The CryoCapsule: Simplifying correlative light to electron microscopy

Abstract

Correlating complementary multiple scale images of the same object is a straightforward means to decipher biological processes. Light and electron microscopy are the most commonly used imaging techniques, yet despite their complementarity, the experimental procedures available to correlate them are technically complex. We designed and manufactured a new device adapted to many biological specimens, the CryoCapsule, that simplifies the multiple sample preparation steps, which at present separate live cell fluorescence imaging from contextual high-resolution electron microscopy, thus opening new strategies for full correlative light to electron microscopy. We tested the biological application of this highly optimized tool on three different specimens: the in-vitro Xenopus laevis mitotic spindle, melanoma cells over-expressing YFP-langerin sequestered in organized membranous subcellular organelles and a pigmented melanocytic cell in which the endosomal system was labeled with internalized fluorescent transferrin.

Introduction

Correlative light and electron microscopy (CLEM) aims at bridging the time and resolution gap between light microscopy (LM) and electron microscopy (EM) [1–4].

A critical step in CLEM is the immobilization of the specimen between the LM and the EM. As electron microscopy for cell biology developed, chemical fixation was extensively investigated as an easy-to-use, affordable and time saving method for ultrastructure observation. But drawbacks inherent in the chemical fixation restrict the interpretation of dynamic events and their ultrastructure at the EM level in a CLEM perspective such as 1) the slowness of the sample fixation (a few seconds to a couple of minutes depending on the specimen thickness and composition); 2) the chemically induced ultrastructure modifications (membrane reticulations, ultrastructure reorganization, shrinking by dehydration and embedding of the specimen); 3) the inefficiency of fixation for some specimens (C. elegans worms are motile for a couple of hours in 10% glutaraldehyde [2]); and 4) the quenching of the fluorescence [5].

To preserve a sample’s molecular and structural integrity, freezing methods have been developed to cryo-immobilize or vitrify specimens. Vitrification occurs within a few milliseconds and preserves a cell’s ultrastructure [6]. To date High Pressure Freezing (HPF) is the sole method allowing vitrification of samples from cells to small organisms [7–11]. In HPF, the pressure is increased to 2100 bars and concomitantly the temperature is decreased to −196°C by liquid nitrogen within a 10ms timeframe. Under these conditions, the physical properties of water are modified, reducing ice nucleation during solidification by freezing [10] that preserves the cell ultrastructure in a “close to native state” [8].

To establish a biologically meaningful CLEM workflow, dynamic fluorescent light microscopy must be rapidly followed by cryo-immobilization to proceed with EM [12]. But High Pressure Freezing machines (HPM) are complex heavy machines [10,13] in which the HPF chamber, where the vitrification occurs is small, constrained and poorly accessible. To ensure proper specimen loading, multiple adaptors need to be assembled that delay the transfer, reducing the biologically relevant time scale of the CLEM method (http://www.youtube.com/watch?v=9kA2lswUpvw). Accelerating and improving CLEM-HPF requires the development of tools that are compatible with the high-end LM standards, physically support the HPF process and facilitate the transfer of the specimen into the EM. Such tools will improve the temporal resolution between the last LM picture and the HPF [1,3] and guarantee high end EM.

Three major HPM instruments are currently used among the cell biology community: the HPM010, the HPM100 and the EMPACT-1 and 2. The EMPACT-2 was designed to facilitate the rapid transfer from LMs to the HPF [3], but not all laboratories are equipped with this expensive equipment and mechanistic constraints of the EMPACT technology limit the sample size to less than 1.4mm [3,13,14]. In this article, we will refer to this machine as the EMPACT-2. On the other hand, the HPM010 can support specimens up to 2mm in diameter while the HPM100 can use specimens up to 5mm in diameter. As many laboratories are equipped with the HPM010 or the HPM100, a convenient tool is required to simplify transfer from LMs to the HPF of these two more versatile machines. In this article, we will refer to these two machines as the HPMs.

We developed the CryoCapsule to accelerate, facilitate and standardize the sample manipulations throughout the CLEM workflow. We imaged our biological samples for 5 minutes before manual transfer from the light microscope to the HPM in a 15 second step. We reduced the non-thermo-conductive mass to the minimum, therefore achieving cryo-immobilization in the culture medium without needing to add cryo-protectants [8,15] to preserve cell physiology. New HPMs adaptors were designed to prevent physical stress before vitrification. Finally, the shape of the CryoCapsule facilitates the handling steps for various specimens throughout the whole CLEM process until the targeted ultramicrotomy [1]. The later stages of sample preparation are common to most EM approaches. We greatly simplified and accelerated the sample preparation process and increased the experimental reproducibility.

To challenge the CryoCapsule at the biological level, we tested three different samples: a fragile in-vitro specimen, the Xenopus laevis mitotic spindle, described as pressure sensitive [16]; melanoma cells (M10 cells) over-expressing a YFP-tagged Langerin that generated a membranous layered structure that is clearly visualized by EM [17]; and a melanocytic cells line (MNT-1 cells) where we followed the endosomal network with fluorescent transferrin bound to endogenous transferrin receptors [18].

Results

The CryoCapsule: a new standard tool for CLEM in cell biology

The CryoCapsule is fabricated using over-molding technology. This process consists of injecting into a mold containing the inserts (the gold spacer ring and the landmarked substrate sapphire disc) a melted polyethylene polymer (PET) at high temperature and then letting it cool down and solidify. This industrial production method [19] ensures the complete reproducibility of the process and guarantees standardized production of the CryoCapsules (Figure 1). The final object is consistently of uniform dimensions (4.5mm diameter, 0.6mm thick, Figure 1 A and D) and exhibits reproducible biophysical properties (sapphire disc, gold spacer ring and polyethylene plastic ring, Figure 1 B). To our knowledge, gold and sapphire are not toxic to the cells. Although the release of PET phthalates is currently being investigated with respect to its impact on embryonic development [20]. However, different cell types (MDCK, HeLa, MNT1 and M10 melanoma cells) grew normally and no increased cell death was observed.

