoids were initially fixed in 4% PFA for 15 min, incubated in mixture of 4% PFA, 30% AA, 0.1% BA, 10% SA, 0.1% V-50, and PBS for 24 h at 4°C followed by 24 h, at RT. Hydrogel embedding, tissue denaturation and expansion were processed similarly to “General MAP protocol”.
Cultured cell experiment
For tubulin imaging in HeLa cells, 8-mm round glass coverslips were coated in 0.1% gelatin in ultrapure water (Millipore). Coverslips were placed in a 48-well plate and seeded with 50,000 HeLa cells overnight. To obtain comparable images before and after MAP processing, cells were washed, fixed with 3% PFA + 0.1% GA in PBS for 10 min, and switched to a solution of 4% PFA, 30% AA in PBS for 8 h at 37°C. Cells were then placed in 0.1% sodium borohydride for 7 min at RT then incubated in 100 mM glycine for 10 min at RT. Cells were washed and stained with anti-tubulin (Abcam, ab6160), Alexa Fluor 594 conjugated secondary (Abcam, ab150152) antibodies and TOTO-1 (Thermo Fisher Scientific). Cells were mounted in 2,2′-thiodiethanol (Sigma) and imaged with a 63×, 1.3 NA glycerol-immersion objective with the Leica microscope system. Cells were washed extensively and embedded into a MAP hybrid polymer by addition of 20 μL of Cell-MAP solution (20% AA, 7% SA, 0.1% BA, 0.5% TEMED, 0.5% ammonium persulfate in PBS). Ammonium persulfate was added last from a freshly prepared 5% stock solution. Cell-MAP solution was quickly added to the coverslip and left to polymerize for 4–5 min. Gels were peeled off the coverslip using forceps, washed extensively and denatured for 30 min in denaturation buffer at 95°C. Cell-MAP gels were washed extensively, restained with anti-tubulin antibody and TOTO-1 and reimaged.
Immunostaining of brain tissue
For typical staining, MAP-processed 100–500-μm-thick mouse brain coronal slices were incubated with primary antibodies (typical dilution, 1:100) in PBS with 1% (wt/vol) Triton X-100 (PBST) at 37°C for 8–16 h, followed by washing at 37°C for 1–2 h in PBST three times. The tissue was then incubated with secondary antibodies (typical dilution, 1:100) in PBST at 37°C for 6–16h,followed by washing at 37°C for 1–2 h in PBST three times. For antibody validation of a given antibody, 100-μm-thick PFA-fixed control and MAP-processed samples were stained with the same titer of primary and, if necessary, secondary antibodies overnight in PBST. See Supplementary Table 1 for the list of antibodies used. To destain for multiplexed labeling, samples were incubated in a denaturation solution 6–16 h at 70°C, and washed with PBST at 37°C for 1–2 h three times.
Mounting and imaging
Samples were mounted on a slide glass. Blu-Tack adhesive was applied on the petri dish or the slide glass, and samples were covered with a glass-bottom Willco dish. The space between the bottom material and the Willco dish around the sample was filled with either shrinkage solution or DI water according to the sample immersion medium. Large expanded samples were additionally placed on a 120-mm-diameter petri dish, and the dish was filled with DI water. Expanded or shrunk samples were stabilized for at least 1 h before imaging. Samples were imaged with either the Olympus FV1200MPE microscope system or the Leica TCS SP8 microscope system. A 10×, 0.6 NA CLARITY-optimized objective (XLPLN10XSVMP; 8.0-mm WD) was used with the Olympus system to obtain wide-field images of shrunk samples and z-stack images of large samples. The images of MAP-processed samples were obtained with a 10×, 0.3 NA water-immersion objective, a 20×, 0.95 NA water-immersion objective, and a 40×, 1.25 NA oil-immersion objective with the Olympus system, or a 10×, 0.3 NA water-immersion objective, a 25×, 0.95 NA water-immersion objective, and 63×, 1.30 NA glycerol-immersion objective with the Leica system. Single-photon confocal laser scanning imaging was performed with 405-, 488-, 559-, and 635-nm lasers (Olympus) or a white-light laser (Leica). Mai Tai DeepSee (Spectra-Physics) was used for multi-photon excitation with 780 nm wavelength. The images were visualized and analyzed with Fiji or Imaris (Bitplane).
