Visualizing the native cellular organization by coupling cryofixation with expansion microscopy (Cryo-ExM)

Cryofixation has proven to be the gold standard for efficient preservation of native cell ultrastructure compared to chemical fixation, but this approach is not widely used in fluorescence microscopy owing to implementation challenges. Here, we develop Cryo-ExM, a method that preserves native cellular organization by coupling cryofixation with expansion microscopy. This method bypasses artifacts associated with chemical fixation and its simplicity will contribute to its widespread use in super-resolution microscopy.

MT (inset 4) (Fig. 2a). Moreover, we found that cryo-ExM can retain and resolve the nanoscale-sized holes in ER sheets as observed previously in live-STED imaging 17 (Fig. 2a,b and Extended Data Figs. 3 and 4a, inset). Besides microtubules, ER has been also involved in the regulation of mitochondrial function and ER-mitochondria contact sites have been extensively studied notably by live SRM 15 . Hence, using NHS-ester staining 18 , a compound that reacts with the primary amines of proteins, we performed a global proteome labeling, consistently revealing the position of mitochondria in the global cellular context 18,19 . Combining NHS-ester to ER labeling, we could also observe that cryo-ExM can capture ER wrapping as well as entanglement of the mitochondria (Fig. 2b,c and Extended Data Fig. 4). Finally, the use of MitoTracker allowed us to demonstrate that both cryo and PFA/GA fixations preserve better the ultrastructure of mitochondria compared to methanol or PFA (Fig. 2d-j).
Cryo-ExM highly preserves the cytoskeleton landscape. We further explored the preservation of the cytoskeletal landscape. First, we looked at the actin network of growth cones in cultured hippocampal neurons, known for their unique cytoskeleton organization 20 . By simultaneously imaging the actin cytoskeleton using β-actin antibodies together with microtubules, we found that both cytoskeletons remain intact in cryo-ExM and their canonical organization is preserved, with internal microtubule bundles and actin structures such as filopodia and ruffles forming waves at the cell periphery ( Fig. 3a and Extended Data Fig. 5). Second, we stained U2OS cells for actin and could unveil the different typical actin networks, the lamellipodia, filopodia and stress fibers 21 (Fig. 3b). Finally, we turned to LifeAct to label filamentous actin 22 and analyzed it under different fixation conditions ( Supplementary Fig. 1). Notably, we found that cryo-ExM gave similar results as the gold standard PFA/GA for actin and did not affect expansion as we noticed minimal distortions of 1.6%, similar or smaller to the distortions observed using other expansion microscopy methods 1,10 (Supplementary Fig. 2).
in mitotic RPE-1 cells and observed that cryo-ExM enabled high preservation of the mitotic spindle as well as astral microtubules, which are difficult to maintain owing to the chemical fixations artifacts 23 ( Fig. 3c and Extended Data Fig. 6a-e). We also found that the mitotic spindle displayed an isotropic ~fourfold expansion by measuring the spindle length before and after expansion as well as the centriole as an internal ruler 11 (Extended Data Fig. 6a-c,g,h). Using NHS-ester staining 18 , we observed that cryo-ExM protects the overall organization of the mitotic cells as chromosomes, intact mitochondria and the midbody could be observed (Extended Data Fig. 6e,f). Second, we looked at nondividing cells where cilia protrude from the cell surface and found that motile cilia from αβ-Tubulin GFP-Sec61β  ependymal cells were fully preserved, displaying the canonical length of 7 μm 24 ( Fig. 3d and Extended Data Fig. 7a,b), as well as primary cilia where we detected that microtubule nucleation sites on the underlying mature basal body 25 were overall better preserved than with regular methanol fixation (Extended Data Fig. 7c-f).
