Novel Contrasting and Labeling Procedures for Correlative Microscopy of Thawed Cryosections

Abstract

One of the major challenges for correlative microscopy is the preparation of the sample; the protocols for transmission electron microscopy (TEM) and fluorescence microscopy (FM) often prove to be incompatible. Here, we introduce 2+Staining: an improved contrasting procedure for Tokuyasu sections that yields both excellent positive membrane contrast in the TEM and bright fluorescence of the probe labeled on the section. 2+Staining involves the contrasting of the immunolabeled sections with 1% osmium tetroxide, 2% uranyl acetate and lead citrate in sequential steps, followed by embedding in 1.8% methyl cellulose. In addition, we demonstrate an amplification of the fluorescent signal by introducing additional antibody incubation steps to the immunolabeling procedure. The methods were validated using the integrated laser and electron microscope (iLEM), a novel tool for correlative microscopy combining FM and TEM in a single setup. The approaches were tested on HL-60 cells labeled for lysosomal-associated membrane protein 2 (LAMP-2) and on sections of muscle from a facioscapulohumeral dystrophy mouse model. Yielding excellent results and greatly expediting the workflow, the methods are of great value for those working in the field of correlative microscopy and indispensible for future users of integrated correlative microscopy.

By combining the strengths of two imaging modalities on a single sample, correlative microscopy has proven to be extremely valuable for research in the field of cell biology. The fluorescence microscope (FM) allows for live cell imaging and, furthermore, for the localization of the region of interest (ROI) over a large field of view. The transmission electron microscope (TEM) has an unprecedented resolution and is not limited to imaging of just the probe but also the cellular context. Therefore, by combining the FM and TEM in the analysis of a sample, cell biology research is reaching new heights.

Correlative microscopy, however, also poses serious challenges. The structure or ROI identified with the FM needs to be traced back in the TEM, which is a difficult and time-consuming process. The next hurdle for correlative microscopy lies in the preparation of the specimen. The FM demands samples to be labeled with a fluorescent probe. With exception of cryo-electron microscopy, TEM imaging requires the biological specimen to be contrasted with heavy metal salts. When the fluorophore is in close proximity to the heavy metal stain, its fluorescence is quenched (Karreman et al. 2009; Valeur 2002). To overcome this, additional sample preparation steps need to be carried out in between imaging the sample with the FM and the TEM (Takizawa et al. 2003; Verkade 2008; van Rijnsoever et al. 2008; Valentijn et al. 2010). However, by performing these steps, the ROI is often lost or distorted, making it difficult to recognize in the TEM. This makes the correlation between the FM and the TEM images very time-consuming and sometimes even impossible. Here, we introduce a new, optimized sample preparation procedure resulting in specimens apt for both FM and TEM imaging without any intermediate specimen preparation steps.

Localization of the ROI in the specimen with both FM and TEM requires immunolabeling with fluorescent and electron-dense probes. This can be achieved by labeling with correlative probes such as quantum dots (Nisman et al. 2004; Giepmans et al. 2005, 2006; Cortese et al. 2012; Dukes et al. 2010) or fluoro(nano)gold (Roth et al. 1980; Robinson and Vandré 1997; Powell et al. 1998; Takizawa et al. 1998, 2000), which are visible in both imaging modalities. In our work, however, we label the sample sequentially with immunogold and immunofluorescence markers (Vicidomini et al. 2008; Cortese et al. 2012; Fabig et al. 2012). Hereto, a protocol similar to immunolabeling of thawed Tokuyasu cryosections can be used. The samples are mildly chemically fixed, cryoprotected with sucrose, frozen in liquid nitrogen, and sectioned at cryogenic temperatures. The resulting cryosections are mounted on a TEM grid, thawed, and immunolabeled with small gold particles (Bernhard and Viron 1971; Tokuyasu 1973, 1980; Slot and Geuze 2007) followed by labeling with a fluorescent antibody. One of the advantages of the Tokuyasu method for immunolabeling is that the use of harsh fixatives is avoided, limiting the introduction of artifacts and structural alterations to the epitopes. Also, the epitope is highly accessible for the antibody because the biological material is not embedded during the immunolabeling procedure.