Figure 1.

Detailed view of the CryoCapsule. A- Picture of the CryoCapsule on top of a 1 Euro cent coin, to illustrate the relative proportions; scale bar 4.5mm. Obverse (left) and reverse (right) magnified view of the CryoCapsule. Concentrically, the plastic ring (1), the gold spacer ring (2) and the landmarked substrate sapphire disc (3); scale bar: 0.5mm. C- Scanning EM of the CryoCapsule; scale bar: 2mm. D- Scaled side view of the CryoCapsule. E- Three dimensional cross section view of the CryoCapsule. The substrate sapphire disc and the gold spacer ring are embedded into the plastic ring and maintained together. The covering sapphire disc fit into the plastic ring and isolates the sample. F- Magnified three dimensional view of the CryoCapsule. The covering sapphire disc does not touch the plastic ring and lies on top of the gold spacer ring. Between the two sapphire discs and the gold spacer ring is the CLEM chamber.

The CryoCapsule is designed to simplify CLEM-HPF experiments. Therefore, we analyzed the thermal properties of the individual elements constituting the CryoCapsule. Sapphire discs of 50μm have a cooling rate of 0.2ms (Supplementary table) while rigid enough to avoid deformations or breakage during cryo-immobilization (depending on the HPM model, see below). Aclar®33C Fluoropolymer sheets, often used as an alternative to sapphire discs [21,22] are flexible and have a longer cooling rate (23.3ms; Supplementary table). We used 50μm sapphire discs to guarantee the best cooling rate for the CryoCapsule. The gold spacer ring has a very fast cooling rate (0.02ms) and its height (50μm) defines the CLEM chamber depth (Figure 1 E–F, Supplementary table). The PET ring surrounding the CryoCapsule is not thermo-conductive and has a long cooling time (167ms; Supplementary table), but is peripheral. Therefore it does not prevent the cooling of the sapphires and the subsequent cryo-immobilization of the biological specimen (Figure 3, 4 and 5) in HPMs (as opposed to the EMPACT that separates pressure and cooling sources [13] as discussed below). Finally, the water depth, defined by the height of the gold spacer ring, is the main thermo-resistant source (from 17.8 to 285.7ms; Supplementary table). All together, the CryoCapsule is designed and composed of components supporting an optimal cryo-immobilization as required for CLEM-HPF.

In the live cell CLEM experiment step, carbon evaporated landmarks on the substrate sapphire disc (Figure 1 B, Figure 2 B) [23] are used to locate the cells throughout the CLEM process [1]. The substrate sapphire disk is inserted 275μm deep into the CryoCapsule (Figure 1 D), therefore long working distance objectives are required (Figure 1 D–F, Figure S2 B). Industrial sapphire has a transmission range from 150nm to 5.5μm and a refractive index (RI) of n=1.67. The commonly used fluorescent proteins have a working spectrum from 350 to 700nm and the objectives are usually corrected for cover glass with a RI n=1.54. Although sapphire is more refractive, we nevertheless obtained high quality fluorescence images, which were used to conduct the live cell imaging correlative experiments (see below Figures 4 and 5, Figure S1 and Video S1, Supporting Information).

The CryoCapsule simplifies both the specimen manipulation from LM to the HPM and the specimen embedding

Currently, most CLEM approaches using HPMs rely on a complex assembly of sub-elements appropriately super-imposed in their respective adaptors before HPF. One possible assembly, derived from the work of J. Hawes et al. [15], is depicted in the Figure 2 C and the Figure S1 B. The total height of the two carriers (500μm, type B) plus the two sapphire discs (50μm) and the gold spacer ring (50μm) generate an assembly of 1150μm that needs to fit into the 1000μm deep sample support (Figure 2 F). Using this assembly, we noticed that the structure of a pressure-sensitive sample like the in vitro Xenopus laevis mitotic spindle [16] was disrupted (Figure S1 A–C).

Figure 2.

The CryoCapsule is an integrated technology that simplifies the manipulation of biological specimen in HPFCLEM approaches. A- Schematic views of the CryoCapsule in use: the substrate sapphire disc of the CryoCapsule is already landmarked by carbon evaporation through a finder grid mask. After sterilization (1) for an experiment, the specimen (red star) is loaded (2) and a 2.5mm diameter covering sapphire disc (blue arrow) is deposited to encapsulate the specimen (3), creating a CLEM chamber. B- Carbon evaporation process used in both the CryoCapsule and the conventional CLEM approaches. A finder grid is deposited onto a substrate sapphire disc (1, 2) and carbon is evaporated through it, creating a negative mask to localize the samples of interest throughout the CLEM process (3). C- Possible sapphire assembly for correlative approaches: a biological specimen is deposited onto a landmarked substrate sapphire disc (1), a spacer ring (2) is laid on top of it followed by a 3mm diameter covering sapphire disc (3). D-E-F- After assembly of the CLEM chamber, the specimen is prepared for live cell imaging to identify a structure of interest prior to HPF. D–E The CryoCapsule is loaded into a bottom-perforated humid chamber for direct imaging using a long working distance objective (1). After identification of the sample, the CryoCapsule is collected and mounted into the HPM010 tip (2) for HPF. E- The CryoCapsule is clipped into the HPM100 middle plate adaptor (1; Pictures of the obverse and reverse sides of the loaded adaptor). F- The sapphire assembly is imaged in a glass bottom humid chamber (1) to identify the region of interest. The assembly is mounted between 2 aluminum carriers (2) and loaded into the HPM adaptors before HPF (3). G- The CryoCapsule is transferred into the freeze-substitution solution (a) and the covering sapphire disc is gently removed, manipulating only the plastic ring and the covering sapphire disc (b). The orientation is done reading the carbon letters or facing the gold spacer ring up (c). After embedding, the removal of the soft plastic ring (d) leaves the sapphire disc accessible for a careful physical removal (e; no heat shock required). The carbon pattern transfers to the resin for a facilitated targeted ultra-microtomy (picture). H- The sapphire assembly is transferred into the freeze substitution solution (a) and disassembled to collect the substrate sapphire disc (b). Orientation of the sample is done exclusively by reading the carbon letters (c). After embedding, the sapphire disc is cleared (d) and removal is done using heat shock or mechanical removal (e). The carbon pattern transfer is used for targeted ultra-microtomy (f).

