Abstract
Correlative imaging of a specific cell with both the light microscope and the electron microscope has proved to be a difficult task, requiring enormous amounts of patience and technical skill. We describe a technique with a high rate of success, which can be used to identify a particular cell in the light microscope and then to embed and thin-section it for electron microscopy. The technique also includes a method to obtain many uninterrupted, thin serial sections for imaging by conventional or energy-filtered transmission electron microscopy, to obtain images for 3D analysis of detail at the suborganelle level.
Keywords
O
Attempts to locate specific factors within the ultra-structure of a cellular organelle have relied extensively on gold-labeled immunological reagents used in pre-or postembedding techniques. Finding rare features with immuno-EM techniques alone can be difficult if not impossible. Correlative light and electron microscopy is essential to make the connection between the spatial map obtainable using fluorescence microscopy and the ultrastructural basis for the organization of the map. By doing so, correlative microscopy techniques provide the advantage of narrowing in on a rare structural feature or an event that occurs in a small temporal window in the cell cycle, or even after stimulation by a chemical or environmental signal. Because most of the information used to describe recent structure-function relationships has come from fluorescence microscopy, and the added advantage that specific proteins involved in cellular events can be followed in living specimens by fluorescence microscopy, the merging of this microscopy technique with electron microscopy is particularly valuable.
The potential of correlative light and electron microscopy is well recognized and has been applied in a number of studies to determine the structures that underlie a particular cellular behavior (Sluder and Rieder 1985; reviewed in Rieder and Cassels 1999), and has been combined with fluorescent probes and immuno-labeling methods (Mitchison et al. 1986; Robinett et al. 1996). We describe a technique here that makes significant improvements over previous approaches, so that the technique can quickly become a routine method in any electron microscopy laboratory. Some of the improvements that we have incorporated include the use of 50-mesh grids to locate cells of interest that are growing on a coverslip, and the use of these grids to identify the cells after embedding in resin. We also introduce the use of Quetol resin for embedding because of its low degree of autofluorescence and properties that ensure long continuous ribbons of serial sections. Finally, our techniques are optimized for energy-filtered transmission electron microscopy and electron spectroscopic imaging (ESI), so that elemental maps can be obtained. These permit delineation of specific biochemical features and provide an opportunity for the development of metal-tagged probes for mapping proteins and other features with both high resolution and high degrees of sensitivity (Malecki et al. 2002).
Materials and Methods
Immunofluorescence Labeling
SK-N-SH neuroblastoma cells were cultured on glass cover-slips (18 × 18 mm) under the recommended conditions (American Type Culture Collection; Manassas, VA). Cells were fixed with fresh 1% paraformaldehyde in PBS (pH 7.5) for 5 min at room temperature (RT) and rinsed twice with PBS. Cells were then permeabilized in PBS containing 0.5% Triton X-100 for 5 min at RT, and rinsed twice with PBS. Cells on coverslips were labeled with an anti-PML protein (promyelocytic leukemia) monoclonal antibody (5E10; Stuurman et al. 1992) (gift from Dr. R. van Driel), for 1 hr at RT (Boisvert et al. 2000). The secondary antibody labeling was carried out with a goat anti-mouse Cy3 antibody (Chemicon; Temecula, CA) for 1 hr at RT. The coverslips were mounted on slides with a glycerol-based mounting medium, 90% (v/v) glycerol in 1 × PBS, sealed with transparent nailpolish. The sealant is to avoid contamination of the mounting medium with cleaning solutions used later.