To allow for image contrast of Tokuyasu sections in the TEM, a commonly employed staining protocol is the Griffiths method. Here, the samples are first stained on drops of uranyloxalate in water followed by embedding of the sections in a thin layer of 0.1% to 0.4% uranyl acetate (UA) in methylcellulose (MC) (Griffiths et al. 1984). Although this protocol yields excellent membrane contrast, the signal of the fluorescence labeling on the section is strongly reduced or lost altogether (Karreman et al. 2009). To image the same Tokuyasu sections both with FM and TEM, an undesirable compromise needs to be made by decreasing the amount of heavy metal stain employed to contrast the specimen and accepting a lower fluorescence signal (Karreman et al. 2009, 2011; Karreman, Agronskaia, et al. 2012). We hypothesized that the final preparation step, embedding the sections in the mixture of UA/MC, was the most detrimental. This strongly increases the concentration of the heavy metal UA, leading to fluorescence quenching. Previously, we had found that briefly staining the sections (mounted on a TEM grid) by placing them face-down onto drops of UA followed by embedding in MC yielded fairly adequate membrane contrast and showed little effect on the fluorescence brightness (Karreman et al. 2009). Here, we introduce a novel protocol for the postcontrasting of Tokuyasu sections: 2+Staining. The contrasting procedure is based on staining the Tokuyasu section with drops of different heavy metals, followed by embedding of the section in MC without UA. In this article, inspired by the contrasting protocols employed for resin-embedded biological material, osmium tetroxide (OsO₄), lead citrate (LC), and UA are employed for the staining of Tokuyasu sections.

The novel staining procedure yields exquisite positive membrane contrast and is of interest not only for users of correlative microscopy but for all those working with the Tokuyasu method. Moreover, the method is indispensible for future users of integrated correlative microscopy procedures such as the integrated laser and electron microscope (iLEM). This novel imaging tool embodies a custom-designed laser-scanning FM mounted on the side port of a conventional TEM, allowing for both FM and TEM analysis of a single specimen (Agronskaia et al. 2008). By integrating both imaging modalities, the ROI identified by FM can be quickly and effortlessly recovered in the TEM and studied with high resolution (Karreman et al. 2009; Karreman, Buurmans, et al. 2012). A similar set-up, the iCorr, has been recently introduced on the market by FEI Company (Eindhoven, the Netherlands). Two major hurdles in correlative microscopy, sample preparation and the retrieval of the ROI in the TEM, are overcome by imaging the specimen stained according to the new protocol with integrated correlative microscopy approaches.

Materials and Methods

Cell Culture, Transgenic Mice, Fixation, Embedding, and Sectioning

The HL-60 cells were cultured as described earlier (Fleck et al. 2005). The HL-60 cells were fixed in 4% formaldehyde (FA; from paraformaldehyde; Sigma-Aldrich, St. Louis, MO) in PHEM buffer (composed of 60 mM PIPES [Merck; Readington, NJ], 25 mM HEPES [Merck], 10 mM EGTA [Sigma-Aldrich], and 2 mM MgCl [Merck], pH adjusted to 6.9) for 2 hr at room temperature, with continued overnight fixation at 4C. After washing away the fixative, the cells were scraped in 1% gelatin (De Twee Torens; Delft, the Netherlands) in PHEM buffer and embedded in 12% gelatin in PHEM buffer. Blocks of ~1 mm³ were cut and cryoprotected by overnight infiltration in 2.3 M sucrose in PHEM buffer. Next, the blocks were plunge frozen in liquid nitrogen. Following trimming of the sample blocks at −101C, 90-nm glossy-looking sections were produced and picked up in a drop of a 1:1 mixture of 2% MC and 2.3 M sucrose in PHEM. Upon thawing, the sections were mounted on TEM grids with a carbon-coated Formvar film.

Transgenic C57BL/6J mice overexpressing human FRG1 were maintained as described previously (Bortolanza et al. 2011). After sacrificing the mice, the vastus lateralis was isolated, cut longitudinally into three pieces, and transferred into the fixative. The muscle was fixed in 2% FA + 0.2% glutaraldehyde (GA, EM grade; Polysciences, Eppelheim, Germany) in PHEM for 2 hr at room temperature and continued overnight at 4C. The muscle was cut into ~1-mm³-sized blocks and embedded in drops of 12% gelatin in PHEM buffer. The sample blocks were cut out from the gelatin drop and further processed as described above. Here, however, the sucrose infiltration was performed for approximately 48 hr instead of overnight.

