FLIPPER, a combinatorial probe for correlated live imaging and electron microscopy, allows identification and quantitative analysis of various cells and organelles
Ultrastructural examination of cells and tissues by electron microscopy (EM) yields detailed information on subcellular structures. However, EM is typically restricted to small fields of view at high magnification; this makes quantifying events in multiple large-area sample sections extremely difficult. Even when combining light microscopy (LM) with EM (correlated LM and EM: CLEM) to find areas of interest, the labeling of molecules is still a challenge. We present a new genetically encoded probe for CLEM, named “FLIPPER”, which facilitates quantitative analysis of ultrastructural features in cells. FLIPPER consists of a fluorescent protein (cyan, green, orange, or red) for LM visualization, fused to a peroxidase allowing visualization of targets at the EM level. The use of FLIPPER is straightforward and because the module is completely genetically encoded, cells can be optimally prepared for EM examination. We use FLIPPER to quantify cellular morphology at the EM level in cells expressing a normal and disease-causing point-mutant cell-surface protein called EpCAM (epithelial cell adhesion molecule). The mutant protein is retained in the endoplasmic reticulum (ER) and could therefore alter ER function and morphology. To reveal possible ER alterations, cells were co-transfected with color-coded full-length or mutant EpCAM and a FLIPPER targeted to the ER. CLEM examination of the mixed cell population allowed color-based cell identification, followed by an unbiased quantitative analysis of the ER ultrastructure by EM. Thus, FLIPPER combines bright fluorescent proteins optimized for live imaging with high sensitivity for EM labeling, thereby representing a promising tool for CLEM.
KeywordsLight microscopy Electron microscopy Correlated microscopy Probes FLIPPER Horseradish peroxidase
The study of the dynamics of molecules in living cells and the determination of their location at the ultrastructural level are important for understanding the molecular mechanisms underlying cell behavior. Genetically encoded fluorescent proteins (FPs) allow the imaging of molecules and organelles in living cells (Shaner et al. 2007) and fluorescent-based microscopic techniques have been developed that permit near-molecular-resolution imaging of biomolecules (Schermelleh et al. 2010). However, with these so-called super-resolution techniques, generally only one or two distinct molecules are identified and visualized. Electron microscopy (EM), on the other hand, reveals the ultrastructural content of cells down to the molecular level. Unfortunately, protein identification with EM in cellular systems can only be performed in fixed dead cells and techniques used for protein determination usually depend on destructive procedures. Correlated light microscopy (LM) and EM, termed “CLEM”, combines the benefits of these two microscopic techniques, resulting in additional synergistic insight (Brown and Verkade 2010; Giepmans 2008). To identify molecules by CLEM, dedicated probes have been designed that not only give spatiotemporal information in living cells but that can also be detected in their well-preserved ultrastructural surroundings by EM. Ideally, these probes should combine the benefits of (1) the ability to use fluorescence in live cells and (2) their visualization by EM without destructive treatment of cellular membranes.
CLEM is often just an overlay of images sequentially obtained by LM and EM showing ultrastructure. When fluorescence is still needed in EM, a compromise is typically made between fluorescent retention and ultrastructural preservation. Examples are the overlaying of green fluorescent protein (GFP) signals in samples for integrated microscopes (Peddie et al. 2014) or the need for ultrastructural context in so-called super-resolution fluorescence microscopy techniques, based both on the precise localization of proteins, such as for PALM (photoactivated localization microscopy; Betzig et al. 2006) and on laser-induced depletion of fluorescence, so-called STED (stimulated emission depletion; Watanabe et al. 2011). CLEM applications of these overlay techniques are still limited and the use of LM-based identification at EM resolution is less accurate than the use of EM-detectable probes.