Large tissue staining
1-and 2-mm-thick mouse brain coronal slices were prepared by “General MAP protocol”, and expanded. A 1-mm-thick slice was chopped to about 3 mm × 3 mm × 1 mm (dimensions before MAP). The sample was stained passively with Alexa Fluor 594-conjugated rabbit anti-GFP antibody (Life Technologies, A21312) for 3 days and Alexa Fluor 594-conjugated donkey anti-rabbit IgG antibody (Abcam, ab150072) for 3 days. 20 μL of antibody was used in 500 μL PBST, and was washed with 40 mL PBST for 1.5 days (three times in total) for each antibody. We used stochastic electrotransport28 to stain a 2-mm-thick slice with Alexa Fluor 647-conjugated rabbit anti-GFP antibody (Life Technologies, A31852) and Alexa Fluor 647-conjugated donkey anti-rabbit IgG antibody (Abcam, ab181347). For each antibody, we first stochastic electrotransported a solution containing 20 μL of antibody in 4 mL of 0.6 M N-cyclohexyl-3-aminopropanesulfonic acid (CAPS), 0.2 M Tris, 100 mM NaCl, 20 mM SDS, 1% BSA with an electrophoresis buffer containing 0.3 M CAPS, 0.2 M Tris, 20 mM SDS, 30% sorbitol for 21 h. It was then stochastic electrotransported in a solution containing 4 mL of 0.3 M CAPS, 0.2 M Tris, 20 mM SDS, 30% sorbitol, 1% BSA, and 1% Triton X-100 with an electrophoresis buffer containing 0.04 M Tris, 0.01 M phosphate, 30% sorbitol for 8 h. These two steps complete the stochastic electrotransport labeling. The stained sample was expanded in a solution containing 0.01% (wt/vol) heparin sodium procine mucosa (Sigma, SRE0027) and 1% Triton X-100 in DI water, and then imaged.
Cell distortion analysis
Before and after MAP processing tubulin images were first registered using a scaled-rotation transformation in Fiji. Non-rigid invertible B-spline registration was performed with an 8 × 8 control point grid in bUnwarpJ. Vectors of different length were subjected to the resulting non-linear transformation, and the input-output difference norm was sorted based on the input vector length and then averaged by root-mean-square. For morphological image processing of nuclei, TOTO-1 images before and after MAP-processing were first registered using a scaled-rotation transformation in Fiji. The registered images were segmented by thresholding and converted to circular particles with equivalent average radii calculated from the “Analyze Particles” function in Fiji. Matched pairs of cells from the before and after images were randomly chosen, and the expansion ratio was calculated from the ratio of the connecting line segment lengths. The fraction in nuclei was obtained by summing the intensity profile along the connecting line segment and averaging the sums from the before and after binary masks.
Tissue distortion analysis
After anesthesia, mice were first washed transcardially with 2% AA in PBS followed by perfusion with 4% PFA and 30% AA in PBS. Thy1-eGFP-M mouse brains were harvested and incubated in 20 mL of the same fixative solution at 4°C overnight and at 37°C for 3 h. Brains were sectioned to 100-μm-thick coronal slices with a vibrating microtome. Slices were stained and imaged to obtain “before MAP” images, and then incubated in a solution containing 4% PFA, 30% AA, 0.1% BA, 10% SA, and 0.1% V-50 in PBS at RT for 8 h. Hydrogel-tissue hybridization was performed in situ by incubating the tissues with nitrogen gas at 45°C for 2 h. Hydrogel-embedded tissues were incubated in denaturation solution at 37°C for 1 h and 95°C for 0.5–1 h. Samples were then stained with the same markers and imaged to obtain “after MAP” images.