Using the CryoCapsule prior to a CLEM-HPF experiment, a 2.5mm diameter covering sapphire disc (CSD) is laid in the 2.8mm CryoCapsule cavity (Figure 1 D) on top of the gold spacer ring to close the isolating CLEM chamber where the sample is situated (Figures 1D–F and 2A, Video S2). Once the CLEM chamber is closed, all subsequent manipulations are done using the plastic ring (Figure 1B,C, Video S2). Moreover, the CryoCapsule is radially symmetric (Figure 1) to simplify the loading into the HPM holders (Figure 2 D–E), and axially asymmetric (Figure 1) to facilitate the sample orientation upon embedding (Figure 1 B, 2 G).

We took into consideration the vitrification mechanism of both HPMs and developed two specific adaptors (Figure 2 D, E, Figure S2). They enable easier transfer of the CryoCapsule into the HPMs after live cell imaging as well as a decrease in the pressure on the specimen prior to HPF, minimizing pre-fixation stress (see below).

In the HPM010, vitrification occurs by direct impact of pressurized liquid nitrogen on the sample from both sides in the HPF chamber [24]. We designed a clamp in which we removed the layer of aluminum on the SSD side as compared to conventional methods (Figure S2 A) to obtain a direct hit of liquid nitrogen during HPF on the substrate sapphire disc. To avoid deleterious shear forces or pressure applied to the specimen before vitrification, the CryoCapsule is maintained in the clamp by its plastic ring (Figure S2 A). We tested the CryoCapsule by repeating the X. laevis experiment and observed no modification of the spindle assembly (Figure S1) indicating that the CryoCapsule preserves the structural integrity of wide and fragile samples.

For CLEM approaches, the CryoCapsule is placed in a humid chamber during live cell imaging (Figure 1 D). Once the event of interest has been identified, the CryoCapsule is mounted in the HPM010 clamp for HPF (Figure D, Figure S2 A). The transfer operation takes approximately 30s depending on the user’s skills.

In contrast to the HPM010, in the HPM100 the liquid nitrogen comes as a side flow, so we left a direct exposure of the sapphire disc to the LN2 from both sides. After deposition of the CSD to close the CLEM chamber, the CryoCapsule is clipped in the middle plate adaptor that we specially designed and produced (Figure 2E, Video S1, S2). The CSD is maintained in place by a lip and the whole assembly stays clipped (Figure S2B, Video S1, S2) unless an external force is applied. For CLEM approaches, the assembly of the CryoCapsule and Adaptor is imaged in a humid chamber. After identification of the event of interest, manual transfer is done in approximately 15s according to the user’s skills, and is therefore suitable to study cell processes happening within a timeframe of at least 20s (See below Figure 5, Video S2).

The EMPACT high pressure freezer uses a fundamentally different mechanism to manipulate and cryo-immobilize the specimen [13]. The cooling, separated from the pressure source, comes to the side of the specimen [13] and the CryoCapsule is surrounded by a thermo-resistant polymer. We expect the plastic ring to prevent proper cooling of the specimen, hence cryo-immobilization. Furthermore, a rapid transfer system (RTS) has already been developed for the EMPACT-2. Thus we only tried the compatibility of the CryoCapsule with HPM010 and −100.

Using both HPMs, the manipulations are reduced to: 1- placing the CSD, 2- loading the closed CryoCapsule in the HPM adaptor, 3- vitrifying. At no time is the specimen exposed to external shear forces or pressure as all the manipulations are done via the plastic ring. Altogether, the CryoCapsule is compatible with both HPMs.

High Pressure Freezing using the CryoCapsule preserves cell morphology

To validate the vitrification efficiency of the CryoCapsule we tested different cell lines using both HPMs (Figure 3).

After sterilization in 70% ethanol, the capsules were placed in culture Petri dishes, allowed to dry for a couple of hours and then HeLa cells, MDCK or MNT1 and M10 melanoma cells were plated (Figure 3–4).

Figure 3.

The CryoCapsule allows the vitrification of a large range of cell lines with both the HPM010 and the HPM100. A- MDCK cells were vitrified using hexadecene as medium filler in a HPM010, freeze substitution: 0.1% uranyl acetate, 0.2% glutaraldehyde and 2% H2O in dry acetone. B- MNT1 cells were vitrified in an HPM100 in their culture medium without cryo-protectants or filler, freeze-substitution: 0.1% uranyl acetate, 0.05% glutaraldehyde, 1% H2O in dry acetone. C- HeLa Kyoto cells were vitrified in a HPM010 without cryoprotectants or filler, freeze-substitution: 0.1% uranyl acetate in dry acetone. D- HeLa Kyoto cells were vitrified in an HPM100 without cryo-protectants or filler, freeze substitution: 0.1% uranyl acetate in dry acetone. Scale bars=10m.

Generally 5 CryoCapsules are prepared per HPF experiment. MDCK cells were vitrified with an HPM010 using hexadecene as a filler to reduce the water content (Figure 3 A; [8]), MNT1 cells were cryo-immobilized on a HPM100 in culture medium containing HEPES (Figure 3B) and finally HeLa cells were cryo-immobilized in culture medium. In each case the experiments were repeated using both HPMs (Figure 3 C, D). M10 Melanoma cells were also cryo-immobilized, using a HPM100, in culture medium containing HEPES (Figure 4). In all cases, the ultrastructure of the cells was well preserved without any detectable segregation artifacts (ice patches, broken membranes). The cytoplasm was homogeneous and the Golgi apparatus, the nuclear membrane and the mitochondria were well preserved (Figure 3). Overall, the CryoCapsule is compatible with cell cultures and preserves the ultrastructure of high-pressure frozen cells without the use of cryo-protectants.