The use of EM-detectable probes for CLEM was pioneered with Lucifer yellow to trace neurons (Maranto 1982). Following injection and LM examination, Lucifer yellow was photoconverted by intense illumination, resulting in the production of reactive oxygen species (ROS). When this technique is performed in the presence of diaminobenzidine (DAB), DAB polymers will form and precipitate in close (nanometer range) apposition to the fluorophore (Sosinsky et al. 2007). During subsequent EM preparation, which includes osmium fixation, electron-dense osmium will bind to these precipitates, which will then stand out in EM. Affinity-targeted photoconversion was first explored by using fluorescent ceramide trapped in the Golgi (Pagano et al. 1989) and later, proteins were also identified by specifically targeting the eosin fluorophore to the actin cytoskeleton (Deerinck et al. 1994). Immunotargeting for CLEM has not only been performed with small fluorophores but also with fluorescent particles, such as quantum dots. These nanoparticles are extremely bright and their electron-dense core is readily detectable by EM. Both pre-embedding (Giepmans et al. 2005; Nisman et al. 2004; see also Vu et al. 2015) and post-embedding (Izdebska et al. 2013) labeling for CLEM are feasible with quantum-dot-conjugated antibodies. Alternatively, both a fluorophore and gold can be coupled to an antibody. For pre-embedding labeling, nanogold that needs signal amplification for EM detection can be used, whereas for post-embedding labeling, colloidal gold combined with fluorophores might be an option (for reviews, see Sosinsky et al. 2007; Fabig et al. 2012). Several approaches have been developed to quantify particle-based detection in electron micrographs but the field of view is typically very limited to allow detection of the nanoparticles, which are usually immuno-targeted to proteins of interest (for a review, see Mayhew 2015 in this issue). The application of genetically encoded fluorescent tags, of which GFP (Tsien 1998) is the best example, revolutionized the imaging approaches possible. However, GFP is not generally suitable for photoconversion in order to generate DAB precipitates, although successful application has been shown (Grabenbauer et al. 2005; Grabenbauer 2012). Therefore, other genetically encoded techniques have been developed, starting with a tetracysteine motive with high affinity for biarsenical-conjugated dyes. One of these, named ReAsH, has proven to be extremely efficient for photoconversion by direct illumination (Gaietta et al. 2002). The affinity step, however, involves a balance between specific binding efficiency and a non-specific background. This obviously accounts for all affinity-based techniques, including the immunolabeling discussed above. In the case of the tetracysteine/ReAsH, specificity is increased by optimizing the affinity (Martin et al. 2005) and by using a Förster/fluorescence energy resonance transfer (Arai and Nagai 2013) by exciting genetically encoded GFP in tandem with the tetracysteine (Gaietta et al. 2006). In a subsequent search for a genuine genetically encoded CLEM tag, mini-SOG has been developed (Shu et al. 2011). The singlet oxygen quantum yield (SO-QY) of miniSOG was originally claimed to be higher than that of ReAsH (Shu et al. 2011). In recent studies, the SO-quantum yield (QY) has been argued to be lower (Pimenta et al. 2013; Ruiz-Gonzalez et al. 2013) but transformation-inducing irradiation leads to a higher QY (Ruiz-Gonzalez et al. 2013), explaining the applicability of miniSOG for CLEM (Boassa et al. 2013; Shu et al. 2011). In order to boost the fluorescence of the relatively dim mini-SOG, it has been used in tandem with FPs (Shu et al. 2011).
In addition to photoconversion, enzymatic reactions for the generation of DAB polymers have been exploited. For EM examination only, horseradish peroxidase (HRP) has been used to enhance the electron density of cells following micro-injection (Valtschanoff et al. 1992) or expression in the Golgi apparatus (Connolly et al. 1994) and to localize proteins in the endocytic pathway (Luik et al. 2006; Sunio et al. 1999). More recently, a genetically encoded EM reporter allowing identification of features in the cytoplasm has been developed based on ascorbate peroxidase (Martell et al. 2012) and optimized for higher activity (Lam et al. 2015). This homodimer, called APEX, has been chosen because of its independence of disulfide bridge formation and calcium and will therefore, in principle, allow the detection in both reducing and oxidizing compartments. To prevent dimerization, monomeric versions have been created by introducing specific mutations based on the crystal structure, although this has been achieved at the expense of activity. For CLEM examination, either immunofluorescence or a tandem binding between APEX and mCherry or GFP has been used (Martell et al. 2012).
We designed a combinatorial probe for CLEM that uses the best of two worlds by combining optimized FPs with highly enzymatically active HRP. With this probe, called FLIPPER (fluorescent indicator and peroxidase for precipitation with EM resolution), several organelle systems can be detected by both LM and EM. With FLIPPER, complete tissue-culture dishes can be straightforwardly labeled for both LM and EM. Here, we combine this strategy with mosaic LM and large-scale EM (Faas et al. 2012; Kuwajima et al. 2013; Ravelli et al. 2013), known as nanotomy, and present quantitative imaging for EM with FLIPPERs.