Identification of the Cell of Interest
Fluorescence imaging was performed on an upright epifluorescence DMRA2 microscope (Leica Microsystems; Richmond Hill, ONT, Canada) using filters optimized for Cy3 (Molecular Probes; Eugene, OR). After immunolabeling, cells were examined using a high-magnification oil-immersion objective lens (X63, 1.4 NA) and imaged using a 12-bit cooled CCD camera (Hamamatsu Orca-ER; Hamamatsu, Japan). When a cell or region of cells of interest had been imaged at high magnification, the region was centered with the aid of an eyepiece reticule and both examined and imaged using a low-magnification objective lens (X10). Despite the presence of aberrations due to imaging cells through the added layer of immersion oil using a low-magnification air lens, an acceptable image map of the distribution of the cells of interest was obtained. After an image record of the cells had been obtained, the immersion oil was washed from the coverslip with a cotton tip, and a 50-mesh copper grid (Electron Microscopy Sciences; Fort Washington, PA) was positioned over the region of interest (Figure 1A). The slide was then removed gently from the microscope and the edges of the grid were fixed to the coverslip with transparent tape. The slide was returned to the microscope and examined using the low-magnification lens to re-confirm that the region of interest was represented, and the precise location with respect to the 16 squares of the grid was noted. Other regions of interest within the area of the first grid could also be identified, or other grids could be used in different regions of the coverslip. The recorded fluorescence images were processed for future reference (see below).
Embedding and Relocating the Cell of Interest
Another grid was attached to the stage of a dissecting microscope with transparent tape. This was aligned with the grid on the top face of the coverslip. The slide was then fixed to the stage of the dissecting microscope with tape. The grid on the coverslip was removed and the coverslip was washed with ethanol. A new grid was placed in the same registration and fixed with transparent tape. This step fixes the grid to the coverslip so that it will not wash off during the dehydration and embedding steps that follow. The coverslip was removed gently from the slide and placed in a 40-mm Petri dish. The coverslip was washed three times each for 60 sec in 1 × PBS to remove the mounting medium. The cells were postfixed with 2% glutaraldehyde in PBS for 5 min at RT. After the postfixation step, the cells were dehydrated with a series of graded ethanol steps of 30, 50, 70, and 95%, with incubations on a shaker for 60 min at each step. One or 2 ml of resin, Quetol 651 (Electron Microscopy Sciences) was then placed on the coverslip and incubated with shaking for 2–3 hr. The resin was removed and replaced with Quetol mix (Quetol 651-NSA Kit) and incubated with shaking for 2-3 hr. This step was repeated. The coverslip was then placed in a Petri dish (Permanox; Nalge Nunc International, Rochester, NY) or a glass Petri dish and covered with a thin layer, 0.5-2 mm, of Quetol mix, followed by polymerization by incubating at 65C for 24 hr (Figure 1B). A thin layer of Quetol mix is preferable if low fluorescence signal is expected.
After polymerization, a second grid was glued onto the resin, with the aid of the dissecting microscope, directly above and in register with the original grid on the coverslip. The Petri dish was placed on a hotplate at approximately 70 EC for a few minutes. When the resin became soft, it was cut around the edges of the coverslip, just within the perimeter of the coverslip area. The resin was then carefully peeled off the coverslip (Figure 1C). The block was then placed on a glass slide and the edges taped down with transparent tape. The region of interest was again checked using a low-magnification lens with the fluorescence microscope. A third grid in perfect register with the second grid was attached to the cell side of the Quetol block with transparent tape (Figure 1D). The point of a sharp scalpel was used to mark the edges of the region to be trimmed, and the region was also marked with a felt-tip pen. The grid on the cell side was then removed, the block was again placed on a hotplate at approximately 70 EC. A large selected area was cut with a razor blade on the hotplate (Figures 1E and 1F). The cut fragment was glued onto the bullet and mounted in the ultramicrotome. If the block surface was lightly wiped with 95% ethanol, the cells could be seen on the surface of the trimmed block with the binoculars of the ultramicrotome. The region of interest could be compared with images captured earlier with the fluorescence microscope, so that the cell of interest could be kept in the center of a keystone formed by trimming with a scalpel (Figure 1F). The shape of the keystone had a high length:width ratio (Figure 1F), which permitted long ribbons of serial sections to remain intact as they came off the diamond knife during the sectioning step with the ultramicrotome (Figure 1F).
Schematic diagram outlining the major preparation steps to identify a region of interest by correlative fluorescence and energy-filtering transmission electron microscopy. Cells of interest are first identified by fluorescence microscopy. (
Serial Sectioning
The region of interest could be seen in the trimmed block with the Ultracut UCT ultramicrotome (Leica Microsystems). When a ribbon of 5–15 serial sections, generally in the 60-90-nm thickness range, had been obtained on the surface of the water, the ribbon could be manipulated so that the sections were perpendicular to the 400-mesh parallel grid bars. As the ribbon was picked up, the region of interest was carefully positioned so that it was located over the grid openings and not on a grid bar.