Cryo-ExM improves epitope accessibility. Next, as it is known that chemical fixations can affect epitopes accessibility in immunostainings 3,5 , we investigated whether Cryo-ExM could alleviate this issue.
To do so, we first compared fixation effects on the staining intensity of the ER using GFP-Sec61β as a proxy. We found that PFA/GA fixation decreased overall fluorescence intensity by 40% compared to Cryo-ExM (Fig. 4a-c). Then, we analyzed the fixation effect on the outer mitochondrial membrane translocase TOMM20 density ( Fig. 4d-g). Notably we observed a greater labeling density using cryofixation compared to PFA/GA, PFA alone or methanol (Fig. 4g).
We also noticed that the use of MitoTracker allowed us to resolve the mitochondrial cristae, highlighting that the inner architecture of this organelle is intact (Fig. 4h-j and Extended Data Fig. 8). Also, we noticed that when staining microtubules and mitochondria together, the cytoplasmic signal for tubulin was absent in the space occupied by mitochondria ( Supplementary Fig. 3). We hypothesize that this could correspond to the cytoplasmic soluble pool of tubulin that is usually precipitated or lost owing to chemical fixation and permeabilization 3 .
Versatility of the Cryo-ExM method. Finally, we investigated the generality of epitope preservation of Cryo-ExM by assessing other cellular structures such as lysosomes/autophagosomes (Lamp1 and LC3), Golgi apparatus (GM130) and nuclear pores (NUP205). We found that all structures could be visualized in Cryo-ExM, demonstrating the wide range of epitope preservation of this method ( Fig. 5a-d and Extended Data Fig. 9). In addition, we assessed whether cryofixation can solve two well-known artifacts of aldehydes fixations: the exclusion of the transcription factor SOX2 from the DNA in mitosis 26 and the cellular distribution of the cell surface glycoprotein CD44 27 . Probably owing to the instantaneous fixation that prevents protein diffusion, we found that cryofixation preserves the correct localization of SOX2 on chromatin of nonexpanded human embryonic kidney (HEK) cells ( Supplementary  Fig. 4a,b) and that CD44 labeling is highly preserved with a pattern colocalizing with actin fibers as previously observed in two-color SRM 28 as well as in consistency with the known colocalization of CD44 with the Golgi apparatus (Supplementary Fig. 4c-f). Last, we further investigated the ability of Cryo-ExM to safeguard the native cellular organization by imaging a soft organelle that is affected by chemical fixation, namely phase-separated organelles 29 . Therefore, we turned to analyze the pyrenoid, a liquid-like droplet organelle from the green algae Chlamydomonas reinhardtii, made of the densely packed CO 2 -fixing enzyme Rubisco, crucial for the photosynthesis process 29 . Using NHS-ester staining 18,19 , we observed that both methanol and cryofixation could preserve structures inside the pyrenoid that most likely correspond to the pyrenoid tubules, previously observed by cryo-electron tomography αβ-Tubulin    after cryo-FIB-milling 29 (Fig. 5e-h). However, we noticed that upon methanol fixation, the expanded Chlamydomonas cells were slightly collapsed and their pyrenoid shape was variable, as indicated by the roundness index and the area (Fig. 5e,eʹ,i and Extended Data Fig. 10). As phase-separated organelles are often perfect spherical droplet 30 , this result might indicate that methanol fixation induces a protein precipitation deforming the pyrenoid. In contrast, we found that applying Cryo-ExM directly on Chlamydomonas cells seems to better preserve the liquid-droplet shape of the pyrenoid as these showed more homogeneous circularity (Fig. 5f,fʹ,i and Extended Data Fig. 10), suggesting that these cellular structures formed by phase separation remain intact under these conditions.