CLEM with the CryoCapsule

Mainly two different approaches exist to achieve CLEM. With the first, the specimen is imaged live, fixed and the cell of interest is recovered in the EM. This method is often applied to time-lapse studies or to narrow down to the cell of interest from a population of cells [1,25]. The second approach consists of cryo-immobilizing the specimen and then imaging the fluorescence directly on the EM sections prior to EM imaging. Using landmarks on the sections, the structure of interest is accurately localized (below 100nm precision) on a physical slice of the cell [26].

The CryoCapsule allows combination of these two approaches into a full CLEM workflow where one can monitor and image the sample between each processing step. We used M10 melanoma cells stably expressing YFP-tagged langerin (the component of the Birbeck granule of Langerhans cells) in its cytoplasmic domain (YFP-langerin) [17]. These cells exhibit a very bright fluorescent signal, easy to detect by light microscopy at low magnification (10x or 40x air extra-long working distance (ELWD) objectives) (Figure 4, Video S1). In these cells, the YFP-langerin accumulates into Organized Smooth Endoplasmic Reticulum (OSER)-like structures [17]. These structures correspond to regions of the ER where the YFP-langerin accumulates by dimerization of the YFP, trapping the fluorescent signal in large layered compartments that also share common morphological features with Birbeck granules [17]. At the EM level, these layered membranous structures sequester the heavy metal salts used to contrast the specimen and are clearly identifiable electron dense and organized regions (Figure 4 G–J) [17].

Figure 4.

Full CLEM workflow. A- The cells (arrow head) in the CLEM chamber are localized at the surface of the substrate sapphire disc using the carbon landmarks; scale bars: 50μm. B- The cells of interest (YFP-langerin stably expressed in M10 melanoma cells) are identified in fluorescent channel (arrowhead); the fluorescent signal of the OSER is voluntarily saturated for illustration purpose and is indicated with a white arrow; scale bars: 50μm. C- Magnified overlay of the cell of interest (dashed line A,B); scale bars: 50μm. D- Schematic representation of the sample preparation for the second imaging step: the cells are serial sectioned (250nm thick) and collected on slot grids. Each individual section is imaged by bright field and fluorescence light microscopy then imaged by EM, and finally overlaid. E- Affine overlay is done with very low magnification image (540x) and used to locate the OSER within the cells (arrowhead); scale bar: 50μm. Flow magnification (2400x) picture of the cell is overlaid with the former overlay (E) to locate the OSER (arrowhead); scale bar: 5μm. G- Medium magnification image of the OSER (arrowhead); scale bar: 5μm. H- High magnification tilted EM (−54°) acquired on a 200kV EM: the layered structure is clearly identifiable (arrowhead); scale bar: 1μm. I- electron tomogram top view slice of the OSER showing a clear layered organization; scale bar 0.2 μm. J- Electron tomogram cross view of the OSER (along the yellow line in I); scale bar 0.2μm.

After identification of the cells of interest by live cell imaging and their localization using the carbon pattern (Figure 4A–C, Video S1), the cells were cryo-immobilized in a HPM100 followed by freeze substitution (FS; Figure 4A,Video S1). As we also planned to do fluorescence imaging on these sections, the cells were embedded in HM20-Lowicryl with low a fixative concentration to preserve the fluorescence of the tagged proteins [5,27]. The CryoCapsule was extracted from the adaptor directly into the FS solution at −90°C; the removal of the CSD was done by simply lifting the CryoCapsule to the side and letting the CSD fall off by gravity (Figure 2 G). PET is not dissolved by acetone; hence the entire structure stands as a whole assembly until embedding in Lowicryl. After polymerization, the PET ring is removed using a razor blade and the substrate sapphire disc is gently mechanically removed. The carbon landmarks transfer to the Lowicryl bloc, allowing a precise targeted ultramicrotomy (Figure 2 G). Serial sections (300nm in Figure 4 or 100nm in Video S1) are collected on slot grids and imaged by fluorescence and transmitted light microscopy (Figure 4 D). The sections are not flat on slot grids, so a z-stack is acquired to collect the entire fluorescent signal. Lowicryl is a rather soft embedding resin and uneven UV-polymerization leads to small imperfections in the bloc. Tiny holes appear on the plastic sections that diffract strongly in transmitted light, inverting from black to white according to the focal plane. These imperfections were used as “natural landmarks” to register the fluorescence in the EM images where the same holes appear white (Figure 4 E). Although not as precise as the fluorescent fiducial markers used by Kukulski et al. [27], they are sufficient to accurately locate the OSER signal within our cells (Figure 4, Video S1). Registration was done using the plugin TurboReg [28] in Fiji. The calculated affine deformation from the transmitted light image was applied to a Z projection of the fluorescent stack after background removal and then overlaid with the EM image at very low magnification (Figure 4E, Video S1). This mapping is then used to retrieve the fluorescent signal within the cell (Figure 4F, Video S1), acquire tilt series images to observe the layered structures described previously [17] and reconstruct electron tomograms of the OSER (Figure 4 G–J). This indicated that the specimen cryo-immobilized with the CryoCapsules were equally well preserved as other cell monolayers using HPF approaches and that the CryoCapsule is suitable for a full CLEM approach.

