Golgi twins in late mitosis revealed by genetically encoded tags for live cell imaging and correlated electron microscopy

Abstract

Combinations of molecular tags visible in light and electron microscopes become particularly advantageous in the analysis of dynamic cellular components like the Golgi apparatus. This organelle disassembles at the onset of mitosis and, after a sequence of poorly understood events, reassembles after cytokinesis. The precise location of Golgi membranes and resident proteins during mitosis remains unclear, partly due to limitations of molecular markers and the resolution of light microscopy. We generated a fusion consisting of the first 117 residues of α-mannosidase II tagged with a fluorescent protein and a tetracysteine motif. The mannosidase component guarantees docking into the Golgi membrane, with the tags exposed in the lumen. The fluorescent protein is optically visible without further treatment, whereas the tetracysteine tag can be reduced acutely with a membrane-permeant phosphine, labeled with ReAsH, monitored in the light microscope, and used to trigger the photoconversion of diaminobenzidine, allowing 4D optical recording on live cells and correlated ultrastructural analysis by electron microscopy. These methods reveal that Golgi reassembly is preceded by the formation of four colinear clusters at telophase, two per daughter cell. Within each daughter, the smaller cluster near the midbody gradually migrates to rejoin the major cluster on the far side of the nucleus and asymmetrically reconstitutes a single Golgi apparatus, first in one daughter cell and then in the other. Our studies provide previously undescribed insights into Golgi disassociation and reassembly during mitosis and offer a powerful approach to follow recombinant protein distribution in 4D imaging and correlated high-resolution analysis.

The Golgi apparatus allows functional diversification of mature proteins by adding and refining carbohydrate chains. In the interphase mammalian cell, the Golgi apparatus is shaped like a ribbon, with stacks of flattened and fenestrated cisternae joined together by a tubular network and anchored in the centrosomal region of the cytoplasm. Before cells enter mitosis, the Golgi apparatus starts a duplication process that may last throughout G₁, S, and G₂, ensuring proper inheritance of the organelle (1, 2). During cell division, it undergoes vesiculation and fragmentation, and its components are found scattered throughout the cytoplasm at metaphase/anaphase in the form of mitotic Golgi clusters and thousands of tiny (≈50-nm) vesicles sized below the resolution of light microscopy (LM) (3–9), often referred to as the “Golgi haze.” The nature of the Golgi haze as observed in LM is controversial, with evidence for (6, 10) and against (11–13) coalescence of these vesicular components with the endoplasmic reticulum (ER). From this haze, the Golgi apparatus has been shown to reconstitute, through a series of poorly understood events, reforming the tubules and stacks that are characteristic of its interphase form. Fusions of Golgi-resident enzymes with fluorescent proteins are commonly used to delineate the Golgi apparatus and follow its dynamics during cell division. However, these fusions are visible only in the LM and usually rely on immunoelectron microscopic approaches for correlated analysis of molecular constituents, with the consequent degradation in quality of the ultrastructural preservation that is associated with antibody labeling (reviewed in ref. 14). To overcome these limitations, we added a small tetracysteine-containing peptide to the carboxyl terminus of green fluorescent protein (GFP) or cyan fluorescent protein (CFP) and fused the resulting fluorescent protein-tetracysteine tag to the first 117 residues of the Golgi resident enzyme α-mannosidase II (MannII). The tetracysteine peptide used in the combinatorial tag, FLNCCPGCCMEP (4C), has an increased affinity and improved ReAsH quantum efficiency compared with earlier tetracysteine peptides (15). The resulting fusion protein, MannII-GFP-4C, was expressed in cultured mammalian cells and used to chronicle the changes occurring to the Golgi apparatus during mitosis. The fluorescent protein allowed for direct live imaging without further labeling, whereas the biarsenical compound ReAsH (15, 16) bound to the tetracysteine component of the tag after acute application of membrane-permeant reducing agents and was used for live cell imaging or for FRET-based photoconversion of diaminobenzidine (DAB) for high-resolution analysis by electron microscopy (EM) (Fig. 1A).

Fig. 1.

Construction and validation of a Golgi marker for EM and LM imaging. (A) Generation of a combinatorial tag by using GFP and a tetracysteine motif and application to correlated LM/EM. FRET between GFP and ReAsH was used to trigger photoconversion of DAB. Because FRET exclusively occurs between closely apposed fluorochromes, only the biarsenical bound to 4C will accept the energy from GFP and produce singlet oxygen, thereby increasing the specificity of photoconversion. (Scale bar: ≈1 nm.) Note that the structure of the tetracysteine is a model, not an experimental determination. (B) The Emerald GFP-4C (eGFP-4C) module was fused to the 117 N-terminal residues of MannII containing cytoplasmic, transmembrane, and part of the intraluminal domains. (C and D) MannII-GFP-4C localizes specifically to the Golgi apparatus. HeLa cells stably expressing MannII-GFP-4C (green) were labeled with an antibody to Giantin (C, red), an integral component of the Golgi membrane, or GM-130 (D, red), a protein of the Golgi matrix. (Scale bars: 10 μm.) (E) The intraluminal tetracysteine tag is labeled only in reducing conditions. HeLa cells transiently expressing cytoplasmic tetracysteine-tagged actin and Golgi resident MannII-CFP-4C were labeled with FlAsH-EDT₂ in the absence of TBP. FlAsH bound to the reduced thiols in tetracysteine-tagged actin but not to the Golgi-luminal oxidized thiols in ManII-CFP-4C. Next, cells were treated simultaneously with ReAsH-EDT₂ and the membrane-permeant reducing agent, TBP. TBP transiently reduces the intraluminal tetracysteine and, thus, allows ReAsH binding. (Scale bar: 20 μm.)

