Correlative light and electron microscopy for the analysis of cell division



    1. Structural Cell Biology Group, Experimental Center, Medical Faculty Carl Gustav Carus, University of Technology Dresden, Dresden, Germany
    Search for more papers by this author

    Corresponding author
    • Structural Cell Biology Group, Experimental Center, Medical Faculty Carl Gustav Carus, University of Technology Dresden, Dresden, Germany
    Search for more papers by this author

Correspondence to: Thomas Müller-Reichert, Structural Cell Biology Group, Experimental Center, Medical Faculty Carl Gustav Carus, University of Technology Dresden, Dresden 01307, Germany. Tel: (+49) 351 458-6442; Fax: (+49) 351 458-6305; e-mail:


Correlative light and electron microscopy (CLEM) has recently gained increasing attention, because it enables the acquisition of dynamic as well as ultrastructural information about subcellular processes. It is the power of combining the two imaging modalities that gives additional information as compared to using the imaging techniques separately. Here, we briefly summarize two CLEM approaches for the analysis of cells in mitosis and cytokinesis.


The combination of light and electron microscopy is a very powerful tool to investigate biological processes. The term correlative light and electron microscopy (CLEM), however, is not a ‘protected term’ in the sense that it refers only to a single approach. The only ‘prerequisite’ for using this term is that the same biological specimen is analysed by two different imaging modalities (for a compilation of current CLEM techniques see: Müller-Reichert & Verkade, 2012). One major motivation to apply CLEM is to stage specimens by light microscopy for further ultrastructural analysis. In early CLEM studies, for instance, living sea urchin embryos and Haemanthus cells were observed during cell division by light microscopy and then fixed at mitotic stages using glutaraldehyde as a fixative for chemical cross-linking at room temperature (Brinkley et al., 1967; Sluder & Rieder, 1985). Further studies by electron microscopy then allowed the analysis of cellular fine structure. Along these lines, we briefly describe two CLEM approaches in the next paragraphs that have been applied to analyse previously staged Caenorhabditis elegans embryos in mitosis and HeLa cells in cytokinesis after high-pressure freezing.

The single-cell approach

The first example shows the advantages of our single-cell approach for CLEM (Müller-Reichert et al., 2007). We have used this approach to investigate centriole duplication in the early C. elegans embryo. C. elegans is a popular model system in modern cell biology, because hermaphrodites, dauer larvae, males, as well as early embryos are transparent, thus allowing the use of fluorescently tagged strains for live-cell imaging. In addition, the genome of this nematode is fully sequenced and annotated, and a genome-wide screen has led to a list of genes involved in various aspects of cell division (Sonnichsen et al., 2005). Last but not least, loss-of-function by RNA-mediated interference (RNAi; Fire et al., 1998), as well as gain-of-function approaches by codon optimization (Redemann et al., 2011) allow systematic structure–function studies by comparison of wild-type and mutant ultrastructure (Pelletier et al., 2006).

Fortunately, mitosis in the early C. elegans embryo can easily be followed by observation of the stereotypic pattern of movement of the pronuclei (Oegema & Hyman, 2006). The two pronuclei form, migrate and rotate within the early embryo in preparation to the assembly of the first bipolar mitotic spindle. The mitotic spindle itself is primarily organized from the two spindle poles (i.e. the centrosomes), which are composed of a central pair of centrioles surrounded by pericentriolar material (PCM). The centriole itself is of rather simple morphology with a central tube surrounded by nine singlet microtubules. The duplication of the centrioles is key to maintain the integrity of cell divisions. The question, however, arose how and when centriole duplication is achieved during the first round of mitosis.

To answer this question, it is crucial to stage embryos and to fix them at specific mitotic stages (Fig. 1). We achieve this by sucking embryos into transparent capillary tubing, which allows light microscopic observation and transfer of staged embryos to a high-pressure freezer for subsequent processing and ultrastructural analysis by electron tomography (Müller-Reichert et al., 2007). A thorough analysis of different time points in the course of the first mitotic division revealed that assembly of the daughter centriole begins with the formation (pronuclear appearance) and elongation of a central tube (pronuclear migration) followed by the peripheral assembly of nine singlet microtubules (pronuclear rotation). This wild-type study was complemented by analysis of mutant embryos lacking specific proteins required for the duplication process (Pelletier et al., 2006).

Figure 1.