Immunolabeling and Staining for iLEM Analysis

Immunolabeling with mouse anti–lysosomal-associated membrane protein 2 (LAMP-2) (BD Biosciences Pharmingen, Franklin Lakes, NJ, diluted 1:150 in the blocking buffer: PBS containing 1% BSA [Sigma-Aldrich]) and rabbit anti–cleaved caspase 3 (Asp175, diluted 1:10 in blocking buffer; Cell Signaling Technology, Danvers, MA) was performed as described previously (Karreman et al. 2009, 2011; Karreman, Agronskaia, et al. 2012). For the immunolabeling of LAMP-2 and cleaved caspase 3, an Alexa–anti-Alexa labeling step was introduced in the protocol. Following incubation with the primary antibody and subsequent washing steps in 0.1% BSA in PBS, the sections were incubated for 45 min with an Alexa 488–conjugated anti-mouse antibody. We employed goat anti-mouse Alexa 488 (diluted 1:200 in blocking buffer; Invitrogen, Carlsbad, CA) for targeting mouse anti-FRG1 and mouse anti–LAMP-2. To target rabbit anti–cleaved caspase 3, goat anti-rabbit Alexa 488 (diluted 1:200 in blocking buffer; Invitrogen) was used. The sections were then labeled with rabbit anti–Alexa 488 (diluted 1:100 in blocking buffer; Invitrogen), followed by labeling with Protein A Gold (PAG; Utrecht Medical Centre, Utrecht, the Netherlands) and finally with goat anti-rabbit conjugated to tetramethylrhodamine isothiocyanate (TRITC) (diluted 1:200 in blocking buffer; Invitrogen). In between the antibody incubations, the unbound antibodies were removed by five 2-min washing steps in 0.1% BSA in PBS. The general steps for the amplified labeling procedure are indicated in the light gray blocks in the flowchart of Fig. 1. Three different staining and embedding methods were employed. Immunolabeled Tokuyasu sections, mounted on a TEM grid, were rinsed, on ice, face-down on a drop of 1.8% MC in water (Sigma-Aldrich) and directly embedded as described before (Karreman et al. 2011); 2+Stained; or stained according to the Griffiths method. 2+Staining entails first the fixation of the immunolabeled sections with 1% OsO₄ (Electron Microscopy Sciences; Hatfield, PA) diluted in water, for 30 min on ice and in the dark. After two very brief washing steps, the sections were stained for five minutes with 2% UA (pH 4, SPI-Chem, West Chester, Pennsylvania) in water. This was followed by five 2-min washing steps in water, after which the sections were stained with LC (prepared according to Reynolds 1963 from sodium citrate [Merck] and lead nitrate [Merck]) for less than 3 min (we note that 1 or 2 min should be sufficient). This step was performed in a closed Petri dish shielded from the air and in the presence of NaOH pellets to prevent lead carbonate precipitation. Before embedding in 1.8% MC as described above, the sections were washed again twice, very briefly, in water. It is of great importance to keep the washing steps as short as possible; the excess of stain should be rinsed away, but the loss of specific stain in the specimen should be limited. The 2+Staining procedure is summarized in Fig. 1 (dark gray boxes). The Griffiths method as employed here involved initial stabilization of the membranes with 2% uranyl oxalate (pH 7, prepared from oxalic acid; J. T. Baker Chemicals NV, Deventer, the Netherlands) and UA (SPI-Chem), both diluted in water, followed by two brief washings steps and embedding in 0.4% UA in MC, as described before (Karreman et al. 2009).

Figure 1.

A flowchart outlining the steps involved in the amplified labeling (light gray boxes) and 2+Staining (dark gray boxes) procedures. iLEM, integrated laser and electron microscope.

FM Analysis of Thick Tokuyasu Sections of FRG1-Transfected Mouse Vastus Lateralis Tissue

For FM analysis of the FRG1-transfected mouse vastus lateralis tissue, 450-nm glossy-looking cryosections were produced from the sample block and mounted on a silan-coated glass slide. The pickup solution was melted away at 37C by washing the sections four times for 5 min with prewarmed PBS, followed by five additional 2-min washing steps in PBS at room temperature. The free aldehyde groups were quenched by washing the sections, mounted on a TEM grid, five times for 2 min face-down onto drops of 0.02 M glycine (Merck) in PBS. To reduce the level of autofluorescence, the sections were then incubated with 0.01 mg mL^–1 sodium borohydride (Roche; Basel, Switzerland) for 5 min, followed by washing thoroughly five times for 2 min in PBS. Next, the nonspecific binding of the antibody was prevented by incubating the sections for 15 min with blocking buffer, followed by a 1-hr incubation with the mouse anti-FRG1 (1:10 in blocking buffer; Santa Cruz Biotechnology, Santa Cruz, CA) or rabbit anti–cleaved caspase 3. The sections were rinsed five times for 2 min in 0.1% BSA in PBS to remove the unbound antibody and incubated with goat anti-mouse Alexa 488, for FRG1-labeled sections, or with goat anti-rabbit Alexa 488, for cleaved caspase 3–labeled sections. After washing another five times for 2 min with PBS, the nuclei were stained by incubation with 0.05 mg mL^–1 DAPI (DAKO; Glostrup, Denmark) in PBS. Finally, after washing five times for 2 min first in PBS and next in water, the sections were embedded in Prolong Gold (Invitrogen).

Imaging, Statistics, and Analysis

FM imaging was performed with a wide-field Olympus AX70 (Olympus, Tokyo, Japan) equipped with a Nikon DXM1200 camera (Nikon, Tokyo, Japan), operating under Nikon ACT-1 software. The immunolabeled thawed thin cryosections, mounted on TEM grids, were imaged in the iLEM (Agronskaia et al. 2008). The iLEM embodies a custom-designed laser-scanning FM mounted on one of the side ports of a Tecnai 12 120-kV TEM (FEI Company). The FM module of the iLEM was equipped with a 532-nm laser (Newport Corporation; Irvine, CA) and an avalanche photo diode (APD; PerkinElmer Optoelectronics, Fremont, CA) detector. The objective lens of the FM has a numerical aperture of 0.55 and can image areas up to 500 µm² in size without significant aberrations. The FM was operated by software written in LabView 8.0 (National Instruments; Austin, TX). The TEM was operated at 80 kV, and images were captured with a bottom-mount TEMCam-F214 (Tietz Video and Image Processing Systems, GmbH, Gauting, Germany) charge-coupled device (CCD) camera.

