|Home | About | Journals | Submit | Contact Us | Français|
Electron microscopy (EM) is the standard method for imaging cellular structures with nanometer resolution, but existing genetic tags are inactive in most cellular compartments1 or require light and are difficult to use2. Here we report the development of a simple and robust EM genetic tag, called “APEX,” that is active in all cellular compartments and does not require light. APEX is a monomeric 28 kDa peroxidase that withstands strong EM fixation to give excellent ultrastructural preservation. We demonstrate the utility of APEX for high-resolution EM imaging of a variety of mammalian organelles and specific proteins. We also fused APEX to the N- or C-terminus of the mitochondrial calcium uniporter (MCU), a newly identified channel whose topology is disputed3,4. MCU-APEX and APEX-MCU give EM contrast exclusively in the mitochondrial matrix, suggesting that both the N-and C-termini of MCU face the matrix.
EM-level localization of proteins has been integral to elucidating many cellular processes. Antibody-based detection5–8, the most common approach, requires either permeablizing treatments, which degrade cellular ultrastructure9,10, or ultracryosectioning of sucrose cryoprotected samples and immunostaining of each individual section11. Genetically encoded tags do not require permeabilization, technically demanding ultracryosectioning, or staining of individual sections because they localize to the site of interest prior to fixation, but the existing genetic tags for EM have limitations. Horseradish peroxidase (HRP) is a sensitive12 tag that catalyzes the H2O2-dependent polymerization of 3,3′-diaminobenzidine (DAB) into a localized precipitate that gives EM contrast after treatment with OsO4 13,14. However, HRP is inactive when expressed in the mammalian cytosol1, likely because its four structurally-essential disulfide bonds and two Ca2+ binding sites do not form in reducing and Ca2+-scarce environments. A genetically encoded tag called “miniSOG”2 provides EM contrast in all cellular compartments via light-dependent generation of 1O2, which produces a DAB precipitate, but the requirement for light can be limiting when large or thick samples need to be stained for EM. ReAsH is another light-dependent tag, but its nonspecific labeling and the challenge of probe delivery into tissue limit its utility15,16.
A re-engineered version of HRP with activity in the cytosol would overcome the limitations of existing methods17, but our attempts to remove its disulfide bonds abolished activity (data not shown). We instead searched for peroxidases that are naturally active in reducing environments. Ascorbate peroxidase (APX) is a class I cytosolic plant peroxidase that lacks disulfide bonds and calcium ions18. At 28 kDa, APX is also ~40% smaller than HRP. However, APX has not previously been tested in mammalian cells, and its natural substrate, ascorbate, has a structure that is very different from DAB (Fig. 1a). Furthermore, wild-type (wt) APX is a constitutive homodimer19, a problem since oligomeric tags can perturb a protein’s natural localization and function20.
We first tested expression and activity of wt APX, using immunostaining and Amplex UltraRed, a fluorogenic peroxidase substrate, in three cellular compartments: cytosol, mitochondrial matrix, and endoplasmic reticulum (ER). For comparison, we tested HRP in the same locations. APX expressed well and was active in all compartments, whereas genetically-targeted HRP was active only in the ER, as expected (Supplementary Fig. 1).
We proceeded to engineer a monomeric APX. Guided by the crystal structure18 and building upon an earlier mutagenesis study19, we individually mutated negatively charged or neutral residues at the dimer interface to Lys to introduce repulsive interactions (Fig. 1b). We also tested three mutants (K14D, K31S, and A233D) on the basis of sequence alignment to an isoform of maize APX that was reported to be monomeric21. Gel filtration chromatography (GFC) was used to assess oligomerization (Fig. 1c and Supplementary Fig. 2). Five mutants were significantly more monomeric than wt, but they all displayed weak dimerization, so we generated double and triple mutants using permutations of the five promising single mutations. APX K14D/E112K exhibited the most monomericity while still expressing well in E. coli and forming minimal high molecular weight aggregates.