Materials and methods
Green fluorescent Golgi-FLIPPER was cloned by replacing the tetracysteine from ManII-GFP-4C (Gaietta et al. 2006) by HRP amplified by the polymerase chain reaction (PCR) from HRP-Stim1 (Luik et al. 2006), a kind gift of R.S. Lewis (Stanford University Medical School), with primers F: GATCGGATCCATGCAGTTAACCCCTACA, and R: GATCGCGGCCGCCTCGAGTTATCCTCCTCCCCTAGAGTTG by using the BamHI and NotI sites. With F: AGTCGAATTCATGGTGAGCAAGGGC and R: AGTCGGATCCCTTGTACAGCTCGTC, mOrange2 (Shaner et al. 2008), mCherry (Shaner et al. 2004; both kindly provided by R.Y. Tsien, University of San Diego, California), or mTurquoise2 (a kind gift of T.W.J. Gadella, University of Amsterdam; Goedhart et al. 2012) were amplified and inserted into the construct above replacing enhanced GFP (EGFP) by using EcoRI and BamHI. For the endoplasmic reticulum (ER)-FLIPPERs, pDsRed2-ER (632409; clontech) was used. First, the DsRed2-ER fragment was ligated into the NheI/XhoI sites of pCDNA3-ManII-GFP-4C plasmid (Gaietta et al. 2006). Next, various colors of FLIPPERs were ligated in frame with the calreticulin fragment at the AgeI site by using primers F: GATCACCGGTCGATGGTGAGCAAGGGCGA and R: GATCCTCGAGTTACAGCTCGTCCTTTCCTCCTCCCCTAGAGTTG, ensuring a C-terminal KDEL sequence. The EcoRI—NotI fragment of Golgi-FLIPPER was ligated in frame with the EcoRI site of EpCAM (epithelial cell adhesion molecule) at amino acid 263 by using F: TAATACGACTCACTATAGGGA and R: GATCCTCGAGTTAACCCTGCATTGAGAATTC on EpCAM, resulting in EpEx-FLIPPER. All EpCAM plasmids used have been described previously (Schnell et al. 2013).
Cell culture, transfection and confocal microscopy
Cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 5 % fetal calf serum (FCS) and penicillin/streptomycin and cultured at 37 °C in the presence of 5 % CO2. HEK293T cells were transfected with Fugene-6 (Promega) according to the manufacturer’s protocol. For multicolor FLIPPER experiments, cells were cotransfected at day 0 with either FLIPPER-mOrange2 together with EpCAM-GFP or FLIPPER-mOrange2 with mutant EpCAM (C66Y) labeled with mCherry (EpCAM[C66Y]-mCherry). At day 1, cells were trypsinized, mixed in a 1:1 ratio and reseeded in glass-bottom Petri dishes (MatTek). On day 2, cells were imaged live under physiological conditions with a confocal system (Zeiss LSM 780). Brightness and contrast were adjusted for presentation purposes. For spectral analysis, fluorescence spectra were recorded in HeLa cells transfected with FLIPPER plasmids by using PPEI. Fluorescence intensities at various wavelengths were measured by using the gamma mode on a confocal scanning system (Zeiss LSM 780, Plan-Neofluar 63×/1.3 Imm Korr DIC M27 lens). Emmision spectra were recorded in lambda mode creating intensity profiles between 416 and 687 nm in approximately 9 nm intervals. Filters and beam splitters used for mTurquoise2, GFP, mOrange2 and mCherry were MBS 458, MBS 458/514, MBS 488 and MBS 488/561, respectively, with a 412–691 filter. Mean image intensities were measured in the imaging data by using Image J Fiji (plot z-axis profile) and normalized to the highest fluorescence intensity measured for the respective fluorescent protein (100 %), yielding relative fluorescence intensity (RFI) plotted as in the graph shown.
By using an automated stage, large high-magnification fields of view were generated by automatically stitching (mosaic) 8 × 8 images recorded in tile scan mode (LCI Plan-Neofluar 63×/1.3 Imm Korr DIC M27). GFP, mOrange2 and mCherry were imaged by using 488-, 514- and 594-nm lasers, respectively, while recording non-overlapping emission. After removal of the culture medium, the cells were fixed with 4 % paraformaldehyde/0.1 % glutaraldehyde in 0.1 M cacodylate buffer for 20 min at room temperature, washed with 0.1 M cacodylate buffer and imaged again in this buffer. With the 20× objective, 5 × 5 tiles were imaged at 1024 × 1024 pixel resolution resulting in an area of approximately 2 × 2 mm.