The M10 melanoma cells that we used to set-up the workflow express an unusually high level of fluorescence and the OSER structures are rather immobile. We aimed to further challenge the use and benefits of the CryoCapsule to overcome the technical limitations of the HPM transfer systems. With this concern in mind, we investigated the dynamic interactions between the endosomal and melanosomal network of pigmented melanocytic cells. Previous studies from Delevoye et al. [18] have shown that in MNT1 cells, endosomes followed with internalized fluorescent transferrin (Tf) interact transiently with melanosomes (~20s) [18]. To track the endosomal network and accurately identify it near melanosomes we acquired wide field images for 3′20″ by bright field and fluorescence microscopy every 25s with a 40X air objective (Figure 5 A). At the top left of the cell, we selected a group of melanosomes (Figure 5A, black arrowhead) apposed to Tf-positive endosomes (Figure 5A, red arrowhead) at the end of the imaging period (2′55″ and 3′20″, insert box). The Tf-fluorescent signal, slightly below the melanosomes (Figure 5 B), has an elongated form with a brighter signal towards the bottom (Figure 5B, inserts). Then we manually transferred the CryoCapsule into the HPM100 for cryo-immobilization in less than 15s and processed it for EM (Video S2). We imaged all the fluorescence on the section on the day of sectioning, since fluorescence is rapidly quenched after sectioning (less than 24 hours). As the Tf-A555 is internalized by binding to the endogenous pool of Tf receptors, the intensity of the fluorescent signal is rather low. This second step of light imaging was done using a 60x oil objective with a numerical aperture of 1.4 in order to easily detect the fluorescence on the EM sections and increase the fluorescence-localization accuracy.

Figure 5.

Correlative microscopy of the endo-melanosomal dynamic networks’ interactions. A- MNT-1 cells were imaged by transmitted and fluorescence light for 3′20″ before a rapid transfer into the HPM100 for cryo-immobilization. In the last two frames, a Tf-positive endosomal structure of tubular appearance (red arrowhead) was observed close to the melanosomes (black arrow head) and was targeted for correlative imaging; the boxes are magnified views of the dashed white box (first frames); scale bar: 50μm. B- Overview of the MNT-1 cell at the last time point before cryo-immobilization; the black box represents the field of view covered in A and C. The Tf-positive tubular structure (red arrowhead) in the white box is slightly below the melanosomes (black arrowhead); the white box corresponds to the magnified field of view in A and C–G; scale bar: 50μm. C- Low magnification EM of the cell of interest. The white box delimits the region of interest, the black arrowhead points the melanosomes; scale bar: 50μm. D- Overlaid magnified view of the white box (A–C); the melanosomes (black bow) are above the endosomes (red arrowhead); scale bar: 20μm. E–G- First 3 sections containing biological material, Z=200nm, 400nm, 600nm. The melanosomes (black bow) and the endosomes (red arrows) are observed; scale bar: 10μm. Below are the magnified views of the black boxes. Scale bars: 0.50μm.

The image alignment was done following the same procedure as described previously but using the black pigmented melanosomes as natural landmarks instead: they are easily distinguishable by bright field imaging even on EM sections. The cell shape and the melanosomes are also used to correlate the live cell imaging with the EM sections. The overview of the cell, taken at low magnification by EM (Figure 5C) displays a comparable distribution of the melanosomes at the cell periphery. The first 13 serial sections (200nm each) imaged both by light and electron microscopy and registered for fluorescence were aligned manually with each other with MIDAS [29] to reconstruct part of the endosomal-melanosomal network (Figure 5 E–G). Fluorescence of the section aligned with the EM (Figure 5D) allows one to identify the labeled endosomal network (Figure 5D, red arrow) apposed to melanosomes (Figure 5D, black bow) from other tubular structures. The melanosomes were concentrated in the bottom 3 serial sections (Figure 5E–G). Tubular endosomal structures expand across the 3 sections and the intensity is brighter towards the bottom (Figure 5G) as observed by LM (Figure 5B). Taken together our data show that the CryoCapsule enables a full CLEM approach on either static fluorescently labeled structures or highly dynamic endosomal tubules that can be followed with a low intensity fluorescent signal, adequate for broad cell biology approaches.

In conclusion, and as compared to the currently available tools and methods, the CryoCapsule simplifies the sample handling from the cell culture step and the sample loading until the targeted ultramicrotomy (Video S2). This simplification allows us to track and image the specimen at almost every processing step, ensuring a reliable integrative full CLEM analysis of dynamic biological process (i.e. within a timescale not less than 20s).

Discussion

CLEM is regarded as a powerful and straightforward method for cell biologists, but it is also considered technically challenging. The handling of the specimen is often complex and requires careful adaptation to each biological object [4,14,25,27,30]. Generally, vitrification by HPF of cell suspensions has a success rate of approximately 80% using cryo-protectants [8]. Live CLEM approaches require identifying the specimen before HPF and retrieving it after sample processing for EM [1,3,31]. The handling complexity of the specimen dramatically reduces the success rate for each experiment.

We created the CryoCapsule to provide a robust and reproducible approach to perform CLEM, while minimizing the external stress factors that could influence the interpretation of the experiment. Hence we addressed the sample manipulation, the cell culture and imaging settings, the fixation conditions, the live imaging time and the dimensions of the sample.

For most studies on cell lines, the cells are cultured on 50μm thick sapphire or Aclar®33C discs [22,23]. Often, the transfer from the light microscope to the HPM is a source of stress for the cells. 50μm thick discs are difficult to handle with tweezers and can be dropped, inducing dehydration of the specimen thus damaging the cells before cryo-immobilization. Furthermore, assembling several small independent parts (B carriers, spacer rings) takes a significant amount of time and skill. This hinders conducting fast correlative experiments, increases the risk of damage to the cells of interest and limits routine use of CLEM on HPMs. The CryoCapsule includes the elements commonly used in CLEM approaches and hold them in a single unit, reducing the handling to a single element. Moreover, the size and shape of the CryoCapsule facilitate the manipulations in the various confined environments inherent to EM (HPF, freeze-substitution, ultramicrotomy). The simplified handling avoids the addition of undesired stress on the specimen, improving the reproducibility of each experiment.

Compared to a naked sapphire disc supporting the cells, orientation of the specimen in the FS media is significantly improved, reducing the risk of damaging the cells with tweezers. Last but not least, determining how best to orientate the SSD in the freeze substitution automate is straightforward (Axial asymmetry of the CryoCapsule, Figure 1B).

The design of the CryoCapsule also allowed us to cryo-immobilize the cells in their culture media [15] (Figure 3). Often in HPF, the use of cryo-protectants such as BSA 20%, Dextran 40KDa 20% or filler like hexadecene is advised to improve the success rate of vitrification [8,9,32,33] but these components introduce the cells to non-physiological conditions. As cryo-immobilization by HPF aims to preserve the ultrastructure as “close to native state” as possible, the adjunction of cryo-protectants may stress and influence cell ultrastructure.

