Correlative light and electron microscopy approaches can reveal the localisation of specific proteins while providing detailed information on the cellular context, thereby combining the strengths of both imaging modalities. The major challenge in combining fluorescence microscopy with electron microscopy is the different sample preparation requirements necessary for obtaining high quality data from both modalities. To overcome this limitation, we combined conventional sample preparation protocols for electron microscopy with post-embedding labelling on ultra-thin sections using antibodies and other specific ligands. We successfully employed STED super-resolution microscopy to image the subcellular distribution of several targets in various specimen including E. coli, T. brucei, S. cerevisiae, human cancer cells and bovine sperm. Thus, we present widely applicable methods facilitating the use of antibodies for correlative super-resolution light and electron microscopy of post-embedding labelled targets.
Electron microscopy requires elaborate and time consuming sample preparation, while providing very high-resolution information about subcellular structures. However, the possibilities to specifically and sensitively label and localise proteins of interest are rather limited. Fluorescence light microscopy enables the investigation of distributions of multiple targets with high molecular specificity. Furthermore, it requires much less elaborate sample preparation protocols and can be applied in living cells. A major drawback of fluorescence microscopy is the limited resolution of conventional light microscopes. This drawback has been rendered obsolete by super-resolution light microscopy techniques which overcome the diffraction limit fundamentally . One shortcoming of fluorescence microscopy remains, however: the lack of contextual information, such as membranes, since only labelled structures are visible in the fluorescence image. This issue can be addressed by correlative light and electron microscopic (CLEM) approaches.
CLEM has been used to find rare events by using the fluorescence signal to identify these events and guiding the EM for detailed structural analysis [2, 3]. Likewise, CLEM has been used to combine information about specific protein localisations using light microscopy with the contextual information about surrounding cellular membranes and structures using electron microscopy [4–6]. This has become even more important in the light of the new diffraction unlimited super-resolution techniques which enable the localisation of fluorescent markers with molecular precision, thereby closing the gap between light and electron microscopic resolution.
Several correlative super-resolution light and electron microscopy approaches have been implemented since their first descriptions in 2006  and 2011 . Most of the studies rely on pre-embedding labelling of the proteins of interest. This means that the labelling of the samples for fluorescence microscopy is carried out before the sample is processed for electron microscopy, e.g. embedded in resin [9, 10]. Permeabilisation of the cell, which is required for intracellular labelling with antibodies or many other specific ligands, degrades membrane structures and is thereby detrimental for electron microscopic analyses. Hence, most CLEM studies employed fluorescent proteins or chemical tags. Still, the use of fluorescent proteins or chemical tags, also comprises certain challenges. These challenges include potential protein mislocalisation by overexpression artefacts and changes of the photophysical properties of the fluorescent label by the sample preparation protocols necessary for EM. Although these potential problems can be addressed using endogenous genomic tagging and clever selection of the fluorescent label combined with sophisticated sample preparation protocols [11–20], the preparation and handling becomes increasingly complex. Accordingly, compromises between sample quality for light microscopy and for electron microscopy have to be made.
In this study, we combined several standard EM fixation and resin embedding techniques facilitating good structural preservation and contrast with post-embedding immuno-labelling protocols allowing the utilisation of antibodies and other specific ligands. Post-embedding has the additional benefit that it results in the highest possible axial resolution, since just the surface of the (ultra-)thin section is labelled. We confirmed that compared to labelling with colloidal gold, the fluorescence labelling yielded a much higher labelling density. The on-section fluorescence labelling allowed us to perform EM and STED super-resolution microscopy on the very same samples, providing high-resolution information about specific protein localisations as well as the cellular ultrastructure, and finally enabled us to perform true correlative super-resolution light and electron microscopy.
Results and discussion
Immuno-labelling and STED super-resolution microscopy on resin sections
The use of light microscopes which can overcome the diffraction barrier of light microscopy demands highest quality of labelling and good structural preservation of the samples. Requirements for suitable samples are: (1) a proper fixation of the biological specimen that sustains a high structural preservation without inhibiting subsequent labelling steps or introducing unspecific (auto-)fluorescence. In case of fixation of cells or tissues that express FP-tagged proteins or similar, the fixation must retain the properties of the fluorophore. (2) When immunolabelling is used to label intracellular targets, usually permeabilisation of the membranes is necessary to facilitate accessibility of these targets. (3) Irrespective of the labelling strategy, a high labelling coverage and a highly specific labelling is required. (4) The photophysical properties of the labels, which have to meet the requirements of the respective super-resolution microscopy method and strongly depend on the environment of the dyes (e.g. the mounting medium, the buffer or, in case of CLEM, the resin), have to be considered as well.