After the imaging step, cells were washed in phosphate-buffered saline (PBS). HRP visualization by DAB polymerization was performed by using standard protocols. Briefly, 5 mg DAB was dissolved in 10 ml PBS or TRIS-buffered saline; the solution was filtered and, after being mixed with H2O2 (3 μl 30 %), was added to the cells. Typically after 5–10 min, depending on the visibility of the reaction product, the reaction was stopped by washing away the DAB and the embedding procedure was carried out.
After the DAB reaction, cells were incubated with 1 % OsO4 in 0.1 M cacodylate buffer (30 min, 4 °C). After being washed with water, the cells were dehydrated through an increasing graded ethanol series and left overnight at room temperature in a 1:1 mixture of ethanol/Epon, which was replaced by pure Epon (3×) and, in the final step, polymerized at 58 °C. The cover glass of the imaging dish was removed with either liquid nitrogen or by using hydrogen fluoride. The imaged area was selected under a stereo microscope by using intrinsic marks. The relevant area containing the cells of interest on the Epon block was sawn out. Either ultrathin (60 nm) sections were obtained and collected on copper grids for classical transmission EM (TEM), or semithin (0.5 μm) sections were cut and collected on a silicon wafer for large-area EM by using the backscatter detector (BSD). In some cases, the sections were contrasted with uranyl acetate and Reynolds lead citrate following standard procedures. TEM images were recorded by using a FEI CM100 at 80 KV with a Morada camera (Olympus-SIS). Unequal illumination of original EM images was corrected in Adobe Photoshop by applying a gradient levels adjustment layer. BSD images were recorded by using a Zeiss Supra55 Scanning Microscope at 3 KV. Large-area scans were generated by using the external scan generator ATLAS (Fibics Canada) and multiple tiles were stitched by using a VE-viewer (Fibics) and exported either as a high-resolution html file or a single TIF.
The region imaged by LM before processing for EM was identified as previously described (Sjollema et al. 2012). The imaged area was marked under a stereo microscope by making scratches in the Epon with a scalpel. In Adobe Photoshop, the LM image was copied as a layer into the EM image and made 50 % transparent. Transformation of the LM image is necessary to match it to the larger scale of the EM image. Alignment was carried out with the aid of the intrinsic marks in combination with the shape of the cells in both the differential interference contrast (DIC) image and the EM image. Subsequently, differently color-coded cells, representing different experimental conditions in the same experiment, could be correlated to their respective EM ultrastructure. For ER measurements, selected ultrastructural images of cells with either full-length EpCAM (FL) or EpCAM-C66Y were analyzed by using ImageJ Fiji software. In 10 different cells expressing either FL or mutant EpCAM, ER thickness was measured at 10 different positions per cell.
Results and discussion
We used FLIPPERs to address whether mutations in EpCAM lead to ER dilation. Mutations in the EpCAM gene have been identified as the cause for congenital tufting enteropathy (CTE), a disease presenting with lethal diarrhea attributable to abnormalities in the intestinal epithelium in affected newborns. Previously, we found that all EpCAM mutations in CTE patients led to either secretion of the protein or to retention and accumulation in the ER (Schnell et al. 2013). We hypothesized that ER retention of EpCAM caused ER stress. In some instances, ER stress might result in the widening of the ER lumen (Ravelli et al. 2013).
Concluding remarks and future outlook
Several probes have now been developed, validated and implemented allowing protein identification at both LM and EM levels. Genetic encoding ensures the maximal specificity of targeting of the protein of interest by use of chimeras. Once the DNA has been delivered, cells do not have to be perturbed to pass large molecules such as antibodies inside cells and organelles and thus, the ultrastructure is superiorly preserved compared with that following immune affinity-based techniques, a hurdle for the localization of proteins by EM. The majority of genetically encoded probes that can be detected at both LM and EM levels are based on continuously improving fluorescent proteins and/or osmiophilic DAB precipitates (Table 1).