A statistical analysis of the qualitative difference between the three different contrasting conditions was performed based on fluorescence intensity and amount of fluorescently labeled structures (fluorescent area) observed in the iLEM. Hereto, the pixels showing specific labeling were selected in ImageJ (National Institutes of Health; Bethesda, MD) via thresholding the image to a specific level of fluorescence signal, which was judged by eye. The mean fluorescence intensity per pixel and the fluorescent area were calculated for similar conditions in different immunolabeling experiments (MC embedding, n=6; 2+Staining, n=4; Griffiths method, n=5). The mean of the fluorescence intensity and the fluorescent area of each condition in each experiment were normalized to the mean of the positive control, embedded in MC. Normalization of all the data to the mean of the control condition permits an analysis of the data with a t-test (Valcu and Valcu 2011). The image quality assessment of the TEM images was performed based on scoring the visibility of various cellular components (see Results section). Normalization of these results was performed as described above, but here the Griffiths method was employed as a positive control method. Microsoft Office 2007 Excel (Microsoft Corporation; Redmond, WA) was employed for the calculation of the t-tests (two-tailed, unequal variance) and the plotting of the graphs. The analysis of the different labeling protocols, direct labeling and via the Alexa–anti-Alexa amplified labeling method, was performed based on two different labeling experiments. The fluorescence intensity and area were calculated for the three different contrasting conditions. For these conditions, the intensities and areas were normalized to the mean intensity of the samples labeled via the amplified Alexa–anti-Alexa 488 protocol and directly embedded in MC. The effect of the labeling protocol was tested for each different condition (two experiments, three different conditions) by performing a two-tailed paired t-test.

All figures were prepared in Photoshop CS3 (Adobe; San Jose, CA). For the production of the FM/TEM overlay images, the γ-value of the FM image was adjusted. The FM and TEM images were scaled for optimal brightness and contrast by applying linear adjustments to the levels of the entire image.

Results

The 2+Staining Method

To validate our staining protocols, human promyelocytic leukemia cells (HL-60) were employed as a test system. Thin Tokuyasu sections were immunofluorescence (IF) and immunogold (IG) labeled for LAMP-2 (also known as CD107B) with TRITC and PAG (see Materials and Methods). LAMP-2 associates to late endosomes (Akasaki et al. 1996), the limiting membrane of lysosomes, and the trans-Golgi system (Chen et al. 1985; Karlsson and Carlsson 1998). We chose this antigen for the validation of our methods, because it shows a clear and recognizable labeling pattern and is highly abundant in the HL-60 cells. Following immunolabeling, the sections were either directly embedded in MC (uncontrasted) or contrasted according to different experimental protocols and embedded. The fluorescence intensity was used as an indicator of the effect of the heavy metal contrasting agents on the quantum yield of TRITC. We tested different combinations and concentrations of heavy metal stains to contrast the Tokuyasu sections to optimize both TEM contrast and TRITC fluorescence quantum yield. On the basis of these experiments, we developed an optimized contrasting procedure. The novel method is named “2+Staining” because it is yields specimens apt for two imaging modalities, FM and TEM, and results in positive (+) membrane contrast in the TEM. The dark gray boxes in the flowchart in Fig. 1 highlight the steps performed for 2+Staining. The immunolabeled sections, mounted on a TEM grid, were first fixed and stained at 4C in the dark, by placing them face-down onto drops of 1% OsO₄ diluted in water for 30 min, followed by five 2-min washing steps with water. Satisfying results can also be achieved when the OsO₄ fixation step is omitted from the protocol. However, in our experiments, postfixation with OsO₄ generally yielded a better membrane contrast and more reproducible results. Next, the sections, mounted on a TEM grid, were stained for 5 min, by placing them face-down onto 2% aqueous UA drops. After briefly washing away the excess of UA on two drops of water, the sections were stained for 3 min on drops of LC. Finally, the sections were rinsed twice in water, embedded in MC (see Materials and Methods), and imaged with the iLEM (Suppl. Fig. S1A–C). iLEM analysis of 2+Stained specimens labeled for LAMP-2 revealed both good TEM contrast and bright fluorescence. Due to the excellent membrane contrast in the TEM, the intracellular localization of LAMP-2 to the lysosomes could be studied. In agreement with earlier results (Chen et al. 1985; Mane et al. 1989), LAMP-2 was mainly found on the limiting membrane of the lysosomes and in multivesicular bodies. Next, we set out to characterize the ultrastructure and the fluorescence intensity of differently stained specimens with the iLEM. A comparison was made between (1) direct embedding in MC, in which no contrasting steps are performed; (2) the Griffiths method, in which the sections were first stained by being placed faced down on drops of aqueous uranyl oxalate for 5 min, briefly washed in water, and then embedded in 0.4% UA in MC; and, finally, (3) the new 2+Staining method. Figure 2 shows typical TEM images of the appearance of various cellular compartments and organelles following the different treatments. When the sections were not postcontrasted with heavy metal stains (uncontrasted; Fig. 2A, D, G, J, M, P), it was extremely difficult to identify the different organelles and membranes in the TEM. By contrasting the sections according to the Griffiths method, the cellular structures were clearly distinguishable (Fig. 2B, E, H, K, N, Q). This method yielded a combination of positive contrast, observed as the dark staining of the heterochromatin, and negative contrast, observed in the membranes (Griffiths et al. 1982, 1984). The novel 2+Staining method yielded a positive contrast of the different organelles and membranes observed in the cell (Fig. 2C, F, I, L, O, R). This staining method resembled strongly the contrast obtained from stained Epon or Lowicryl resin-embedded specimens. This is unsurprising because this method is similar to the staining protocol employed for these types of specimens.