Correlative Fluorescence and Electron Imaging of Serial Sections
The grid with the serial sections was then placed on a glass slide, covered with a glass coverslip, which was taped down with transparent tape, and imaged with the fluorescence microscope at both low magnification, to determine the relative orientation and position of the sections with respect to the grid center, and at high magnification (X63, NA 1.4) to obtain the highest-resolution spatial map of fluorescence labeling present within the sectioned specimen that the light microscope could provide with the given specimen preparation. Because the sections were typically in the 60–90-nm range of thickness, the images were truly confocal. The absence of out-of-focus haze permitted very high resolution and signal-to-noise not seen even with the laser scanning confocal microscope or with digital deconvolution of serial optical sections. To provide stability to the sections under the sometimes intense electron beam conditions required for elemental mapping, the sections on a grid were used to pick up a thin (4 nm) carbon film. The carbon films were prepared by electron beam evaporation onto mica. The films were floated onto water and picked up with the resin sections. Electron microscopy of the regions of interest was then carried out with a Tecnai 20 (FEI; Eindhoven, The Netherlands) transmission electron microscope equipped with an electron imaging spectrometer (Gatan; Pleasanton, CA), and operated at 200 kV. The technique of electron spectroscopic imaging (ESI) has been described elsewhere (Bazett-Jones et al. 1999). Serial sections were first stabilized to the beam by imaging with a low flux of electrons at a low magnification. Net phosphorus ratio maps were produced from pre- and post-edge images recorded at 120 and 155 eV (LII,III edge), and net nitrogen ratio maps were produced from pre- and post-edge images recorded at 385 and 415 eV (Nk edge). Elemental maps were generated by dividing the element-enhanced post-edge image by the pre-edge image. The recording times required to obtain the pre-edge and post-edge images are in the range of 10-30 sec, and result in electron exposures of the specimen to approximately 105 electrons/ nm2.
Image Processing
Fluorescence images were processed using Adobe Photoshop v 5.0 and ESI images were processed using Digital Micrograph v. 2.5. The fluorescence images were re-sampled based on differences in magnification and pixel size so that a comparison between the fluorescence spatial map and the EM image could be made. Rotational differences between the fluorescence and EM images were also taken into consideration. This was necessary so that the region of interest in the low-magnification electron microscope image could be found at the high magnification, with certainty that the same region was being examined at high magnification.
Results
The ability to track a particular cell of interest is essential for a correlative light and electron microscopy technique to be successful. When the in situ distribution of a protein is to be examined in fixed, immuno-labeled cells, in this case for the PML protein, cells were examined first at high magnification to identify a cell of interest (Figure 2B) and then at low magnification to identify a region of interest (Figure 2A). The same regions were re-imaged after embedding in resin but before thin sectioning (Figures 2C and 2D). The strong fluorescence signal present in the postembedded sample indicates that the fluorophore Cy3 is stable under the dehydration and resin polymerization steps. In both the pre- and postembedded images, the seven strongly labeled structures represent PML bodies, subnuclear domains that accumulate PML protein. The nuclei of SK-N-SH cells are quite flat (∼5 μm in thickness), and therefore most of the PML bodies are restricted to a relatively thin plane within the nucleus. The maintenance of the fluorescence signal after embedding provides the opportunity to monitor regions of interest after the embedding step and before sectioning, a valuable tool that enables examination of specimen integrity before investing time in further preparation steps.
Pre- and postembedding fluorescence images of cells immunolabeled with an antibody directed against PML protein. (
The next step is to prepare a series of thin sections by ultramicrotomy. After preparing a ribbon of serial thin sections, the ribbon was picked up on a 400-mesh grid with only horizontal grid bars. If the ribbon is maintained parallel with the grid bars, but with the individual sections perpendicular to the grid bars, a cell of interest can be visualized in several continuous sections. A low-magnification fluorescence image of such a ribbon is shown in Figure 3. Two individual cells, one in each of two lanes (arrows), are represented in 15 consecutive sections. The ability to monitor both the position of the cell of interest within the section and the section ribbon on the grid by fluorescence microscopy is key to determining whether one should pursue collecting multiple EM images of that particular sample for future high-resolution 3D analysis. Long ribbons of continuous sections can be routinely cut. At least six serial sections of a cell of interest can be obtained more than 80% of the time. Ribbons of at least 15 consecutive sections at an 80% success rate can be achieved with coarser mesh grids. These sections, however, are less stable in the beam unless stabilized by picking up on formvar films, which would interfere with an attempt to obtain elemental maps by ESI.