discussion
In this work, we introduce a new method to perform SRM by coupling cryofixation of a biological specimen with ExM. With this, we provide a universal framework to visualize subcellular compartments without chemical fixation artifacts such as structural alteration and loss of antigenicity. Moreover, Cryo-ExM alleviates the issues of optimizing fixation conditions required to visualize specific structures and is therefore expected to enable more accurate protein localization. Notably, this method also demonstrates that the classical cryo-substitution protocols developed for EM are compatible with expansion microscopy by replacing the EM resin with hydrogel monomer solutions. Therefore, this approach may also be applicable on tissues cryofixed by high-pressure freezing as well as in hydrogel-based tissue clearing. Finally, as expansion microscopy is also compatible with SIM, STED or dSTORM 31-33 , our method now allows all these microscopy modalities to image easily cells in their native state, paving the way for further studies of complex cellular processes.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41592-021-01356-4.

Methods
Reagents and reagent preparation. Cell culture. Chlamydomonas reinhardtii. The cell-wall-less Chlamydomonas strain CW15 was grown in liquid Tris acetate phosphate medium (containing Trace) at 22 °C or on Tris acetate phosphate plates with 1.5% agar, as previously described 34 . Cells were either fixed by immersion in −20 °C chilled methanol for 5 min or cryofixed with plunging in liquid ethane/propane mix (see below).
Mouse neuronal cell culture. Primary cultures of hippocampal neurons were obtained according to the procedure described previously 35 . Hippocampi were dissected from E18.5 mouse embryos in HBSS (Invitrogen) containing HEPES 10 mM, streptomycin 10 µg ml −1 , penicillin 10 U ml −1 , treated with 0.25% trypsin-EDTA for 10 min at 37 °C and disrupted by 10-15 aspirations/ ejections through a 5-ml pipette, followed by ten cycles through a micropipette tip. Dissociated hippocampal neurons were seeded in DMEM (Invitrogen) supplemented with 10% heat-inactivated horse serum at 50,000 cells cm −2 in six-well plates on 12-mm glass coverslips precoated overnight with 50 μg ml −1 poly-d-lysine (Sigma) at 37 °C. At 20 h after seeding, the medium was changed to the culture medium (Neurobasal (Invitrogen), B27 supplement 2%, sodium pyruvate 1 mM, l-glutamine 2 mM, streptomycin 10 µg ml −1 , penicillin 10 U ml −1 ) and neurons were imaged between 2-4 d in vitro. These experiments were carried out in accordance with the Institutional Animal Care and Use Committee of the University of Geneva and with permission of the Geneva cantonal authorities.
U2OS expressing GFP-sec61β (133-291) or LifeAct-GFP were transiently transfected with JetPRIME following the manufacturer's instructions. After 24 h of expression, cells were fixed as described below.
Cell fixation. Cells grown at desired confluence were washed in PBS and either cryofixed (see below) or with the following protocols when specified: (1) by immersion in −20 °C chilled methanol for 5 min, (2) in 4% PFA for 15 min at room temperature, (3) in 3% PFA + 0.1% GA for 20 min at room temperature or (4) in 1% PFA + 0.2% GA (Supplementary Fig. 4c-f) for 20 min at room temperature.
Cryo-ExM protocol. Plunge-freezing and freeze substitution. The 12-mm coverslips containing the sample were held halfway with a thin tweezer (Dumont 5, Sigma F6521-1EA), the excess of remaining medium was strongly blotted with a filter paper and coverslips were rapidly plunged with a homemade plunge freezer into liquid ethane or an ethane/propane mix cooled with liquid nitrogen (Extended Data Fig. 1a-c). Note that the homemade plunger is a classical system used by most of the cryomicroscopy laboratories but an automatic system might work similarly. No difference could be observed between pure ethane and an ethane/propane mix, the latter mix being more convenient because it does not solidify at the temperature of liquid nitrogen 38 . Coverslips were then rapidly transferred into a 5-ml Eppendorf tube containing 2.5 ml of liquid nitrogen-chilled acetone (Extended Data Fig.  1d). Tubes were placed on dry ice with a 45° angle and agitated overnight to allow the temperature to rise to −80 °C (Extended Data Fig. 1d). Samples were further incubated without dry ice for 1.5 h until the temperature reached ~0 °C. Samples were then rehydrated in successive ethanol:water baths, 5 min each, as follows: ethanol 100%, ethanol 100%, ethanol 95%, ethanol 95%, ethanol 70%, ethanol 50% and PBS. Cells were stored in PBS until expansion or directly processed for immunostaining (Extended Data Fig. 6c and Supplementary Figs. 2 and 4a,b).
Ultrastructure expansion microscopy. Expansion of the cells was performed as previously described 39 . Briefly, fixed cells (cryo, PFA, PFA/GA or methanol) were incubated for 3 to 5 h in 2% AA and 1.4% FA diluted in PBS at 37 °C before gelation in monomer solution containing 0.5% tetramethylethylendiamine and ammonium persulfate. Next, cells were incubated for 5 min on ice followed by 1 h at 37 °C and incubated for 1.5 h at 95 °C in denaturation buffer. Gels were washed twice in ddH 2 O. Note that the original U-ExM protocol without previous fixation 10 depolymerizes cytoplasmic microtubules 11 . In contrast, cryofixation before U-ExM protocol preserves cytoplasmic microtubules.
Note that the quality of sample preservation using the plunger was also compared to manual immersion. As shown in Supplementary Fig. 5a-d, manual immersion leads to wavy and broken microtubules, whereas the plunger fully preserves their native structures. Note also that sample fractures could be sometimes observed, as classically observed in cryomicroscopy 40 ( Supplementary Fig. 5e-h).
To visualize the actin network, we either used β-actin antibodies or LifeAct-GFP. Comparison of LifeAct-GFP to β-actin staining patterns confirmed that both could be used to faithfully visualize actin in human cells using Cryo-ExM ( Supplementary Fig. 1).