Figure 2.

The appearance of different cellular organelles in uncontrasted cryosections and sections stained according to the Griffiths and 2+Staining methods. The 90-nm-thin Tokuyasu sections of HL-60 cells were immunolabeled for lysosomal-associated membrane protein 2 (LAMP-2). Next, the sections were either directly embedded in methylcellulose (uncontrasted) (A, D, G, J, M, P), stained according to the Griffiths method (B, E, H, K, N, Q), or the 2+Staining method (C, F, I, L, O, R) and imaged with the transmission electron microscope (TEM). TEM images reveal the nuclear envelope (NE) and the endoplasmic reticulum (ER) (A–C), the Golgi (G) (D–F), mitochondria (M) (G–I), lysosomes (L) (J–L), the plasma membrane (PM) (M–O), and the centriole (C) (P–R). The scale bars in all panels represent 200 nm.

Qualitative Analysis of 2+Staining for Correlative Microscopy

Processing the sections according to the three distinct methods yielded a striking difference in fluorescence intensity of the LAMP-2 IF label (Fig. 3). When the sections were not stained with heavy metals (uncontrasted; Fig. 3A, B), the fluorescence of TRITC was bright. This signal was, however, strongly reduced in intensity when the sections were contrasted according to the Griffiths method (Fig. 3C, D). The effect of 2+Staining (Fig. 3E, F) on TRITC’s brightness was much less detrimental compared with the Griffiths method. To quantify these effects, in a series of different labeling experiments (n>5), we measured the fluorescence intensity for the different contrasting conditions (Fig. 4A). For each condition, a total of 110 to 185 FM images obtained from different specimens were analyzed. The fluorescence intensity values were normalized to the mean of the fluorescence intensity of the MC-embedded specimen (uncontrasted; normalized fluorescence intensity [NFI] = 1.00 ± standard error [SE] = 0.05, number of images analyzed (n) = 185). The sections contrasted according to the Griffiths method showed a significant reduction in fluorescence intensity of 44% (NFI = 0.56 ± 0.01; n=140; t-test, p=0.03; Fig. 4A, “Griffiths”). Importantly, the novel 2+Staining method did not affect the fluorescence signal (NFI = 1.06 ± 0.03, n=112): No significant difference in intensity (t-test, p=0.8) was found between uncontrasted and 2+Stained sections.

Figure 3.

2+Staining yields excellent membrane contrast and minimally influences the intensity of on-section immunofluorescence labeling. Tokuyasu sections were immunolabeled for lysosomal-associated membrane protein 2 (LAMP-2) with Protein A Gold and tetramethylrhodamine isothiocyanate and subsequently directly embedded in methylcellulose (uncontrasted) (A, B), stained according to the Griffiths method (C, D), or the 2+Staining method (E, F) and imaged with the integrated laser and electron microscope. The left column (A, C and E) shows transmission electron microscopy images of sections treated according to the different contrasting conditions. The right column (B, D, F) depicts fluorescence images of 210-µm² areas on section. The scale bars represent 500 nm for panels A, C, and E and 30 µm for panels B, D, and F. C, centriole; G, Golgi; L, lysosomes; M, mitochondria; N, nucleus.

Figure 4.