These requirements are very challenging to reconcile with subsequent typical sample preparation procedures for electron microscopy. We decided to rely first on a post-embedding labelling strategy [21, 22] which consists of chemical fixation of the cells, progressive lowering of temperature (PLT), dehydration and embedding in a Lowicryl resin (HM20) [23, 24], followed by polymerisation and ultrathin sectioning. The immunolabelling was then performed (post-embedding) on the plastic sections (figure 1(a)). In a first attempt, we processed the parasite Trypanosoma brucei and labelled resin sections with an antibody against alpha-Tubulin (figure 1). In contrast to a pre-embedding approach, which would have required a detergent treatment leading to the destruction of the membranes, our approach did not affect the membrane integrity in the sample. Diffraction-limited confocal and diffraction-unlimited STED microscopy reveal a high labelling efficiency on the resin sections (figures 1(b) and (c)). In the STED images, individual microtubules in the flagella were resolved and orientations from transverse to longitudinal sections of the flagella could be observed (figure 1(c)). These sections through the flagella of T. brucei demonstrate another benefit of recording post-embedding labelled resin sections, namely the superior axial resolution, resulting from the fact that only the surface of the resin-section is labelled.
In all STED images throughout this study, we used a vortex phase plate for shaping the STED beam, which provides the best lateral resolution but does not improve the axial resolution. With this imaging modality, the optical z-resolution is diffraction-limited (500 nm). However, with post-embedding labelled sections the effective z-resolution is determined by the labelling itself. Post-embedding labelling on the resin sections results in a strict surface labelling due to the fact that only the epitopes on the very surface of the sections are accessible for the antibodies. This makes it possible to resolve the individual microtubules even in more inclined cross-sections through the flagella, which otherwise would have been blurred by the confocal z-resolution of the microscope (figure 1(c)).
We conclude that post-embedding labelling of alpha-Tubulin in Lowicryl-embedded T. brucei is highly compatible with super-resolution STED microscopy. This finding raises the question if post-embedding labelling strategies can be used in diverse types of samples, facilitating a broader applicability of this approach.
STED microscopy and TEM on various specimen
Since we were able to image tubulin on ultrathin Lowicryl-sections of the parasite T. brucei using STED, we next aimed to use post-embedding labelling strategies to label several targets in various specimen. To this end we used chemical fixed Escherichia coli cells and stained, after embedding in Lowicryl resin and ultrathin sectioning, for the bacterial outer membrane protein OmpA (figure 2(a)). This resulted in a highly specific membrane staining of these bacteria. The resolution enhancement in the STED images was clearly visible.
However, for optimal ultrastructural preservation especially of larger volumes and specimen with thick cell walls that are difficult to pass for chemical fixative, high-pressure freezing (HPF) in combination with low temperature embedding is the best method of choice [25–28]. To establish post-embedding labelling with HPF and low temperature embedding, we used HeLa cells, bovine sperm and the yeast Saccharomyces cerevisiae, employed HPF to cryofix the specimen and stained them after freeze substitution and sectioning for several targets (figures 2(b)–(e)). We were able to successfully label tubulin and DNA in the HeLa cells (figures 2(b) and (c)), tubulin in the bovine sperm (figure 2(d)) and, using an anti-GFP antibody, GFP-tagged histone Htb1 in yeast cells (figure 2(e)). Performing STED on these samples, we could resolve microtubules of the spindle apparatus of a HeLa cell during mitosis (figure 2(b)), the structure of DNA patches in the nucleus of a HeLa cell (figure 2(c)), microtubules in the tail of bovine sperm (figure 2(d)), and the structure of histone Htb1 patches in the nucleus of yeast cells (figure 2(e)). Importantly, all of the samples possess a high degree of structural preservation as exemplified in the TEM images of these samples (supplementary figure 1 (stacks.iop.org/JPhysD/52/374003/mmedia)). We note, however, that new antibodies that have been successfully used in standard immuno-fluorescence labeling are not guaranteed to work on resin-sections and need to be individually tested. In general, antibodies directed against abundant and continuous targets like, for example, abundant membrane proteins, cytoskeletal elements or DNA are more likely to give meaningful results since in these cases enough epitopes are accessible at the surface of the ultra-thin resin-section.