Dedicated genetically encoded probes for correlated light microscopy and electron microscopy (CLEM) based on diaminobenzidine (DAB) deposition. Whereas FLIPPER (fluorescent indicator and peroxidase for precipitation with EM resolution) has been specifically developed for multispectral analysis (this report), other fusions with fluorescent proteins (FPs) have been introduced to enable (APEX; Martell et al. 2012; Lam et al. 2015) or increase (miniSOG; Shu et al. 2011) fluorescence or to obtain more specific photoconversion by using fluorescence energy resonance transfer (FRET; ReAsH; Gaietta et al. 2006)
Includes affinity-step in labeling, very small
Relatively dim, small
Weaker than horseradish peroxidase (HRP), cytoplasmic activity
Secretory route only, bright and active
The key is to create CLEM probes that are suitable for general use and that are, thus, widely applicable, similar to the achievements for FPs, starting with GFP two decades ago. The combinatorial tag allows not only to perform live-cell imaging but also to construct chimeric proteins including initial fast LM validation by using routine methods. The added values of genetically encoded probes for CLEM are (1) highly specific targeting; (2) high-quality preservation of the ultrastructure; (3) cheap amplification and easy sharing within the scientific community; (4) pre-embedding labeling suitable for three-dimensional EM; (5) compatibility with labeling in living cells in vitro and in vivo, even including animals (Shu et al. 2011). Further efforts will be needed to engineer a cytoplasmic FLIPPER-like module or APEX variants with higher activity and, by extension, higher sensitivity. A future alternative to direct tagging of proteins of interest might be the fusion of genetically encoded CLEM probes to genetically encoded single-chain antibodies to targets (Rothbauer et al. 2006). Finally, the parallel development of probes based on various principles will help current developments in simultaneous labeling of multiple structures in EM (“multi-color EM”) and will allow better quantitative analysis of ultrastructural data.
We thank our laboratory members, especially P. de Boer, K. Seerangaiyan and R. Hoffmann, for feedback and technical assistance and T.W.J. Gadella (Goedhart et al. 2012), R.Y. Tsien and R.C. Lewis for providing plasmids.
- Boassa D, Berlanga ML, Yang MA, Terada M, Hu J, Bushong EA, Hwang M, Masliah E, George JM, Ellisman MH (2013) Mapping the subcellular distribution of alpha-synuclein in neurons using genetically encoded probes for correlated light and electron microscopy: implications for Parkinson’s disease pathogenesis. J Neurosci 33:2605–2615CrossRefPubMedCentralPubMedGoogle Scholar
- Deerinck TJ, Martone ME, Lev-Ram V, Green DP, Tsien RY, Spector DL, Huang S, Ellisman MH (1994) Fluorescence photooxidation with eosin: a method for high resolution immunolocalization and in situ hybridization detection for light and electron microscopy. J Cell Biol 126:901–910CrossRefPubMedGoogle Scholar
- Faas FG, Avramut MC, Berg BM van den, Mommaas AM, Koster AJ, Ravelli RB (2012) Virtual nanoscopy: generation of ultra-large high resolution electron microscopy maps. J Cell Biol 198:457–469Google Scholar
- Gaietta GM, Giepmans BN, Deerinck TJ, Smith WB, Ngan L, Llopes J, Adams SR, Tsien RY, Ellisman MH (2006) Golgi twins in late mitosis revealed by genetically encoded tags for live cell imaging and correlated electron microscopy. Proc Natl Acad Sci U S A 103:17777–177782CrossRefPubMedCentralPubMedGoogle Scholar
- Goedhart J, Stetten D von, Noirclerc-Savoye M, Lelimousin M, Joosen L, Hink MA, Weeren L van, Gadella TW Jr, Royant A (2012) Structure-guided evolution of cyan fluorescent proteins towards a quantum yield of 93 %. Nat Commun 3:751Google Scholar
- Igdoura SA, Herscovics A, Lal A, Moremen KW, Morales CR, Hermo L (1999) Alpha-mannosidases involved in N-glycan processing show cell specificity and distinct subcompartmentalization within the Golgi apparatus of cells in the testis and epididymis. Eur J Cell Biol 78:441–452CrossRefPubMedGoogle Scholar
- Mayhew (2015) Quantitative immunocytochemistry at the ultrastructural level: a stereology-based approach to molecular nanomorphomics. Cell Tissue Res (this issue)Google Scholar
- Peddie CJ, Blight K, Wilson E, Melia C, Marrison J, Carzaniga R, Domart MC, O’Toole P, Larijani B, Collinson LM (2014) Correlative and integrated light and electron microscopy of in-resin GFP fluorescence, used to localise diacylglycerol in mammalian cells. Ultramicroscopy 143:3–14CrossRefPubMedCentralPubMedGoogle Scholar
- Vu T, Lam WY, Hatch E, Lidke D (2015) Quantum dots for quantitative imaging: from single molecules to tissue. Cell Tissue Res (this issue)Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution License, which permits any use, distribution and reproduction in any medium, provided the original author(s) and the source are credited.