High-resolution fluorescence images are needed to help pinpoint and correlate the distribution of subcellular or, in the case of PML bodies, suborganellar structures observed using the light microscope with the ultrastructure observed using the electron microscope. Consequently, serial sections of the cell of interest identified in Figure 2 were imaged at high magnification (Figure 4B). Each section contains only a portion of one of the seven PML bodies. Because the sections were typically in the 70-nm range of thickness, the images were truly confocal. The absence of out-of-focus haze permitted very high resolution and signal-to-noise not seen even with the laser scanning confocal microscope or with digital deconvolution of serial optical sections (Pombo et al. 1999; Robinson et al. 2001). There is, however, some degradation in lateral resolution due to introduction of spherical aberration caused by refractive index mismatch. Nevertheless, fluorescence imaging of the thin sections provides an excellent map to find the structure of interest in the electron microscope.
A series of ultrathin sections containing cells of interest. Shown is a ribbon of 70-nm-thick serial sections picked up on an EM grid of 400-mesh horizontal bars. The individual sections are perpendicular to the horizontal bars, such that cells of interest in the serial sections fall in the spaces between the grid bars. Two cells of interest in two adjacent spaces are shown with arrows. Each cell is represented in 15 serial sections in this field. Two horizontal grid bars are identified by pairs of vertical arrowheads.
Next, the sections are examined using a transmission electron microscope, by conventional means or in energy-filtering mode, depending on whether the samples were stained for TEM after the high-magnification fluorescence images were obtained. Here we show imaging of unstained sections using energy-filtering transmission electron microscopy. A low-magnification electron micrograph of each section was first recorded at 155 eV energy loss (Figure 4A). The low-magnification EM image provides a spatial map of cell ultra-structure that is used to correlate with the fluorescence map to pinpoint the structure of interest. The fluorescence microscopy image is superimposed on the low-magnification energy-filtered TEM image to identify the PML body in the electron micrograph. A number of image processing steps are required to overlay the light and EM images (see Materials and Methods, and Figure 5). In addition, the intensity of the fluorescence signal from the PML bodies provides a good indication of whether a particular PML body is being sampled in the section, and whether a grazing section or a section through the body's interior is being examined. The diameter of PML bodies is quite variable (200–800 nm), such that some are represented in several 70-nm-thick sections whereas others are represented in only two or three 70-nm-thick sections. If a cell represented in the entire series of sections did not overlap with a grid bar, serial sections of every PML body in the cell can be obtained. This is very desirable for attempts to reconstruct the 3D structure of the bodies. To reduce the chance of losing a cell of interest by being situated on a grid bar, a coarser mesh grid could be used. Although the unsupported sections are less stable on such grids, they can be stabilized with form-var if conventional TEM is going to be used, or stabilized with a 3-nm carbon film if ESI is to be used. We routinely produce such films by electron beam evaporation onto mica (Bazett-Jones et al. 1999). The films can be floated on water and picked up with the serial section ribbon after it has been picked up with the 400-mesh horizontal grid. We have found that stabilization of the sections with a 3-nm carbon film prevents distortion in the shape of the sections when they are irradiated with the high beam intensities required for ESI. If the sections change shape, the structures will be distorted and the EM images will not superimpose over large fields onto the fluorescence images for identification of structures of interest. An example of this type of distortion is observed by comparing panels 2, 3 and 4 in Figure 4A. The nucleus in panel 3 is distorted, seemingly compressed in the y-axis and extended in the x-axis, compared to the nucleus in panels 2 and 4. The distortion in panel 3 is particularly prevalent due to the irregularity in that particular section from an unfortunate deposit of dirt. Therefore, the use of a stabilizing film is recommended. Overall, many undistorted serial sections were available that covered the depth of a number of PML bodies, sufficient to analyze the 3D ultrastructure of these subnuclear bodies when imaged at high magnification using the electron microscope.