To quantify distortion errors, regions of approximately 3 mm × 2.5 mm in size that included cortex and hippocampus were stained with DyLight 594-conjugated lectin (Vector Laboratories, DL-1177) before (8 μL in 200 μL PBST for up to 8 samples) and after (2 μL in 200 μL PBST for each sample) MAP processing. Samples were incubated in RI matching solution after staining and mounted and imaged before MAP processing. Five samples were repeated for incubation, mounting and imaging to measure mounting errors. After MAP processing, six samples were stained, expanded in DI water, and imaged. Keypoints of the vasculature in volumetric images were detected and matched between two image sets with a MATLAB code implementing the 3D Harris Corner Detector and 3D SIFT algorithm as described previously13. Using custom-built graphical user interface software developed with Delphi XE4 (Embarcadero Technologies), redundant keypoints closely located to each other and keypoints at tissue margins were removed. Tissue sizes were estimated by the area defined by a convex hull encompassing all keypoints, and the expansion ratio was calculated as the ratio between two squared roots of the areas. The correspondence information was used to generate a regularly spaced deformation mesh using a 3D thin plate spline code written by Yang, Foong, and Ong (http://www.mathworks.com/matlabcentral/fileexchange/37576-3d-thin-plate-spline-warping-function). Lengths between each pair of grid points were calculated in both pre-MAP and post-MAP images, considering the expansion ratio. The difference between the two lengths was measured as a distortion error. After averaging the squared errors for each measurement length, the square root of the averages was collected from the samples to obtain statistical values of error.
Neurofilament tracing
Tracing of individual neurons was performed using either Fiji or Imaris. For manual tracing of a long TH fiber, a representative fiber was chosen during confocal imaging acquisition and traced by moving the motorized stage and adjusting the z-level. During the tracing, ambiguous crossovers were resolved by obtaining high-magnification subvolume images. After tile-scanning, the target fiber was re-identified from the image volume using Fiji and marked to be displayed in a two-dimensional plane. For the semi-automatic tracing of SMI-312 fibers in a dense region, the entire 800 μm × 800 μm × 150 μm (expanded) dataset was loaded along with a filament tool into a ‘Surpass’ instance using Imaris. An auto path calculation was performed using a single starting point as indicated in Supplementary Video 8. The fiber endpoint was designated by selecting the portion of the fiber exiting the imaged volume. The fiber representation was changed to a cone representation to visualize the filament diameter as well as the tracing path. The tracing fidelity was confirmed by inspection. To trace multiple fibers in a dense region, the full dataset was cropped to the region indicated in Figure 5o using the ‘3D crop’ tool, and similar auto path calculations were performed for each fiber. For traceability comparison between before and after MAP, same tissues before and after MAP processing were imaged at the same resolution near the optical sampling limit with the Leica system and a 25×, 0.95 NA water-immersion objective. Two individuals not involved in imaging acquisition performed manual tracing using auto depth assistance in Imaris, and discordance ratios between manual tracing results before and after MAP processing were calculated.
Synaptic and fiber intensity profiles
The regions of interest were imported into Fiji. Lines were drawn perpendicular to the synaptic junction or near the fibers of interest and intensity profiles were obtained with adjusting the line width. For synaptic intensity profiles, two Gaussian distributions were fit to the distinct peaks by simultaneous minimization of the sum of squared residuals. Synaptic densities for bassoon and VGluT2 were calculated from three non-overlapping xy-images (235 μm × 235 μm) of expanded samples. The images were segmented by thresholding, and individual synaptic structures for each channel were counted using the “Analyze Particles” function in Fiji. The synaptic densities were calculated based on the frame area, and the standard deviation was calculated from the 3 replicates.
Supplementary Material
Acknowledgments
The authors thank the entire Chung laboratory for support and helpful discussions. K.C. was supported by Burroughs Wellcome Fund Career Awards at the Scientific Interface, the Searle Scholars Program, Packard award in Science and Engineering, JPB Foundation (PIIF and PNDRF) and NIH (1-U01-NS090473-01). Resources that may help enable general users to establish the methodology are freely available online (http://www.chunglabresources.org).