The effect of the different contrasting and labeling methods on the fluorescence signal intensity, the fluorescence area ratio, and the quality of the transmission electron microscopy (TEM) image contrast. The immunofluorescence (IF) intensity (A), the relative amount of IF-labeled structures (B), and the TEM image contrast quality (C) were analyzed from thin 90-nm Tokuyasu sections that were immunogold (IG) and IF labeled for lysosomal-associated membrane protein 2 (LAMP-2). The sections were directly embedded in methylcellulose (MC; uncontrasted, left), stained according to the Griffiths method (Griffiths, middle), or the 2+Staining method (2+Stain, right). The quantification of the fluorescence intensity, area, and the TEM contrast was performed as outlined in the Results section and the Material and Methods section. Sections of HL-60 cells were immunolabeled for LAMP-2 via the Alexa–anti-Alexa amplified labeling procedure (“Amplified Labeling,” outlined bars) or via the conventional labeling protocol (“Direct Labeling,” solid bars). Following immunolabeling, the sections were directly embedded in MC (“Uncontrasted”), contrasted according to the Griffiths method (“Griffiths”), or the 2+Staining method (“2+Stain”). The IF intensity (D) and the relative amount of observed fluorescence labeling (E) were calculated for the different conditions as described in the Results and the Materials and Methods sections. The error bars in both bar graphs indicate the standard error of the results. Statistically relevant differences are marked with asterisks (t-test, p<0.05). The two asterisks indicate a p-value below 0.01.

While performing the fluorescence measurements, a difference was noticed in the number of positively labeled structures for the different contrasting protocols. Unstained or 2+Stained sections showed more fluorescently labeled structures per section area compared with specimens stained according to the Griffiths method. Importantly, the different contrasting protocols yielded no obvious difference in amount of IG labeling (data not shown). We presume that, by employing the Griffiths method, low levels of LAMP-2 labeling became invisible for FM. The already weak fluorescence signal of the scarcely labeled structures was now below the detection level of the FM or lost in background noise. In uncontrasted or 2+Stained sections, the dim fluorescence signal was preserved, and therefore more fluorescent structures could be observed; both the weak and the brightly labeled structures were visible. To quantify this effect, the ratio of fluorescently labeled pixels and the total number of pixels in a section area was calculated (see Materials and Methods). The number of cells and the number of lysosomes per section area were not fixed. However, because we analyzed 110 to 185 FM images per condition, this variation was minimized. The measurements indicated a highly reproducible fluorescence area to total section area ratio of 0.2:1 for uncontrasted sections (Fig. 4B, “Uncontrasted,” 0.2 ± 0.01, n=185). For the Griffiths staining, the fluorescence area ratio was reduced to 0.09:1 (Fig. 4B, “Griffiths,” 0.09 ± 0.01, n=140). This result indicates that more than half of the positively labeled structures, which were readily observed in uncontrasted specimens, became undetectable using the FM after staining the specimens according to the Griffiths method. 2+Stained sections, however, showed the same fluorescence to total area ratio of 0.2:1 as the uncontrasted sections (Fig. 4B, “2+Stained,” 0.2 ± 0.01, n=112). Compared with uncontrasted Tokuyasu sections, 2+Staining did not lead to a decrease in IF labeling intensity or influence the detection sensitivity of faintly fluorescing structures.

In an attempt to quantitate the quality of TEM contrast yielded by the different methods, we scored the appearance of various cellular structures. The features that were evaluated included the nuclear envelope, the Golgi, the vesicles around the Golgi, the mitochondria, the lysosomal membranes, the endoplasmic reticulum membrane, the plasma membrane, and the centriole (Fig. 2). Each of the eight features was rated on a scale of 0 to 2 for their visibility: indistinguishable (score: 0), discernible (score: 1), or clearly visible with high structural detail (score: 2), yielding a maximum score of 16 per condition (Fig. 4C). In five different labeling experiments, two or more specimens were analyzed per staining condition. The quality of the contrast was quantified for each method based on the averaged scores of the different features observed in the TEM. The Griffiths staining method yielded the highest quality of TEM contrast (Fig. 4C, “Griffiths,” 15.8 ± 0.12, n=72), followed closely by the 2+Staining method (Fig. 4C, “2+Stain,” 15.04 ± 0.46, n=60). When the specimen was not stained, the image contrast was insufficient to recognize the features (Fig. 4C, “Uncontrasted,” 2.25 ± 0.36, n=132). The quality of the membrane contrast was significantly lower than the contrast resulting from the Griffiths and 2+Staining methods (t-test, p=5.75 × 10^–12 and p=5.46 × 10^–05, respectively).

Combined, these results indicate that 2+Staining of Tokuyasu sections yields superb TEM image contrast without affecting the fluorescence signal of the immunolabeling on section.