The fluorescence microscopy image was superimposed on the low-magnification energy-filtered TEM image to identify the PML body in the electron micrographs (Figure 5). Obtaining a high-magnification/ high-resolution electron micrograph of the structure of interest is the principal reason for employing correlative microscopy. Energy-filtered transmission electron microscopy and, in particular, ESI has the added benefit of examining elemental distributions along with structural features. Employing ESI enabled us to generate high-resolution phosphorus and nitrogen maps, which reveal that the PML body contains a phosphorus-depleted, nitrogen-rich core structure (Figure 5; compare 5C with 5D and 5E with 5F). The comparison of phosphorus and nitrogen maps indicates that the core of the structure is primarily protein-based and not nucleic acid-based. Adjacent serial sections imaged in this way can be used to reconstruct the protein-based substructure of the body (data not shown) and to relate the structure of the core to other protein-based components in the surrounding nucleoplasm (Boisvert et al. 2000). Likewise, the path of chromatin fibers found at the periphery and that surround PML bodies could be reconstructed in 3D to provide insight into the local ultrastructural environment in which PML bodies are found.
Fluorescence and electron spectroscopic images of a cell of interest after sectioning. (
Discussion
The technique described here provides significant improvements over previous approaches to correlative microscopy. First, the use of 50-mesh grids at various stages of the procedure provides a means to readily identify a particular cell of interest through all of the preparation steps, from first identifying a cell on a coverslip to the final trimming of the block on the ultramicrotome. The ability to easily observe a particular cell through the entire process accounts for the very high success rate. The use of such grids is more accurate than scoring the glass or resin surfaces with diamond pencils or other markers. Second, the use of Quetol resin had the advantage that Cyanine and Alexa fluorophores are stable under the polymerization conditions. Moreover, the degree of background-contributing autofluorescence of Quetol is relatively low, and is able to limit the thickness of the resin introduced onto the cell layer, so that background from the resin is even further minimized. Another advantage of Quetol resin is the ease with which serial sections can be cut. Long continuous ribbons of serial sections of a desired thickness chosen from the range of 30 nm to over 100 nm can be routinely obtained. This is dependent, however, on another technical introduction, that the blocks are cut with a high length:width ratio. If the ribbons are picked up on grids with longitudinal bars, serial sections of the same cell can be obtained without the cell becoming obscured on grid bars. Finally, by combining fluorescence microscopy with ESI, we have eliminated the need for heavy atom contrast agents which limit resolution and can lead to over- or under-representation of particular features due to biochemical-specific interactions of stain with the specimen.
ESI images of a PML body in the cell of interest. (
The technique described here readily permits the identification of structures of interest in situ in fluorescently-labeled cells but is also applicable to live cell imaging using the fluorescence microscope where cells are eventually fixed and embedded using the preparation described here. Because serial sections of specific regions of interest can also be obtained, structures that are rare in the cell or are temporally restricted can easily be detected, even in single sections, after embedding and ultramicrotomy. Therefore, a major advantage of the technique is that such structures can be found very quickly in the TEM; large areas and large numbers of sections do not have to be screened with the TEM. In addition, we described a method for producing and picking up long ribbons of serial sections. The imaging of structures represented in these sections can be used to reconstruct 3D maps. The ability to reliably prepare such long series of sections on a grid with minimal obstruction from grid bars is key to performing 3D ultrastructural analysis of rare structural features in a statistically meaningful manner. We hope to extend the technique by combining the information generated using this correlative microscopy approach with tomographic data sets from energy-filtered images. Ultimately, we would like to provide very high-resolution information in three dimensions of the relationships between protein-based and nucleic acid-based structures in the cell nucleus.
Footnotes
Acknowledgements
Supported by operating grants from the Canadian Institutes of Health Research and with funds from the Cancer Research Society, Inc. D.P.B.-J. is the recipient of a Canada Research Chair in Molecular and Cellular Imaging.
We thank Dr. Roel van Driel for the gift of the 5E10 PML monoclonal antibody.