To investigate the consequences of physical stress applied to the specimen prior to HPF, we used the Xenopus laevis mitotic spindle and simulated its immobilization by HPF. Xenopus laevis mitotic spindle has been described as sensitive to pressure and shear forces [16]. We loaded the specimen in the CryoCapsule, closed the HPM010 clamp and transferred the clamp with the enclosed CryoCapsule into the HPM chamber but did not proceed to HPF (Figure S1 D–E). The sample was recovered and imaged again under the same light microscope. We observed a moderated specimen shift with regard to the carbon landmark used to localize and identify the mitotic spindle (Figure S1 F). We assume this shift is the consequence of the dynamic instability of the microtubules that locally adhere to the substrate, inducing displacements of the structure. If shear forces or pressure were applied they were not sufficient to disassemble the spindle [16].

Using HPF, the main limiting factor to vitrification itself is the water content of the specimen (Supplementary table). Although our experiments show that we can cryo-immobilize many cultured cells without cryo-protectant, all cell types show different requirements for vitrification and we cannot exclude the need for cryo-protection in some cases. Also, with deeper CryoCapsules, cryo-protection might be required. Combining the limited physical stresses and cryo-immobilization in growth media, the specimens are kept very close to native state until cryo-immobilization. The observations done at EM level are hence very comparable to their “native state”.

Once cryo-immobilized, the sapphire disc in the CryoCapsule prevents performing cryo-sectioning and evaluating a “vitreous state”. We used the cell morphology to evaluate the fixation quality.

To conduct live cell imaging in the CryoCapsule, we added HEPES to the culture medium as our stage controls the temperature but not CO₂. When the experimental temperature was maintained at 37°C, a progressive evaporation occurred causing the appearance of air bubbles after 6 or 7 minutes of observation. As well as the deleterious dehydration of the specimen, air bubbles, even tiny, prevent the raising to full pressure in the sample chamber (compression of the gas) compromising proper cryo-fixation. Hence we kept the imaging time within 5min. In this time frame, experiments were successful. For longer experiments, further adaptation and developments will need to be implemented.

The newly developed CryoCapsule is 50μm deep, limiting its use to cell cultures or very thin specimens. In the future, thicker spacer rings will be available to use depending on the biological requirements e.g. tissues, up to the vitrification’s physical limit of 300μm. The imaging window width is 2mm. Although for many cell biology experiments it might be sufficient, some specimens might require a larger support. We evaluate the feasibility of using a 5mm imaging window CryoCapsule based on the 6mm sapphire discs that is used on the HPM100. The larger size of this CryoCapsule opens possibilities to using immersion objectives (as discussed below).

Correlative microscopy for HPMs as we described here relies on two fluorescent imaging phases: first, the cells are imaged live before HPF. So far in this first phase, most approaches are based on live cell imaging on a substrate supported in a Petri dish surrounded by culture medium. In that case, the light path goes through a first diffraction barrier of the Petri dish glass bottom (RI: n~1.52) and the liquid phase of the culture medium (Water: n~1.33), the substrate support (Sapphire: n~1.67; Aclar: n~1.45) before reaching the specimen. As none of these materials have the same refractive index, the light path is strongly perturbed reducing significantly the detection level. The EMPACT-2 recently circumvented these constraints with the inverted RTS setup [34]. For the HPMs, the CryoCapsule as a standalone element, permits direct imaging through the sapphire substrate, removing at least one phase transition. The current use of air objective (Air: n=1) is not optimal and the 6 mm CryoCapsule should allow the use of immersion objectives to improve significantly the light sensitivity.

Bearing in mind all these technical considerations, we analyzed two different types of samples in the CLEM method presented here. First, the M10–11G provided an ideal cell model to set-up the workflow because of their very bright and stable fluorescent signal [17]; long working distance objectives were used because fluorescence was not a limiting factor. Setting up this method, we decided to use the same light microscope in every fluorescence step to obtain directly comparable images. As we aimed for a rapid transfer of the specimen from the light to the EM, the use of immersion objectives was avoided for two reasons: 1) Oil is thermo-resistant and requires solvents to be efficiently removed and so is inappropriate for “live cell” CLEM. Water is also a thermo-resistant although blotting would be possible, but prevents a rapid transfer. 2) The shape of the actual CryoCapsule leads the substrate sapphire disc to be 275μm deep in the CryoCapsule. So a water immersion objective with at least 300μm working distance would be necessary. Considering the very bright signal and large stable structures of interest in our specimen, we estimated the use of 10X or 40X for live cell imaging to be sufficient for setting up our method. For the fluorescence on sections, the method described by Kukulski et al. [27] uses a 100x oil immersion objective on single sections. We optimized our method for a full cell correlative approach, where we image the fluorescence on every section across 50 serial sections, sometimes imaging several cells on each section. As the fluorescent signal is very bright and our initial aim was not to image small subcellular structures, a broader field of view was more convenient and facilitated the localization of the cells on each section.

Although we used the first cell system to more easily establish the setup, our goal was to be able to image highly dynamic structures. The second sample we analyzed was more challenging with respect to its dynamic nature and the fluorescence intensity levels. We followed smaller subcellular structures, such as endosomes, that can be labeled with an internalized fluorescent ligand bound to a pool of endogenously expressed receptors. The early endosomal network in MNT-1 cells is composed of vacuolar endosomes and 70nm wide endosomal tubules that can extend over a micrometer in close proximity to pigmented melanosomes [18]. We used the 40x air objective for live cell imaging for the reasons already stated above. But a better detection of the fluorescence across several sections was necessary to investigate this complex network. Using a 60x oil immersion objective, the signal sensitivity and the localization were significantly improved opening post-processing image analysis solutions going towards super-resolution imaging [35] while keeping a reasonably large field of view, helping to orient the specimen. Last, for this second sample, we used a protein coupled to Alexa-555 and we were able to detect a bright fluorescent signal. We are not aware of another publication reporting the efficiency of this fluorophore after HPF so it adds to the growing list of fluorescent labels that can be used for correlative experiments.