We compared the oligomerization of APX K14D/E112K (=mAPX, for monomeric APX) to that of wt APX, miniSOG, and several fluorescent proteins (mApple, EYFP, and mEos2) across a range of concentrations (Fig. 1d). mAPX was not as monomeric as mApple, but it was comparable to EYFP and much more monomeric than both wt APX and mEos220. We fused both mAPX and wt APX to connexin43 (Cx43), a protein that is sensitive to oligomeric tags (Fig. 1e)20,22. Whereas wt APX gave rise to abnormal retention inside cells, both Cx43-mAPX and Cx43-GFP-mAPX localized properly to gap junctions, indicating that despite its weak residual dimerization at high concentrations, mAPX does not disrupt Cx43 trafficking.
We compared the activity of wt APX and mAPX in mammalian cells. mAPX and wt APX were equally active in vitro and comparably active in some cellular contexts, such as the mitochondrial matrix, which houses the last step of heme biosynthesis (Fig. 2). However, when expressed throughout the cytosol via fusion to a nuclear export sequence (NES), mAPX was much less active than wt APX toward both DAB (Fig. 2) and Amplex UltraRed (Supplementary Fig. 3a–b). Culturing the cells with heme after transfection significantly boosted the activity of mAPX-NES. We reasoned that the poor activity of mAPX-NES is likely due to less efficient heme incorporation in the cellular context, perhaps a result of decreased thermal stability associated with monomerization19.
In order to improve the performance of mAPX in mammalian cells, we engineered a monomeric APX with improved activity toward DAB in the hope that faster kinetics could compensate for diminished heme binding. We compared the substrate binding site of APX to that of HRP, which has faster kinetics and broader aromatic substrate tolerance23. In order to make APX more HRP-like, we designed seven mutants of APX that incorporated mostly aromatic residues into the active site (Fig. 2a). We employed a colorimetric assay using guaiacol, a prototypical aromatic peroxidase substrate, to determine Michaelis-Menten kinetic parameters 24 (Fig. 2b and Supplementary Fig. 4). All of the mutations improved APX activity, with the most promising mutant showing >25-fold enhancement over wt APX in terms of kcat/KM. The most HRP-like mutant (W41F/G69F/D133A/T135F/K136F) was also the most active, validating our engineering approach.
To generate a monomeric APX with improved activity, we added the two mAPX mutations (K14D/E112K) to each of our activity-enhanced mutants. The monomerized versions of some activity-enhanced mutants performed poorly in several cellular contexts (Supplementary Fig. 5), but mAPX + W41F expressed well in E. coli, displayed minimal high molecular weight (MW) aggregation, and localized comparably to mAPX in cells. We therefore selected mAPX + W41F (K14D/E112K/W41F), which we call “APEX” (for Enhanced APX), as our optimized reporter for subsequent experiments. Importantly, APEX-NES gave markedly stronger staining in cells than mAPX-NES, using both DAB (Fig. 2c) and Amplex UltraRed (Supplementary Fig. 3) as substrates. APEX is as monomeric as mAPX in vitro (Fig. 1d) and behaves well as a fusion to Cx43, both alone and in tandem with GFP (Fig. 1e).
We proceeded to test the utility of APEX for EM imaging in mammalian cells. We first used APEX to stain two cellular compartments with distinct EM signatures: the mitochondrial matrix and the ER lumen. Mitochondria expressing APEX showed strong contrast and well-defined cristae structures relative to non-expressing mitochondria (Fig. 3a). When localized to the ER lumen, APEX provided dense staining throughout the ER tubular network (Fig. 3a). By comparison, the ER lumen of untransfected cells gave weak contrast.
We next performed EM imaging of APEX fused to three mammalian proteins: histone 2B (H2B), vimentin, and connexin43 (Cx43). APEX-H2B revealed detailed chromatin structures, both along the nuclear envelope and on the periphery of nucleoli (Fig. 3a). These features were not discernible in untransfected cells. APEX-H2B properly incorporated into chromatin throughout all stages of mitosis (Fig. 3a and Supplementary Fig. 6), demonstrating that the weak dimerization of APEX is non-perturbing in this context. Vimentin-APEX highlighted intermediate filaments throughout the cell, and subfilamentous repeating densities could be discerned (Fig. 3a), underscoring the excellent contrast and high resolution afforded by APEX. Cx43-GFP-APEX gave high-resolution images of gap junction plaques, revealed by both thin sections and electron tomography (Fig. 3a and Supplementary Movie 1). In some cases, gap junction plaques were closely apposed to tubular segments of ER membrane25, but the DAB reaction product exhibited minimal spread even in the absence of membrane enclosure (Fig. 3a).