We conclude that post-embedding labelling for super-resolution STED microscopy on typical EM samples is applicable to different specimen and various targets. Though the STED images lack any contextual information.
Context information through dual-colour STED microscopy on resin sections
In order to analyse protein localisations, contextual information about surrounding structures is beneficial. In fluorescence microscopy, contextual information can be achieved through labelling of additional reference structures with fluorophores of a different emission wavelength. The accuracy of the colocalisation of the two detection channels, however, depends on the imaging scheme and has to be taken under consideration. In STED microscopy, the colocalisation accuracy depends on the depletion beam since it defines the position of the fluorophores which are still allowed to fluoresce. Hence, when using two different lasers for stimulated emission depletion, the positioning of the two STED beams has to be carefully monitored and potential deviations must be corrected . To avoid this potential problem and to ensure colocalisation without any need for correction, we used a single STED beam for both detection channels . To get structural context for the tubulin labelling in T. brucei we co-labelled this specimen with an antibody directed against the cell surface protein 'variant surface glycoprotein' (VSG), which helps this parasite to evade the hosts immune response . Two-colour STED microscopy (using a single STED laser for depletion) shows the labelling of the cell surface in the VSG channel and clearly separable the underlying subpellicular tubulin and the flagellar tubulin (figure 3(a)). In another example, the acrosomal region of bovine sperm was stained using fluorophore conjugated WGA (wheat germ agglutinin). This lectin binds N-acetyl glucosamine or sialic acid, on the plasma membrane of sperm called WGA receptors . We were able to specifically label the WGA receptors on the Lowicryl sections and confocal and STED microscopic images of these stainings show the outline of the sperm head (figures 3(b) and (c)). In the STED images the WGA receptors are clearly separable from the tubulin content of the sperm (figure 3(b)) and the DNA in the sperm-head (figure 3(c)). The staining of the plasma membrane of the sperm head provides a frame of reference for the antibody-labelled cytoskeletal structures and the highly condensed DNA.
Taken together, we could show that post-embedding labelling of various structures and dual-colour super-resolution STED microscopy is possible on resin sections. Next to various antibodies, also other specific ligands like lectins (e.g. WGA) can be used to label these samples, making this technique useful for numerous applications.
Correlative dual colour STED and dual gold label scanning electron microscopy
Decorating EM sections with fluorescent labels and imaging them using diffraction unlimited STED microscopy has benefits such as a superior z-resolution and the possibility to use the same samples for electron microscopic investigations. However, the key strength of being able to label and image EM sections using a STED microscope is the possibility to perform direct correlative (super-resolution) light and electron microscopy.
To explore correlative super-resolution light and scanning electron microscopy, T. brucei resin sections were transferred onto indium tin-oxide coated (ITO) coverslips suitable for both imaging modalities . The sections were labelled first with primary antibodies against alpha-tubulin and VSG. Next, secondary antibodies coupled to colloidal gold with two sizes (6 and 12 nm) were applied. Finally, fluorophore labelled secondary antibodies were added. The samples were first imaged with the STED microscope. The super-resolution data shows the membrane localisation of VSG and the subpellicular and flagellar alpha-tubulin (figure 4). The labelling was, regardless of the prior labelling with colloidal gold coupled antibodies, as specific and dense as in the samples that have not been labelled previously with immunogold conjugates (figures 3(a) and 4). Subsequently, the samples were imaged using a scanning electron microscope. The EM image shows the parasites membranes and the gold labelling indicates the localisation of VSG and alpha-tubulin (figure 4, supplementary figure 2). Comparing the data of the two imaging modalities, the advantage of immuno-fluorescence labelling over immuno-gold labelling becomes apparent (figure 4(b)). Even though the labelling with colloidal gold is specific and relatively efficient, immunofluorescence labelling provides a much higher labelling density combined with high specificity and unrivalled contrast. The labelling of the two separate structures (alpha-tubulin and VSG) with two gold-particle sizes is possible. Yet the discrimination of the two labels is challenging.