Correlative light and electron microscopy reveals the course of centriole duplication in the early embryo of C. elegans. (A) Adult worms are sliced open to obtain embryos of different developmental stages. (B) An embryo of the desired developmental stage is sucked into a transparent capillary tube. (C–D) An embryo inside the capillary tube can be observed by light microscopy until high-pressure freezing. (E–F) Fluorescent images of C. elegans embryos labelled with ®-tubulin- and histone-GFP at different stages (E, metaphase; F, late anaphase) imaged inside capillary tubing (Müller-Reichert et al., 2007). (G) Centriole duplication in the C. elegans embryo during the first cell cycle. The central tube of the newly forming daughter centriole appears at the time of pronuclear (PN) appearance. This central tube is then elongated during PN migration. At the time of PN rotation, microtubules start to assemble around the central channel (Pelletier et al., 2006). Scale bars = 100 ìm (D), 10 ìm (F).

The same early embryo/CLEM approach can also be applied to analyse assembly of the mitotic spindle. The microtubule cytoskeleton has to undergo extreme reorganizations during the course of mitosis to form a bipolar spindle and to segregate chromatids. Within the spindle, the organization of microtubules is very defined. Microtubule minus-ends of microtubules are organized from microtubule organizing centres (MTOCs), known as spindle poles or centrosomes. The plus-ends of microtubules are radiating out from the MTOCs towards the cell cortex (astral microtubules), the aligned chromosomes (kinetochore microtubules), or making contacts to microtubules emanating from the opposite MTOCs (interdigitating microtubules). The dynamic properties of mitotic spindle assembly have been studied intensively by light microscopy and are well described. A detailed structural analysis of spindle organization, however, is essential for a complete understanding of the mechanisms underlying this spindle formation. We are in the process of generating such a three-dimensional (3-D) reconstruction of an entire mitotic spindle.

As mitosis is a very dynamic process, it is essential, to have the ability to collect samples for the electron microscope at precise stages of mitosis. This can be done with the same approach as described above. For this, embryos are collected in transparent tubes and observed by light microscopy until they reach the stage of interest (Fig. 2A). Thus, collecting in capillary tubing allows high-pressure freezing at a very precise time point. Embryos are then embedded in resin and sectioned, at which each single section containing the spindle has to be collected for further electron tomography. Montage tilt series that cover large areas of the spindle can be obtained using the SerialEM software (Fig. 2B) and 'supermontage' tomograms are computed and assembled using the IMOD software packages (Boulder Lab for 3D Electron Microscopy of Cells, Boulder, CO, USA; Kremer et al., 1996). Microtubules in the reconstructed tomograms can then be semi-automatically segmented using AMIRA (Weber et al., 2011). Compared to manual tracking, it is expected to obtain large-scale reconstructions of entire spindles within a much shorter period of time. Our goal is to get a 3-D model of an entire mitotic spindle (Fig. 2C). Information about the number of microtubules connected to the kinetochores, for instance, will be combined with data on microtubule dynamics as obtained by light microscopy. Such a combination should provide a detailed model of the mechanics of spindle assembly.

Figure 2.

Full 3-D reconstruction of a mitotic spindle in the C. elegans one-cell embryo using a correlative light and electron microscopy approach. (A) C.elegans embryo in anaphase labelled with ®-tubulin- and histone-GFP. An embryo is filmed until early anaphase and subsequently transferred to the high-pressure freezer. (B) Supermontage of tomograms of a C. elegans anaphase spindle. To acquire the entire anaphase spindle of the C. elegans embryo, montages of tomograms (here 2 × 3 tomograms depicted in yellow, orange and magenta) have to be combined to a supermontage (3 × 2 × 3). (C) Using the AMIRA microtubule-tracing tool, microtubules (green) can be semi-automatically traced within the tomograms of such supermontages. In order to obtain a 3-D volume of the entire spindle, several supermontages have to be joined in Z direction. Scale bars = 10 ìm (A), 1 ìm (B and C).