We used Cx43-GFP-APEX for correlated light microscopy (LM) and EM9 by imaging GFP fluorescence in fixed cells, followed by DAB deposition and EM processing (Fig. 3a). Similar constructs utilizing mCherry instead of GFP, in two different orientations, also formed gap junctions and gave bright fluorescence and strong DAB staining (Supplementary Fig. 7).
Because APEX remains active after membrane-preserving cell fixation and generates a minimally-diffusive reaction product that does not cross membranes, we envisioned that APEX could be useful for determining the topology of membrane proteins. The mitochondrial calcium uniporter (MCU) is a recently-discovered channel protein responsible for calcium uptake into the mitochondrion 3,4. Each MCU subunit contains two transmembrane segments that span the inner mitochondrial membrane. Using a protease accessibility assay on purified mitochondria, Baughman et. al. proposed that both the N- and C-termini of MCU protrude into the mitochondrial matrix3. In contrast, DeStefani et. al. reported the opposite topology, with both the N- and C-termini facing the intermembrane space (IMS)4 (Fig. 3b). To resolve this discrepancy, we constructed N- and C-terminal fusions of APEX to MCU, performed DAB staining, and imaged the samples by EM. Consistent with the model of Baughman et. al., both constructs gave clear EM staining in the matrix, but not the IMS (Fig. 3c and Supplementary Fig. 8). Separate controls showed that APEX is active in the IMS (Supplementary Fig. 8).
Interestingly, the matrix stain was strongest for both constructs at sites where cristae were closely stacked (within ~25 nm). To investigate whether this localization is an artifact of APEX dimerization at high concentrations, we repeated the experiment using a tandem dimer of APXW41F (description and characterization in Supplementary Fig. 9) as well as miniSOG in place of APEX. Identical EM staining patterns were obtained for all constructs (Supplementary Fig. 8), although we still cannot completely rule out artifactual dimerization because miniSOG, like APEX, shows weak oligomerization at concentrations above ~20 μM (Fig. 1d). We also repeated the experiment with stable HeLa cells expressing MCU-APEX or APEX-MCU in place of endogenous MCU and obtained the same result (data not shown). In a functional assay for Ca2+ uptake, all fusions of MCU to APEX, tdAPXW41F, and miniSOG performed comparably to MCU appended to a small Flag epitope tag (Fig. 3d), indicating that none of these tags perturb function. The MCU topology revealed by APEX EM imaging has implications for MCU interaction partners and regulatory proteins, as it is now clear that most of the soluble portions of the protein reside in the matrix, rather than the IMS.
In summary, we have developed APEX, an engineered variant of APX that is monomeric, can be targeted through transfection, and provides strong EM contrast in all cellular compartments thus far tested, including the cytosol, where HRP is inactive as a genetic tag. Although HRP-antibody conjugates provide sensitive EM staining of cytosolic targets, the permeablizing reagents required for immunostaining irreversibly damage cellular ultrastructure and allow the DAB reaction product to spread. In contrast, cells expressing APEX can be strongly fixed with 2% glutaraldehyde in the absence of detergent, and the reaction product is generated at 4 °C, enabling optimal preservation of ultrastructure and tight localization of the EM stain.
The APEX staining protocol is straightforward, requiring only the addition of DAB and H2O2. Photosensitizers such as miniSOG require light, which limits their utility in complex tissues and whole organisms, whereas APEX should be fully applicable for large-scale EM. Additionally, since APEX fusions provide a high-resolution 3-D reaction product, they are directly applicable to electron tomography (as demonstrated here), serial section analyses, and serial blockface scanning EM. MiniSOG has two advantages over APEX: it is smaller (12 kDa vs 28 kDa) and possesses intrinsic fluorescence, which facilitates correlated LM and EM. However, miniSOG’s fluorescence is weak, and correlated imaging is best performed with miniSOG-fluorescent protein fusions26. Here we show that APEX also tolerates fusion to fluorescent proteins on either terminus, making correlated LM and EM straightforward. Because the APEX reaction conditions are orthogonal to miniSOG reaction conditions, it may be possible to utilize the two tags in combination to generate differentiable depositions of DAB for “two-color” EM.