Taken together, we present an approach to perform correlative super-resolution light and electron microscopy. This method provides superior ultrastructural preservation enabling high quality electron microscopic investigations. It simultaneously facilitates efficient fluorescence labelling and diffraction unlimited imaging of various structures thus representing an alternative to immunogold labelling. We note that also other super-resolution microscopy approaches like dSTORM/GSDIM or DNA-Paint could benefit from this approach since the fluorophores are not embedded in the resin and can be exposed to any imaging buffer necessary for single molecule switching.
Methods and materials
Cultivation and preparation of cells
Cells of the budding yeast S. cerevisiae expressing genomically tagged Htb1-GFP (Yeast GFP clone collection, Life Technologies) were grown at 30 °C in liquid YP lactate medium (10 g l−1 yeast extract, 20 g l−1 pepton from casein, 2% (w/v) lactate, 20 mg l−1 uracil, 20 mg l−1 adenine sulfate) and harvested by centrifugation (3 min, 1500 g) during the logarithmic growth phase (OD ~1).
Human HeLa S3 spinner cells were grown in suspension culture using a BioBench Reactor (Applikon Biotechnology, Delft, Netherlands) in Dulbecco's modified Eagle medium (DMEM; PAA Laboratories), supplemented with 10% (v/v) dialyzed fetal bovine serum (PAA Laboratories) and 1 × penicillin/streptomycin and were harvested by centrifugation (5 min, 300 g).
Bloodstream forms of T. brucei clone MITat 1.2 (clone 221) were isolated from mice as described .
E. coli wild type cells and E. coli cells overexpressing OmpA were cultured and harvested as described .
Chemical fixation, dehydration and embedding
Trypanosoma and E. coli cells used for embedding in methacrylate resins were fixed with 2% (w/v) formaldehyde, 0.05% (w/v) glutaraldehyde in PBS for 1 h on ice and embedded in 2% (w/v) agarose. Specimen blocks were dehydrated with ethanol by PLT to −35 °C and infiltrated with Lowicryl K4M or HM20 (Polyscience Europe, Eppelheim, Germany). Polymerisation was carried out by irradiation with UV light for 24–48 h at −35 °C.
High pressure freezing, freeze-substitution and embedding
For fixation, mammalian cells, budding yeast cells and bovine sperms were transferred onto aluminum planchettes with a cavity of 150 µm and vitrified in a Leica HPM100 high pressure freezer (Leica Microsystems, Vienna, Austria). The vitrified specimens were freeze-substituted using a Leica EM AFS freeze substitution unit (Leica Microsystems) in 0,1% (w/v) uranyl acetate and 1% (v/v) water containing acetone. After removing the pellet from the planchette, it was embedded in HM20 (Polyscience Europe, Eppelheim, Germany) at −30 °C and polymerised at −50 °C for 72 h [26, 27].
Ultrathin sections (50–70 nm) were prepared using a Leica EM UC6 (Leica Microsystems) or a Reichert-Jung Ultracut E (C. Reichert, Vienna, Austria).
Immunolabelling for fluorescence and electron microscopy
A detailed protocol for all steps is given by Schwarz and Humbel, 2014 . In short: ultrathin sections were mounted on coverslips, particularly indium tin oxide coated ones (Optics Balzers, Liechtenstein), if subsequent imaging in a SEM was intended. After incubation in 5% (w/v) bovine serum albumin (BSA) in PBS or 0.2% (w/v) gelatine, 0.5% (w/v) BSA in PBS Trypanosoma sections were decorated using primary mouse-antibodies against alpha-tubulin (Sigma T5168, Sigma-Aldrich, St Louis, USA) and rabbit antibodies against VSG (a gift from Dr George A M Cross, Rockefeller University, New York).
HeLa cell sections were decorated using primary antibodies against alpha-tubulin and dsDNA (61014, Progen, Heidelberg, Germany).
Sections of yeast cell expressing Htb1–GFP were decorated using primary antibodies against GFP (ab6556, Abcam, Cambridge, UK).
Sections of E. coli cells were labelled using primary antibodies against OmpA (MPI for Biology, Tübingen, Germany).