Footnotes
Author Contributions: T.K., J.S, J.-Y.P., and K.C. designed the experiments and wrote the paper with input from other authors. T.K. stained and imaged mouse samples. J.S. performed the gel and cell experiments. T.K. and J.S. analyzed the data. J.-Y.P. prepared mouse tissues. J.-Y.P. and V.M. processed mouse MAP samples. A.A. performed the cell and organoid experiments. E.M., Y.-G.P., and T.K. performed the antibody validation test. J.H.C. performed stochastic electrotransport staining. Y.-G.P. and T.K. obtained synaptic images. J.-Y.P., V.M., T.K., and J.S. performed tracing. J.C. performed the gel experiment. K.C. supervised all aspects of the work.
Competing Financial Interests: The authors declare competing financial interests: K.C., T.K., J.S. and J.-Y.P. are co-inventors on patent application owned by MIT covering the MAP technology (U.S. Provisional Patent Application 62/330,018).
Note: Any Supplementary Information and Source Data files are available in the online version of the paper.
References
- 1.White JG, Southgate E, Thomson JN, Brenner S. The structure of the nervous system of the nematode Caenorhabditis elegans. Philos Trans R Soc Lond B. 1986;314:1–340. doi: 10.1098/rstb.1986.0056. [DOI] [PubMed] [Google Scholar]
- 2.DeFelipe J. From the connectome to the synaptome: an epic love story. Science. 2010;330:1198–1201. doi: 10.1126/science.1193378. [DOI] [PubMed] [Google Scholar]
- 3.Helmstaedter M. Cellular-resolution connectomics: challenges of dense neural circuit reconstruction. Nat Methods. 2013;10:501–507. doi: 10.1038/nmeth.2476. [DOI] [PubMed] [Google Scholar]
- 4.Craddock RC, et al. Imaging human connectomes at the macroscale. Nat Methods. 2013;10:524–539. doi: 10.1038/nmeth.2482. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Oh SW, et al. A mesoscale connectome of the mouse brain. Nature. 2014;508:207–214. doi: 10.1038/nature13186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Phizicky E, Bastiaens PI, Zhu H, Snyder M, Fields S. Protein analysis on a proteomic scale. Nature. 2003;422:208–215. doi: 10.1038/nature01512. [DOI] [PubMed] [Google Scholar]
- 7.Hanash S. HUPO initiatives relevant to clinical proteomics. Mol Cell Proteomics. 2004;3:298–301. doi: 10.1074/mcp.R400004-MCP200. [DOI] [PubMed] [Google Scholar]
- 8.Agaton C, Uhlén M, Hober S. Genome-based proteomics. Electrophoresis. 2004;25:1280–1288. doi: 10.1002/elps.200405846. [DOI] [PubMed] [Google Scholar]
- 9.Berglund L, et al. A genecentric Human Protein Atlas for expression profiles based on antibodies. Mol Cell Proteomics. 2008;7:2019–2027. doi: 10.1074/mcp.R800013-MCP200. [DOI] [PubMed] [Google Scholar]
- 10.Uhlen M, et al. Towards a knowledge-based human protein atlas. Nat Biotechnol. 2010;28:1248–1250. doi: 10.1038/nbt1210-1248. [DOI] [PubMed] [Google Scholar]
- 11.Micheva KD, Smith SJ. Array tomography: a new tool for imaging the molecular architecture and ultrastructure of neural circuits. Neuron. 2007;55:25–36. doi: 10.1016/j.neuron.2007.06.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kay KR, et al. Studying synapses in human brain with array tomography and electron microscopy. Nat Protoc. 2013;8:1366–1380. doi: 10.1038/nprot.2013.078. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Murray E, et al. Simple, Scalable Proteomic Imaging for High-Dimensional Profiling of Intact Systems. Cell. 2015;163:1500–1514. doi: 10.1016/j.cell.2015.11.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Dodt HU, et al. Ultramicroscopy: three-dimensional visualization of neuronal networks in the whole mouse brain. Nat Methods. 2007;4:331–336. doi: 10.1038/nmeth1036. [DOI] [PubMed] [Google Scholar]