Wt APX is a strong homodimer, but we have engineered APEX to be predominantly monomeric. APEX still shows residual dimerization at high concentrations, comparable to EYFP and miniSOG. Despite its weak dimerization, APEX behaves well as a fusion to numerous proteins that are sensitive to oligomerization. We did observe perturbation of ER structure when APEX was overexpressed on the ER membrane facing the cytosol (data not shown), similar to observations published for EGFP constructs27. To avoid dimerization artifacts, we recommend expressing APEX at the minimum level required to obtain EM contrast.
One potential limitation of APEX is its requirement for heme. In many cell types and compartments, the endogenous heme available is sufficient to confer high APEX activity; all images in this work, except those in Figure 2c where indicated, were acquired without exogenous heme supplied. In heme-poor environments, however, such as the cytosol of some cell types, the activity of APEX toward DAB may be insufficient for EM, perhaps a result of diminished heme binding resulting from decreased stability19. In these cases, we recommend using APXW41F instead of APEX if a dimeric tag can be tolerated, or supplementing the cell media with purified heme prior to fixation to increase the heme occupancy of APEX.
Beyond EM, APEX should have utility for a variety of other imaging and biotechnological applications, such as those for which HRP is currently used28. For example, the activity of APEX toward numerous aromatic substrates (DAB, Amplex Red, and guaiacol are illustrated here) provides the opportunity not only for EM contrast, but also for colorimetric, fluorescent, and chemiluminescent readouts.
The pTRC99A bacterial expression vector encoding pea cytosolic APX with a His6-tag appended to its N-terminus has been described previously32. Mutants of APX were generated using QuikChange mutagenesis (Stratagene) or overlap extension PCR33. Some mutants were subcloned into the pET21a bacterial expression vector (EMD Chemicals) between the NdeI and XhoI restriction sites, which enabled higher protein yield than pTRC. Fusions of APX to mammalian proteins and localization signals were generated using standard restriction cloning methods. Supplementary Table 1 presents an overview of all mammalian expression constructs used in this study. For most of these mammalian expression constructs, the original pea APX gene was used. For some constructs, however, we used a pea APX gene with humanized codons, synthesized by GenScript. We avoided the use of HA epitope tags for immunofluorescence detection because the HA tag, being tyrosine rich, is oxidatively damaged by APX in the presence of H2O2.
Purified APX, miniSOG, or fluorescent proteins were diluted to the desired concentrations using chilled (4 °C) phosphate buffered saline (PBS, 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2.0 mM KH2PO4, pH 7.4) and then allowed to incubate at room temperature for 30 min to overnight. APX was not prone to aggregation on the timescale of hours at room temperature. Samples (100 μL) were run on a Waters HPLC system over a Superdex 75 10/300 column (GE Healthcare) that had been pre-equilibrated in PBS, pH 7.4. Samples were run isocratically at room temperature in PBS at a flow rate of 0.5 mL/min. Elution of APX was monitored at 280 nm and 405 nm (heme absorbance). Empower software (Waters) was used for analysis of elution profiles. Apparent MWs were determined by referencing low molecular weight standards (GE) (Supplementary Fig. 2). For proteins that exhibited no significant change in apparent MW between 0.25 and 1 μM, the apparent MW of the monomer was estimated to be equal to the apparent MW at 0.25 μM. This estimation was applied for all proteins in Figure 1d except for wt APX, which was predominantly dimeric at 0.25 μM. The apparent MW of the monomer for wt APX was assumed to be equal to the apparent MW of mAPX at 0.25 μM.