Sections of bovine sperms were labelled with primary antibodies against alpha-tubulin and dsDNA and, in addition with wheat germ agglutinin (WGA) coupled to Alexa Fluor 594 (Life Technologies). Primary antibodies were detected by secondary antibodies (sheep anti-mouse, goat-anti-rabbit; Jackson ImmunoResearch Laboratories, West Grove; PA; USA) custom labelled with the fluorophores Atto 590 (ATTO-TEC, Siegen, Germany), Alexa Fluor 594 (Life Technologies, Carlsbad, CA, USA) or Abberior STAR RED (Abberior, Göttingen, Germany) [36, 37]. For electron microscopy, primary antibodies were detected using goat antibodies against rabbit or mouse-IgG coupled to 6 or 12 nm colloidal gold (Jackson ImmunoResearch Laboratories, West Grove; PA; USA). For CLEM primary antibodies were detected using secondary antibodies labelled with colloidal gold, first. Then free epitopes of the primary antibodies were decorated as indicated above for light microscopy. After several washing steps with PBS, the samples were mounted on coverslips in Mowiol containing DABCO.
Light microscopy and image processing
Overview images were generated using a Leica DM6000 epifluorescence microscope (Leica, Wetzlar, Germany). Super-resolution microscopy of the samples was performed using custom built STED microscopes described previously [30, 38] or a 2-colour STED 775 QUAD Scanning microscope (Abberior Instruments, Göttingen, Germany). Except for (lagrange) interpolation, smoothing and contrast stretching, no further image processing was applied.
Transmission electron microscopy
Sections shown in supplementary figure 1 were counterstained with 1% (w/v) uranyl acetate and lead citrate and examined using a Philips CM 120 BioTwin transmission electron microscope (Philips, Eindhoven, Netherlands).
Scanning electron microscopy
For scanning electron microscopy and correlative microscopy, thin sections were dried on indium-tin oxide (ITO) coated coverslips (Optics Balzers, Liechtenstein) and then labelled as described . After acquisition of the STED images of the labelled sections, the coverslips were released from the glass slide by immersing into PBS to soften the mowiol. Then they were washed with water and stained with 1% (w/v) uranyl acetate. The ITO coverslips were mounted on SEM stubs and care has been taken that the conductive ITO surface is well connected to the stub. For this we used carbon paint and adhesive copper strips. The sections were analyzed in a Helios Nanolab 650 DualBeam microscope (FEI Company, Eindhoven, Netherlands). Imaging was driven by a correlative and automated acquisition software (Maps, FEI Company Eindhoven, Netherlands). We used the software for guiding the instrument to the area of interest but the acquisition was done with the UI of the microscope. Images were taken at 1–3 keV acceleration voltage with a current of 400–800 pA, at a working distance of 1–3 mm. The best conditions were 2 keV, 800 pA, with 1–3 µs dwell time at a working distance of 3 mm. Images were recorded with the through the lens secondary electron detector (SE-TLD), the mirror detector (MD) and the in-column detector (ICD) using the immersion lens mode. The scan size was 6144 × 4096 pixels with pixel resolution of about 8 Å to visualise the gold particles.
A part of the work was supported by a grant from the Deutsche Forschungsgemeinschaft (SFB 1286, project A5, to SJ) and by an R'Equip Grant 316030_128692 of the Swiss National Science Foundation (to BH). We thank T Conrad for providing mammalian cells and W Wemheuer for providing the bovine sperm. We are grateful for the donation of antibodies from Dept. Henning (MPI for Biology, Tübingen, Germany). Infrastructure development support was provided by the Swedish Foundation for Strategic Research (RIF14-0091, to DCJ).
- Müller M and Moor H 1984 Cryofixation of thick specimens by high pressure freezing Science of Biological Specimen Preparation 1983 ed J Revel et al (Chicago, IL: SEM Inc, AMF O’Hare) pp 131–8
- Humbel B M and Schwarz H 1989 Freeze-substitution for immunochemistry Immuno-Gold Labeling in Cell Biology ed A J Verkleij and J L M Leunissen (Boca Raton, FL: CRC Press) pp 115–34
- Humbel B M and Müller M 1984 Freeze substitution and low temperature embedding Proc. 8th European Congress on Electron Microscopy (Budapest,) vol 3 pp 1789–98
- Freudl R, Schwarz H, Stierhof Y D, Gamon K, Hindennach I and Henning U 1986 An outer membrane protein (OmpA) of Escherichia coli K-12 undergoes a conformational change during export J. Biol. Chem. 261 11355–61