- 15.Richardson DS, Lichtman JW. Clarifying Tissue Clearing. Cell. 2015;162:246–257. doi: 10.1016/j.cell.2015.06.067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rust MJ, Bates M, Zhuang X. Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM) Nat Methods. 2006;3:793–795. doi: 10.1038/nmeth929. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Betzig E, et al. Imaging intracellular fluorescent proteins at nanometer resolution. Science. 2006;313:1642–1645. doi: 10.1126/science.1127344. [DOI] [PubMed] [Google Scholar]
- 18.Hess ST, Girirajan TP, Mason MD. Ultra-high resolution imaging by fluorescence photoactivation localization microscopy. Biophys J. 2006;91:4258–4272. doi: 10.1529/biophysj.106.091116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Maglione M, Sigrist SJ. Seeing the forest tree by tree: super-resolution light microscopy meets the neurosciences. Nat Neurosci. 2013;16:790–797. doi: 10.1038/nn.3403. [DOI] [PubMed] [Google Scholar]
- 20.Chen F, Tillberg PW, Boyden ES. Expansion microscopy. Science. 2015;347:543–548. doi: 10.1126/science.1260088. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Chung K, Deisseroth K. CLARITY for mapping the nervous system. Nat Methods. 2013;10:508–513. doi: 10.1038/nmeth.2481. [DOI] [PubMed] [Google Scholar]
- 22.Puchtler H, Meloan SN. On the chemistry of formaldehyde fixation and its effects on immunohistochemical reactions. Histochemistry. 1985;82:201–204. doi: 10.1007/BF00501395. [DOI] [PubMed] [Google Scholar]
- 23.Sung HW, Hsu HL, Shih CC, Lin DS. Cross-linking characteristics of biological tissues fixed with monofunctional or multifunctional epoxy compounds. Biomaterials. 1996;17:1405–1410. doi: 10.1016/0142-9612(96)87282-6. [DOI] [PubMed] [Google Scholar]
- 24.Kosaka T, Hama K. Three-dimensional structure of astrocytes in the rat dentate gyrus. J Comp Neurol. 1986;249:242–260. doi: 10.1002/cne.902490209. [DOI] [PubMed] [Google Scholar]
- 25.Khakh BS, Sofroniew MV. Diversity of astrocyte functions and phenotypes in neural circuits. Nat Neurosci. 2015;18:942–952. doi: 10.1038/nn.4043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Yuan A, et al. Neurofilament subunits are integral components of synapses and modulate neurotransmission and behavior in vivo. Mol Psychiatry. 2015;20:986–994. doi: 10.1038/mp.2015.45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Feng G, et al. Imaging neuronal subsets in transgenic mice expressing multiple spectral variants of GFP. Neuron. 2000;28:41–51. doi: 10.1016/s0896-6273(00)00084-2. [DOI] [PubMed] [Google Scholar]
- 28.Kim S, et al. Stochastic electrotransport selectively enhances the transport of highly electromobile molecules. Proc Natl Acad Sci USA. 2015;112:E6274–E6283. doi: 10.1073/pnas.1510133112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Uhlen M, Ponten F. Antibody-based proteomics for human tissue profiling. Mol Cell Proteomics. 2005;4:384–393. doi: 10.1074/mcp.R500009-MCP200. [DOI] [PubMed] [Google Scholar]
- 30.Tomer R, Ye L, Hsueh B, Deisseroth K. Advanced CLARITY for rapid and high-resolution imaging of intact tissues. Nat Protoc. 2014;9:1682–1697. doi: 10.1038/nprot.2014.123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Ragan T, et al. Serial two-photon tomography for automated ex vivo mouse brain imaging. Nat Methods. 2012;9:255–258. doi: 10.1038/nmeth.1854. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Economo MN, et al. A platform for brain-wide imaging and reconstruction of individual neurons. eLife. 2016;5:e10566. doi: 10.7554/eLife.10566. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Albanese A, Chung K. Neuroimaging: Whole-brain imaging reaches new heights (and lengths) eLife. 2016;5:e13367. doi: 10.7554/eLife.13367. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Schindelin J, et al. Fiji: an open-source platform for biological-image analysis. Nat Methods. 2012;9:676–682. doi: 10.1038/nmeth.2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Lancaster MA, Knoblich JA. Generation of cerebral organoids from human pluripotent stem cells. Nat Protoc. 2014;9:2329–2340. doi: 10.1038/nprot.2014.158. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.