The Alexa–Anti-Alexa Amplified Labeling Procedure

Immunolocalization on thin sections often yields disappointing IG and IF labeling levels. The occupancy of the epitope in the specimen might be low, or immunolabeling is hindered due to the sample preparation protocol. Here, we set out to increase both the immunofluorescence and immunogold labeling signals by the addition of extra steps to the existing labeling protocol. Following primary antibody incubation with mouse anti–LAMP-2, the sections were incubated with a goat anti-mouse Alexa 488 secondary antibody. Next, this antibody was targeted with a rabbit anti–Alexa 488 antibody, the Alexa–anti-Alexa step. Finally, the specimen was labeled with PAG, for localization in the TEM, and with TRITC conjugated to a goat anti-rabbit secondary antibody for FM imaging (see Materials and Methods and the light gray boxes in Fig. 1). To quantify and compare the fluorescence intensity between the different experiments and conditions, the results were normalized to the mean intensity of TRITC labeled via the amplified labeling method on uncontrasted sections. Via the Alexa–anti-Alexa labeling step, an increase in both the fluorescence and the immunogold signal is created. We presume that, with each additional labeling step, the number of antibodies connected to the epitope is increased, which amplifies the signal. The normalized fluorescence intensity of unstained (1.00 ± 0.04, n=61) and contrasted (Griffiths: 0.72 ± 0.02, n=55, and 2+Stained: 1.27 ± 0.04, n=26) sections was higher when the amplified labeling method (Fig. 3D, outlined bars) was employed compared with unstained (0.79 ± 0.02, n=51) and contrasted (Griffiths: 0.48 ± 0.02, n=51, and 2+Stained: 0.74 ± 0.02, n=26) sections labeled via the direct method (Fig. 4D, solid bars). Furthermore, the amplified labeling method yielded a higher fluorescence area ratio (uncontrasted: 0.21 ± 0.01, n=61; Griffiths: 0.10 ± 0.01, n=55; 2+Stained: 0.24 ± 0.01, n=26; Fig. 4E, outlined bars) compared with direct labeling (uncontrasted: 0.13 ± 0.01, n=51; Griffiths: 0.08 ± 0.04, n=51; 2+Stained: 0.13 ± 0.01, n=26; Fig. 4E, solid bars). The observed difference was significant for both the fluorescence intensity (paired t-test, p=0.017) and the fluorescent area ratio (paired t-test, p=0.039). We note that with the extra steps added to the immunolabeling protocol, the chance of background labeling also increases. We did not observe any significant increase in nonspecific labeling events in the samples compared with direct labeling (data not shown). In conclusion, a brighter fluorescence signal could be observed by employing the Alexa–anti-Alexa amplified labeling protocol compared with the direct labeling method.

Applications of the 2+Staining Method in the Study of FSHD

The 2+Staining procedure was applied in the research of facioscapulohumeral dystrophy (FSHD), the third most common form of inherited muscle dystrophy worldwide. FSHD region gene 1 (FRG1) is one of the proteins proposed to be involved in FSHD (Van Deutekom et al. 1996). FRG1-overexpressing mice are an animal model for FSHD, displaying a phenotype that strikingly resembles the functional and structural characteristics of FSHD (Gabellini et al. 2006; Bortolanza et al. 2011). Thick Tokuyasu sections of vastus lateralis muscle of FRG1-overexpressing mice were imaged with FM. Low levels of cleaved caspase 3 (CC3) were observed in a subset of the FRG1-positive nuclei (Suppl. Fig. S2A, B). To investigate the structural characteristics of these cells, thin cryosections of FRG1-overexpressing mouse vastus lateralis were IG and IF labeled for CC3 and 2+Stained. iLEM revealed the labeling of CC3 in the cytoplasm but also in the nuclei of the cells, in agreement with previous findings (Olney et al. 2002; Kamada et al. 2005; Karreman et al. 2009; Hunt et al. 2011). CC3-positive nuclei were readily found with the iLEM (Fig. 5A). Figure 5B shows a nucleus positively labeled for CC3, found in the periphery of the muscle fiber. This muscle cell did not show any morphological hallmarks of apoptosis yet, indicating that it was not far into the process of programmed cell death and CC3 levels were therefore still comparatively low (Karreman et al. 2009). However, the activated form of caspase 3 in the nucleus was still detectable both with IF and IG labeling (Fig. 5A–C), demonstrating the high sensitivity of the 2+Staining and amplified labeling methods.

Figure 5.

Integrated laser and electron microscope (iLEM) imaging of thin 90-nm Tokuyasu sections of FRG1-overexpressing mouse vastus lateralis, immunogold (IG) and immunofluorescence (IF) labeled for cleaved caspase 3. (A) A fluorescence image of an area of 210 × 210 µm on the grid. We note the fluorescence signal from folds in the section and dirt (full arrowheads). Furthermore, background labeling can be observed on necrotic cells (open arrowheads). These “false positives” demonstrate the value of the iLEM; it is impossible to identify the difference between the aspecific and specific labeling based on the fluorescence signal alone and requires further investigation with transmission electron microscopy (TEM). The fluorescence signal from the area indicated with the white square is overlaid on a TEM image of the same area (B), showing various muscle cells (MC), a positively labeled nucleus (N), and a capillary (C). (C) A high-magnification TEM image of the area indicated in panel B with the black square. At this magnification, the localization of the IG labeling for cleaved caspase 3 in the nucleus (N) becomes apparent; the nucleolus (Nu) is free from labeling. 2+Staining clearly reveals the nuclear envelope (NE) and the cross section of the muscle fibrils (MF). Scale bars indicate 30 µm in panel A, 5 µm in panel B, and 500 nm in panel C.