CLEM of HeLa cells grown on solid supports

The second example illustrates the use of CLEM for the study cells grown on a solid support. We used this approach to study the physical separation of forming postmitotic daughter cells. The last step of cytokinesis, that is, abscission, is currently a field of intensive research. How do cell pairs terminally split and what is the molecular mechanism of this separation? Again, a correlative approach is advantageous here to image cells during cytokinesis and to proof/disproof current models of abscission, namely the cortical constriction, membrane fusion or cell rupture model (Guizetti & Gerlich, 2010). Cultured HeLa cells are frequently used to study abscission. Abscission in HeLa cells occurs at ∼1-μm wide intercellular bridge connecting the postmitotic daughter cells. Terminal splitting of the daughter cells, however, often occurs long after ingression of the cleavage furrow. A method is, therefore, necessary to synchronize cells to this stage and to select cell pairs at intermediate stages of abscission. When grown on a solid support, culture cells can be used for live-cell imaging and subsequent EM analysis.

In principle, there are two approaches available. First, HeLa cells can be grown on an ACLAR support on which a grid pattern had been scratched (Fig. 3A). This grid pattern leaves a negative imprint on thin-layer embedded samples, allowing identification of cells selected from time-lapse imaging (Fig. 3B). Thin-layer embedding of chemically fixed cells then allows the ultrastructural analysis of specific stages of abscission by EM (Fig. 3C; Guizetti et al., 2010). As a side point, the use of ACLAR for high-pressure freezing had been developed by Jimenez and coworkers (Jimenez et al., 2006). Second, HeLa cells can be grown on sapphire discs as routinely used for high-pressure freezing. Instead of scratching a grid pattern to allow the identification of cell pairs, a pattern on the sapphire disc can be created by carbon evaporating (for details see: McDonald et al., 2010). The combination of CLEM and electron tomography has led to the visualization of ESCRT-III-dependent ∼17-nm wide filaments, which are thought to drive abscission through cortical constriction (Fig. 3D). The exact composition of these filaments, however, is currently unknown and needs further investigation (Guizetti et al., 2011).

Figure 3.

CLEM approach to investigate abscission in HeLa cells. (A) HeLa cells are grown on ACLAR with a scratched grid pattern. This grid pattern allows a relocation of cell pairs after thin-layer embedding. (B) Cells shown in the white boxes in (A) are continuously filmed by time-lapse imaging. (C) Intercellular bridge of a postmitotic cell pair, chemically fixed for electron microscopy. Abscission takes place in the neck region of the intercellular bridge as indicated by the insert (Guizetti et al., 2010). (D) Three-dimensional model of the neck region during abscission. Microtubules are shown in red; filaments that assemble in the neck region in dependency of ESCRT-III proteins are shown in blue and green. HeLa cells for electron tomography are grown on sapphire discs for high-pressure freezing (Guizetti et al., 2011). Scale bars = 0.5 mm (A), 0.5 ìm (C), 0.2 ìm (D).

The next step: labelling structures of interest in 3-D

In addition to a comparative analysis of wild-type and mutant ultrastructure, it is often of interest to study the precise subcellular location of specific proteins, preferably in three-dimensions. To obtain such 3-D information, gold-conjugated antibodies are currently introduced to the specimen prior to resin embedding. However, in order to get ultrasmall gold into the specimen, large ‘holes’ have to be created in the cell membranes through treatment with detergents or solvents. This can destroy fine cellular structures or in the worst case, result in an extraction of the antigen. Alternatively, one would like to take advantage of tags, which on one site are fluorescent and on the other site can be converted into a visible electron-dense signal.

A recent example of such a tag is ascorbate peroxidase (APEX). APEX can be genetically fused to a target protein, is visible by fluorescent light microscopy and catalyses the oxidative polymerization of diaminobenzidine (DAB) upon addition of H2O2. Using this tag, contrast is generated by staining of the DAB polymer with osmium tetroxide (Martell et al., 2012), similar to other methods using benzidine substrates (Grabenbauer et al., 2005; Shu et al., 2011). However, this method is expected to give a rather diffuse precipitate that can be used best when labelling proteins in chemically fixed cellular compartments, such as the Golgi apparatus or mitochondria. It will be interesting to test whether this tag will give sufficient electron-dense signal for a precise and unambiguous three-dimensional localization of centrosomal proteins, kinetochore components or abscission structures.


The authors would like to thank Drs. Eileen O'Toole and Kent McDonald for a critical reading of the manuscript. Research in the Müller-Reichert lab is supported by grants from the Deutsche Forschungsgemeinschaft (DFG: MU 1423/3–1 and MU 1423/4–1), the Human Frontier Science Program (RGP 0034/2010) and the SMWK (Sächsisches Ministerium für Wissenschaft und Kunst).