Image acquisition.
Pieces of gels were mounted on 24-mm round precision coverslips (1.5H, 0117640, Marienfeld) coated with poly-d-lysine for imaging. Image acquisition was performed on an inverted Leica TCS SP8 microscope or a Leica Thunder DMi8 microscope using a ×631.4 NA oil objective with Lightening or Thunder LVCC (large volume computational clearing) mode at max resolution, adaptive as 'Strategy' and water as 'Mounting medium' to generate deconvolved images. Three-dimensional stacks were acquired with 0.12 μm z-intervals and an x, y pixel size of 35 nm (Leica TCS SP8) or 0.21 μm z-intervals and an x, y pixel size of 100 nm (Thunder DMi8).
Quantifications. For each gel, a caliper was used to accurately measure its expanded size. The gel expansion factor was obtained by dividing the expanded size by the original size of the coverslip (12 mm in this work). Each measurement was divided by the calculated expansion factor and reported as such in the graphs or figure scale bars, except in Extended Data Fig. 6g,h where lengths and diameters are indicated as expanded and after rescaling.
Length of the motile cilia. Cilia were manually traced using the segmented line tool of ImageJ 42 . The total measured length was divided by the expansion factor of the gel and reported as a dot plot using GraphPad (https://www.graphpad.com/).
Area and roundness of the pyrenoids. Methanol-fixed and cryofixed pyrenoids were manually delineated using the polygon selection tool of ImageJ. Roundness and area were calculated and reported as dot plot using GraphPad.
GFP-Sec61β intensity signal measurement. For the comparison of GFP-Sec61β signal intensity obtained after cryofixation versus PFA/GA fixation, we measured the mean intensity of identical regions of interest (200 × 200 pixels) from nondeconvolved images of each condition. Five regions of interest per cell were quantified and averaged. Dot-plots were generated using GraphPad.
Mitochondrial area and TOMM20 density measurement. For comparison of mitochondrial morphology under different fixative conditions (cryofixation, methanol, PFA and PFA/GA), we binarized deconvolved images on the MitoTracker channel and applied a mask to draw the outlines of the mitochondrial network. We then quantified the surface area of the mitochondrial network and expressed it as a percentage of the whole cell surface area. For the quantification of the TOMM20 density under different fixative conditions (cryofixation, methanol, PFA and PFA/GA), we manually counted the number of TOMM20 dots in 3-5 mitochondria per cell and expressed it as dots μm −2 . The surface of mitochondria was calculated on the MitoTracker channel as above. Dot-plots were generated using GraphPad.
Mitotic spindle measurements. RPE-1 cells were cryofixed and either directly stained with α/β-tubulin antibodies and DAPI or processed for U-ExM and post-stained with α/β-tubulin antibodies and DAPI. Quantification of the mitotic spindle length at different mitotic stages (prophase, metaphase and anaphase) was performed with the straight-line tool of image and plotted with GraphPad.
Plot profiles. Plot profiles from Fig. 4j, Extended Data Fig. 4f and Supplementary  Fig. 3f were obtained using the straight-line tool of ImageJ and plotted using GraphPad. The distance between either half of the maximal (Extended Data Figs. 4f and 6g) or peak-to-peak (Extended Data Fig. 6h) distance was calculated.
RMS calculation on U2OS cells expressing LifeAct-GFP. U2OS were cultured, transfected with LifeAct-GFP plasmid and cryofixed as described above. GFP-positive cells in a restricted area at the center of the coverslips were rapidly acquired and the coverslips were directly processed for U-ExM as described above. The center of the gel was stained with anti-GFP (see immunostaining section) and acquired with the same microscope as previous expansion. To estimate the sample deformation after expansion, we calculated the r.m.s. error between two images of the same structure before and after expansion, following the protocol described by Chozinski et al 43 . This protocol also provides the scale factor between the images, thus giving the expansion factor of the experiment.
Statistical analysis. The comparison of two groups was performed using a two-sided Student's t-test or its nonparametric correspondent, the Mann-Whitney U-test, if normality was not granted either because of not being checked (n < 10) or because it was rejected (D' Agostino and Pearson test). The comparisons of more than two groups were made using one-way ANOVA followed by post hoc tests (Kruskal-Wallis test) to identify all the significant group differences. n indicates independent biological replicates from distinct samples. Data are represented as scatter-plots with center line as mean. The graphs with error bars indicate 1 s.d. (±) and the significance level is denoted as usual (*P < 0.05, **P < 0.01, ***P < 0.001). All statistical analyses were performed using Prism7 (GraphPad v.7.0a).

Reproducibility.
All experiments were performed at least three times, except for the ependymal cells and neurons, which were performed only once. Representative images are shown for each experiment.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

data availability
The data that support the findings of this study are available as 'source data' provided with the manuscript. Further request can be sent to the corresponding authors. Source data are provided with this paper.