Discussion

Sample Preparation for Correlative Microscopy

The 2+Staining method results in specimens that are compatible with both FM and TEM analysis; this makes the protocol very well suited for (integrated) correlative microscopy. Currently, correlative microscopy is increasingly employed in cell biology research, making the demand for apt sample preparation techniques more and more pressing. Although many advances have been made in sample preparation for correlative microscopy, most protocols require processing or staining of the specimen in between the transfer from the FM to the EM (Roessler et al. 1991; Ren et al. 2003; Takizawa and Robinson 2003; van Rijnsoever et al. 2008; Watanabe et al. 2011; Kopek et al. 2012). Protocols yielding resin-embedded specimens that can be imaged both with FM and TEM were recently developed (Nixon et al. 2009; Kukulski et al. 2011; Karreman, Agronskaia, et al. 2012). However, correlative imaging of Tokuyasu sections was hitherto not possible without intermediate processing steps.

Postcontrasting Tokuyasu Sections

A large number of different approaches to achieve satisfying TEM image contrast of thawed cryosections have been tested since the introduction of the method. Tokuyasu (1978, 1980) tested a large variety of combinations with UA, phosphotungstic acid (PTA), lead, and OsO₄ but was never fully satisfied with the results. One of the encountered problems was the irreproducibility of the membrane contrast (Tokuyasu et al. 1981; Griffiths et al. 1982). The heavy metal stain was lost while rinsing, leading to destaining of the specimen. Although destaining is also a threat to the success of 2+Staining, we can achieve excellent and highly reproducible TEM contrast. Indeed, during the development of the protocol, we encountered some variation in quality of the TEM contrast, but this was overcome by keeping the washing steps as short as possible (see Materials and Methods).

We observed differences in contrast between 2+Stained specimens and sections stained according to the Griffiths method. This is due to the employment of different contrasting agents; in the Griffiths method, the TEM image contrast mainly originates from UA, whereas 2+Staining combines OsO₄, LC, and UA to contrast the specimen. It is generally believed that only the reduced form of OsO₄, osmium dioxide or osmium black, yields image contrast of biological specimens in the TEM. Reduction of OsO₄ takes place at the protein and fatty acid level, on sulfide groups and amino groups, and at double bonds, leading to stained proteins and lipids (Bahr 1954; Hayat 2000). Moreover, OsO₄ is proposed not only to have a role as a contrasting agent but also to work as an enhancer of LC staining. Lead is bound by OsO₄, resulting in improved staining of membranes. In addition, LC stains carbohydrates and RNA-containing structures in the cell. Finally, UA is employed in the 2+Staining because it associates with phosphate groups, yielding stained membranes and nucleic acids, the latter not being a target of OsO₄ (Hayat 2000). By combining these different contrasting agents in 2+Staining, a wide variety of structures present in the cell are contrasted, creating a rich image of the cellular content.

Prospects

The small effect of the 2+Staining on the immunofluorescence intensity is surprising. No compromise was made in the use and concentration of the heavy metal stains to retain the fluorescence signal. By creating a sufficient distance between the heavy metal stain and the fluorescent probe, influence of the stain on the fluorescence brightness is limited. We propose that by washing away the excess contrasting agent following the staining steps, only the metals bound to the biological structures remain on the specimen. Likely, the distance between the bound stain and the fluorescent probe is sufficient to prevent quenching from occurring. We showed that IF labeling intensity could be enhanced by introducing the Alexa–anti-Alexa step in the immunolabeling protocol. Brighter signals were observed following this amplified labeling compared with the conventional labeling method, irrespective of the subsequent staining protocol employed. We argue that the secondary and tertiary labeling steps, like in the amplified labeling procedure, further increase the distance between the fluorophore and the heavy metal stain.

Here, a novel method is introduced for the heavy metal staining of thawed cryosections prepared according to the Tokuyasu method. 2+Staining yields excellent positive contrast and shows little to no effect on the fluorescence signal of the immunolabeling on the section. In addition, an adapted labeling protocol was developed that yields a higher level of immunogold and immunofluorescence labeling by introducing an extra labeling step. The novel methods are very well suited for the preparation of Tokuyasu sections destined for imaging with both FM and TEM. Furthermore, the 2+Staining procedure makes the intermediate processing steps between imaging the specimen in the FM and the TEM, commonly employed in correlative microscopy, redundant. This not only speeds up the workflow, but completely avoids the introduction of structural alterations due to the extra preparation steps.