As CLEM grows more important for biological analysis, the CryoCapsule is a novel tool that simplifies the sample preparation and manipulation along the several steps required for CLEM. One potential application is for the cryo-CLEM, observing the vitrified cells in a cryo-stage for later super-resolution microscopy approaches (after vitrification, in cryo-stages).

Materials and Methods

Cell Culture

The CryoCapsules (CryoCapCell, Paris, France) were stored for 72 hours at +60°C to harden the carbon layer, sterilized in 70% Ethanol a day in advance and left to dry in the petri dish where the cells were plated.

HeLa Kyoto cells were grown in 4.5 g/L glucose DMEM (Gibco), 10% Fetal Calf Serum (FCS), 1% penicillin/streptomycin, 1% glutamine and plated into the CryoCapsules two days before HPF.

MDCK cells were cultured to confluence in MEM supplemented with 10% FCS and plated onto the CryoCapsules two days before HPF.

M10–11G cells were cultured in RPMI 1640 (Gibco), 10% FCS, 0.5% geneticyn (Gibco) for two to three days before vitrification to ensure good attachment of the cells. Cells were grown so they would grow separately to simplify the cells identification and location in CLEM experiments.

MNT-1 cells were cultured in DMEM, 4,5g/L glucose, 20% FCS, glutamine, penicillin, streptomyosin, for 3 days before HPF. Cells were not grown to confluence. For the CLEM experiment, the cells were starved for 1h and incubated with Alexa-555-labeled Tf (Tf-A555. 10 μg/ml) for 30 min prior to live cell imaging.

In vitro mitotic spindle preparation and imaging

The mitotic cytostatic factor free spindles were assembled in vitro according to the procedure first described in [36]. They were directly pipetted into the CryoCapsules without pre-treatment. The spindles were imaged on a Zeiss Axio Observer Z1, with a 20x LD plan-NeoFluar (NA 0.2; WD 8.5), LED module 540–580nm.

High Pressure Freezing

Validation of cryo-immobilization was done on a HPM010 (Abra Fluid) at ABRA FLUID AG, Widnau, Switzerland. The correlative approach plus cryo-immobilization validation was done on a HPM100 (Leica Microsystem), Institut Curie, Paris France.

Sample orientation of the specimen during Freeze substitution

After cryo-immobilization, the specimens are freeze substituted with Lowicryl (HM20) in an AFS-2 for further analysis by EM. In the freeze-substitution solution at −90°C, the CSD is gently removed and the CryoCapsule orientation is elucidated by reading the carbon pattern from the substrate sapphire disc (SSD), manipulating exclusively the plastic ring and thus avoiding tweezers manipulations damages to the sample (Figure 1 G). Conveniently, the CryoCapsule fits in the Leica flow-through chambers so samples, once oriented, can be deposited at the bottom and embedded in a similar manner to other HPF samples.

In comparison, the sapphire assembly (Figure 1 C) tends to dismantle in the freeze substitution solution (Figure 1 H), sometime damaging the specimen and adding complexity to the sample orientation for embedding.

Freeze substitution and sectioning

Freeze-substitutions were performed in an AFS-2 Leica Microsystem. Samples were kept for 12 hours at −90°C in the FS solutions (according to each protocol), the temperature was then raised by 5°C per hour to −45°C, rinsed in acetone and embedded in 25%, 50%, 75%, 100% Lowicryl HM20/acetone in 2 hours steps. A second 100% Lowicryl embedding for 12 hours was pursued before UV polymerization for 48 hours. The temperature was raised 5°C per hours until +20°C and another 48 hours polymerization was conducted under UV. Sectioning was performed on a Leica ultracut UCT and 70 nm sections were collected in Cu/Pd slot grids coated with pioloform for ultra-structural observations. 250 nm sections were collected similarly for tomography purpose.

Fluorescence microscopy

CLEM Live cell imaging was performed on a Nikon Te2000u, with 10x ELWD plan apo (NA 0.28; WD 33.5), or 40x ELWD S plan Fluor (NA 0.6; WD 3.5) mounted on a piezzo-electric, phase Ph2, GFP or Cy3 dichroic filter.

Fluorescence on sections

Lowicryl sections were imaged in distilled water on a glass slide for fluorescence imaging on the same microscope with 40x ELWD S plan Fluor (NA 0.6; WD 3.5) mounted on a piezzo-electric, phase Ph2, GFP dichroic filter for the M10–11G experiments, and with a 60x 1.4 Oil DIC N2, mounted on a piezzo-electric, phase Ph2, Cy3 dichroic filter for the MNT-1 experiment.

Z-stacks of 20μm with steps of 1.4μm were acquired.

Electron microscopy imaging

After fluorescence on Lowicryl sections was performed, post-contrasting 3min of 4% uranyl acetate in H₂O plus 2min on lead citrate was performed on all slot grids.

Electron micrographs of the HPM010 cryo-immobilized specimens were acquired on a Biotwin CM120 Philips - equipped with a bottom-mounted 1K CCD Camera (Keen View, Olympus, Münster, Germany). Electron micrographs of the HPM100 cryo-immobilized specimens were acquired on a CM120 Philips equipped with a side-mounted 11 MegaPixels CCD camera (Morada, Olympus, Münster, Germany) for the M10–11G experiments. For the MNT-1 experiments, electron micrographs were acquired on a Tecnai Spirit equipped with a 4k camera CCD camera (Quemesa, Olympus, Münster, Germany)

Scanning electron microscopy

Samples were directly mounted on stubs and then gold-sputtered (Scancoat Six, Edwards). Observation was carried out with a conventional SEM operating at 15kV (Cambridge Stereoscan S260).

Electron tomography

Tilted microscopy and electron tomography were done on a Tecnai20 equipped with a bottom mounted TemCam-T416 camera (TVIPS, Gauting, Germany). Images from −65 to +65 degree were acquired using TVIPS EM-MENU software and reconstruction was done using eTomo [29].