We attempted to establish an assay to monitor the APX-catalyzed polymerization of DAB in vitro, but could not determine the kinetic parameters for this complex process. Amplex UltraRed was also not utilized for kinetic measurements since the KM of APX toward aromatic substrates is typically between 1 and 10 mM, and solutions of Amplex UltraRed cannot be economically prepared at these high concentrations. We utilized guaiacol as the substrate to measure peroxidase kinetics instead. Good activity toward guaiacol generally correlated well with robust polymerization of DAB in vitro (data not shown). Guaiacol assays were performed on a Nanodrop 2000c UV-vis spectrophotometer (Thermo Scientific) using its cuvette reader. HRP (Sigma) was prepared as a 10 μM stock in PBS, flash frozen and stored at −80 °C, and thawed immediately before use. Guaiacol peroxidase activity was measured according to the protocol of Lad et. al.24 Briefly, guaiacol (Sigma) was diluted in room temperature PBS, pH 7.4, to concentrations ranging from 0.25 to 30 mM. Solutions were vortexed thoroughly to ensure the guaiacol was completely dissolved. H2O2 was added to a final concentration of 90 μM, followed by addition of APX to a final concentration of 20 nM. For HRP, when 90 μM H2O2 was used, all of the H2O2 was consumed so quickly that the initial rate of guaiacol turnover could not be determined on our spectrophotometer. Therefore, 10 mM H2O2 was used for HRP. Under these conditions, the initial rate of turnover was maintained for several seconds and could be readily determined. Oxidation to tetraguaiacol was monitored by absorbance at 470 nm (ε470 = 22 × 103 M−1cm−1)24. kcat and KM were calculated by nonlinear regression fitting to the Michaelis-Menten equation using OriginPro. All values are reported as the mean ± standard deviation for 2–3 independent Michaelis-Menten fittings, with each fitting using at least 5 values of substrate concentration ranging between 0.25 and 30 mM.
HEK293T, HeLa, and COS-7 cells were cultured as a monolayer in Modified Eagle’s Medium (MEM, Cellgro) supplemented with 10% fetal bovine serum (PAA Laboratores). Cells were maintained at 37 °C under an atmosphere of 5% CO2. For EM imaging experiments, cells were grown on poly-d-lysine coated glass-bottom dishes (P35GC-0–14-C, MatTek Corp.). To improve the adherence of HEK293T cells, the dishes were also pretreated with 50 μg/mL fibronectin (Millipore) for 1 hour at 37 °C before cell plating. Cells were transfected at 60–90% confluence using Lipofectamine2000 (Life Technologies), typically with 0.7 uL Lipofectamine and 100 ng plasmid per 300,000 cells. Cells were labeled and/or fixed 18–24 h after transfection. For incubation of HEK293T cells with heme (Figure 2c), the media was supplemented with a heme-bovine serum albumin (BSA) complex (3 or 7 μM)34 for 16–24 h, after which time the media was removed, cells were rinsed 1× with non-heme-containing media, then processed for imaging. Note that all images, except those in Figure 2c where indicated, were acquired without exogenous heme supplied. APEX generally affords strong EM staining with no requirement for exogenous heme addition, so we recommend supplementing the media with heme only in cases where the DAB stain is not sufficiently strong for EM without heme addition.
Transfected cells were fixed using room temperature 2% glutaraldehyde (Electron Microscopy Sciences) in buffer (100 mM sodium cacodylate with 2 mM CaCl2, pH 7.4), then quickly moved to ice. Cells were kept between 0 and 4 °C for all subsequent steps until resin infiltration. After 30–60 min, cells were rinsed 5 × 2 min in chilled buffer, then treated for 5 min in buffer containing 20 mM glycine to quench unreacted glutaraldehyde, followed by 5 × 2 min rinses in chilled buffer. A freshly-diluted solution of 0.5 mg/mL (1.4 mM) 3,3′-diaminobenzidine (DAB) tetrahydrochloride or the DAB free base (Sigma) dissolved in HCl were combined with 0.03% (v/v) (10 mM) H2O2 in chilled buffer, and the solution was added to cells for 1 to 15 min, depending on the sample. A summary of cell types and DAB reaction times is presented in Supplementary Table 2. The generation of reaction product could be monitored by transmitted light microscopy. To halt the reaction, the DAB solution was removed, and cells were rinsed 5 × 2 min with chilled buffer. Post-fixation staining was performed with 2% osmium tetroxide (Electron Microscopy Sciences) for 30 min in chilled buffer. Cells were rinsed 5 × 2 min in chilled distilled water, then placed in chilled 2% aqueous uranyl acetate (Electron Microscopy Sciences) overnight. The samples were then dehydrated in a cold graded ethanol series (20%, 50%, 70%, 90%, 100%, 100%) 2 min each, rinsed once in room temperature anhydrous ethanol to avoid condensation, and infiltrated in Durcupan ACM resin (Electron Microscopy Sciences) using 1:1 (v/v) anhydrous ethanol and resin for 30 min, then 100% resin 2 × 1 h, then into fresh resin and polymerized in a vacuum oven at 60 °C for 48 h.
DAB-stained areas of embedded cultured cells were identified by transmitted light, and the areas of interest were sawed out using a jeweler’s saw and mounted on dummy acrylic blocks with cyanoacrylic adhesive (Krazy Glue, Elmer’s Products). The coverslip was carefully removed, the block trimmed, and ultrathin (80 nm thick) sections were cut using an ultramicrotome (Leica Ultracut UTC6). Electron micrographs recorded using a JEOL 1200 TEM operating at 80 keV.
Transfected cells on glass coverslips were moved to ice, then treated with a solution of 50 μM Amplex UltraRed (Molecular Probes) with 0.02% (6.7 mM) H2O2 in Dulbecco’s phosphate-buffered saline (DPBS). The Amplex UltraRed solution was freshly diluted from a 10 mM stock in dimethyl sulfoxide. After 5 to 30 min, depending on the sample, the Amplex UltraRed solution was removed and replaced with DPBS. In some cases, cells were imaged live. Strong resorufin signal was present in cells expressing APX, although the signal was not tightly localized to the site of origin. Alternatively, Amplex UltraRed-stained cells were fixed using freshly depolymerized 4% formaldehyde (Electron Microscopy Sciences) in PBS for 30 min on ice, rinsed 5 × 2 min in chilled PBS, then treated with methanol for 5–10 min at −20 °C. Samples were blocked using 1% (v/v) bovine serum albumin (BSA, Fisher Scientific) in PBS at 4 °C for 30 min, then treated overnight at 4 °C with a 1:500 dilution of mouse-anti-Flag antibody (Agilent) or chicken-anti-c-myc antibody (Life Technologies) in PBS with 1% BSA. Cells were rinsed 4 ×5 min in PBS, then treated with a 1:750 dilution of AlexaFluor 488 goat anti-mouse IgG or AlexaFluor 568 goat anti-chicken IgG (Life Technologies) for 15 min at 4 °C. Cells were rinsed 4 × 5 min in PBS, then imaged by confocal microscopy. At this point, much of the resorufin had been washed away, but the remaining label was more closely localized to the site of origin.
We thank Amy Keating for use of her GFC system and Andrea Thor, Angela Cone, and Masako Terada for assistance with electron tomography. The IMS-APEX EM data were obtained by Hyun-Woo Rhee, Peng Zou, J.D.M., and Eliza Vasile (Microscopy and Imaging Core Facility, Koch Institute at MIT). Chaysith Uttamapinant (MIT) and Hunter Fraser (Stanford) provided helpful feedback on the manuscript. Funding was provided by NIH grants DP1 OD003961 (A.Y.T.), P41RR004050 (M.H.E.), P41GM103412 (M.H.E.), GM065937 (G.E.S.), GM072881 (G.E.S.), GM077465 (V.K.M), and GM42614 (T.L.P.). J.D.M. was supported by NSFGR and NDSEG fellowships.
AUTHORS’ CONTRIBUTIONS STATEMENTJ.D.M. and A.Y.T. designed the research, analyzed the data, and wrote the paper. All authors edited the paper. M.H.E. and G.E.S. oversaw EM experiments and analyzed the results. T.L.P. provided guidance on peroxidases and peroxidase assays. J.D.M. and T.J.D. performed EM sample preparation and imaging. Y.S. and V.K.M. performed calcium uptake assays and prepared MCU stable cells. J.D.M. performed all other experiments.
COMPETING INTERESTS STATEMENT
Massachusetts Institute of Technology is seeking to file a patent application covering part of the information contained in this article.