FM/EM overlay, signal registration

Fluorescence registration was done in FIJI using TurboReg [28]. The transmitted light Z stack is targeted to the low magnification electron micrograph. The saved transform is then applied to the maximum intensity Z-projection of the fluorescent channel with background removal. After color merging of the electron micrograph and the fluorescent overlay, manual adjustment of the fluorescent signal is adapted for visualization purposes. The macro is available if requested by e-mail.

Serial Sections alignment

The CLEM overlays were grouped into a stack (EM/LM). Both channels were split, the serial EM sections manually aligned using MIDAS from the Imod package and the transformations were applied back to the LM stack. Both stacks were then merged back in FIJI.

Supplementary Material

Supp Figure S1-S3

Supplemental figure 1. No pressure is applied to the specimen prior to the HPF shot using the CryoCapsule as compared to other CLEM approaches. Vitrification simulation with conventional tools: A- In vitro X. laevis mitotic spindle structure deposited between substrate and covering sapphire discs (SSD, CSD) separated by a gold spacer ring (GSR) (see also fig. 1 C). It is located near the X negative carbon landmark (white arrow); Cy3-Tubulin in red, Chromosome mass labeled with Hoechst. B- The assembly is loaded in the conventional HPM010 clamp between two aluminum carriers and HPF is simulated (fig. 1 F). C- The sapphire assembly is recovered and re-imaged by LM. The chromosome mass (blue) has shifted, lost its structure (from a plate to a rounded shape) and the mitotic spindle structure of the Cy3-tubulin can no longer be seen (white arrow); all the fluorescent channels are electronically saturated for visualization purpose. Vitrification simulation with the CryoCapsule: D- Mitotic multipolar spindle from X. laevis trapped in a CryoCapsule and localized close to the positive carbon landmark (white arrow, N letter in carbon partially degraded). The chromosome mass was not labeled. E- Transfer of the CryoCapsule in the HPM010 tip and simulation of HPF. F- The CryoCapsule is recovered and re-imaged by LM. The displacement observed corresponds to the natural movements of mitotic spindles. The structure is not altered nor dismantled. Scale bars 50 μm.

Supplemental figure 2 Specific adaptors have been designed to use the CryoCapsule with the HPM010 and the HPM100. A- The HPM010 adaptor: the CryoCapsule is deposited upside down, the covering sapphire disc facing the 200$m aluminum support membrane. The revolving part of the HPM010 clamp is folded and maintains the CryoCapsule by the plastic ring. The substrate sapphire disc faces the opened side of the clamp for a direct liquid nitrogen contact upon HPF. Both sapphire discs are 50$m thick, but for visualization purpose, the covering sapphire disc is drawn thicker. On the side of the clamp, two LN2 flow-through channels are used to help cooling the periphery of the CryoCapsule with direct liquid nitrogen contact. B- The HPM100 adaptor: the CryoCapsule is clipped up-side up in the middle plate adaptor and can be directly imaged in this assembly using a long working distance objective. The covering sapphire disc is maintained in place by a ceiling lip that prevents movements and shear forces on the sample during the various HPF steps. Opened on both sides, the liquid nitrogen flows around the specimen and contact directly both sapphire discs to optimally drag the temperature of the sample down. The substrate sapphire disc is drawn thicker for visualization purpose.

Supplemental figure 3

Cooling rates of the different sub-elements usually used in high pressure freezing of cell monolayers. A-The commonly available sapphire discs for high pressure freezing are 50_m thick. Laser-etched sapphire discs are thicker (170_m). Thin sapphire disc has the best cooling rate of all. The aluminum carriers have a very high heat transfer (metal) and cool down very quickly. The water is the most thermo-resistant material and hence is the main limiting factor in vitrification. The heat resistance grows with the square of the thickness and should be one critical parameter to consider when proper vitrification should be achieved. Aclar®, often used to grow cells, is highly thermo-resistant. Thinness gives the vitrification properties. BCooling time calculation formulas: cooling time (θ=e2/a), were calculated using the constants presented here and the physical formulas of thermal diffusivity: a=(κ/ρ.Cp).

Supp Video S1. Supplementary video.

Live cell correlative microscopy workflow: M10 Melanoma cells expressing YFP-langerin where imaged for 5 minutes in the CryoCapsule every 5 seconds using fluorescent and transmitted light. Just after the last fluorescent image, the assembly, composed of the CryoCapsule and the Leica adaptor, was transferred into the HPM100 for vitrification. The samples were freeze substituted to preserve the fluorescence and serial sectioned. Each section was imaged by fluorescence and transmitted light microscopy and overlaid with the electron micrographs (430x) to map the fluorescence. Low magnification pictures were then aligned to map the fluorescence more precisely and identify the structures of interest. The live cell, fluorescence on section and EM are then aligned to illustrate the full correlative microscopy approach.

Supp Video S2

Video S2: Tutorial movie about the CryoCapsule manipulation. The covering sapphire disc in laid in the CryoCapsule in the culture petri dish. The assembly is blotted on a filter paper and clipped in the middle plate adaptor for the HPM100. The middle plate is loaded in an imaging chamber containing wet filter paper to keep the humidity high during live cell imaging. The CryoCapsule is centered above the hole of the imaging chamber to guarantee a direct light path for high quality data collection. After fluorescence live cell imaging, the sample is rapidly transferred into the HPM100 for HPF.

Abbreviations

EM

Electron Microscopy

LM

Light Microscopy

CLEM

Correlative Light to Electron Microscopy

CSD

Covering Sapphire Disc

SSD

Substrate Sapphire Disc

GSR

Gold Spacer Ring

OSER

Organized Smooth Endoplasmic Reticulum

HPF

High Pressure Freezing

HPM

High Pressure freezing Machine

PET

PolyEThylene polymer

RI

Refractive Index

FS

Freeze Substitution

FPs

Fluorescent Proteins

ELWD

Extra Long Working Distance