Results

Labeling Tetracysteine Tags with ReAsH-EDT₂ in Oxidizing Environments.

MannII-GFP-4C (Fig. 1 A and B) was stably expressed in HeLa cells, where it localized in the Golgi apparatus and codistributed with the Golgi proteins Giantin (Fig. 1C), GM-130 (Fig. 1D), and α-mannosidase II (data not shown). The tetracysteine residues in MannII-GFP-4C are exposed to the oxidizing environment of the Golgi lumen and would normally be unavailable to FlAsH or ReAsH, because binding of biarsenicals to tetracysteine tags requires the cysteine residues to be completely reduced (17, 18). Thus, when HeLa cells expressing tetracysteine-tagged actin (15), used as a cytoplasmic protein control, and MannII-CFP-4C were incubated with saturating concentrations of FlAsH-EDT₂ in absence of tributylphosphine (TBP), only actin was labeled (Fig. 1E Left). Coadministration of membrane-permeant reducing agents TBP or triethylphosphine (TEP) allowed ReAsH labeling of MannII-CFP-4C in the Golgi lumen of living cell (Fig. 1E Middle and Right). CFP rather than GFP was used here because of spectral overlap between FlAsH and GFP.

MannII-GFP-4C Distribution During Mitosis: Correlated Microscopy and FRET-Based Photoconversion.

To examine the Golgi disassembly and partitioning during mitosis in correlated LM and EM studies, we labeled cells stably expressing MannII-GFP-4C with ReAsH-EDT₂ in the presence of TEP, imaged multiple fields by using a high-speed, two photon laser-scanning microscope equipped with a motorized stage, and then photoconverted the imaged areas (Fig. 2). We examined the progression through cell division by using the maximum intensity projection of MannII-GFP-4C (GFP fluorescence displayed in negative contrast) (Fig. 2A; see Movie 1, which is published as supporting information on the PNAS web site). This experimental setup allowed us to record multiple fields in parallel over long periods of time (i.e., days), capturing cells at different stages of division. We preferred this approach to chemical treatments aimed at synchronizing the cell population at a specific mitotic stage, which might lead to alterations of the Golgi apparatus, cellular hypertrophy, and artifacts in the distribution and dynamics of Golgi proteins. Upon completion of live imaging, sometimes when a critical stage of mitosis was seen, cells were fixed with 2% glutaraldehyde and processed for photoconversion of DAB (Fig. 2 B–E). Photoexcitation of ReAsH generates singlet oxygen, which locally polymerizes DAB into a fine precipitate that is subsequently rendered electron-dense by treatment with osmium tetroxide (19). Here, we indirectly excited ReAsH through FRET from GFP (Fig. 1A). The electron-dense product was found localized to the medial and trans-Golgi stacks, with traces in the cis-Golgi stacks of interphase cells. This is a distribution similar to that reported by others for the native form of α-mannosidase II in HeLa (20) and other (21) cells. The nucleus, nuclear envelope, ER, and mitochondria were not labeled (Fig. 2F). MannII-GFP-4C was found at prophase (Fig. 2 G and H) and prometaphase (Fig. 2 I and J) in vesicles and clusters similar in size and morphology to those described (3, 4, 6, 7, 9), which were generally separated from the membranes of the rough ER. At late metaphase, the majority of the fusion protein was found in small vesicles ranging from 50 to 70 nm (as described in refs. 4 and 7), and distributed at the periphery or close to the metaphase plate (Fig. 2 K–M; see Movie 2, which is published as supporting information on the PNAS web site). In our initial time-lapse studies, total fluorescence was seemingly less during mitosis. Movement out of the defined Z sections due to the cell rounding was found to cause this effect. The total GFP fluorescence is nonetheless conserved even though the haze looks much dimmer than the organized Golgi stacks (see below). We found the vesicles along the metaphase plate to be clearly separated from ER tubules (Fig. 2K), whereas those at the cell poles were sometimes close to ER-like membranes (Fig. 2 L and M).

Fig. 2.

Golgi vesicles throughout mitosis. (A) MannII-GFP-4C dynamics in HeLa cells. A maximum intensity projection of the GFP fluorescence is shown for selected time frames (see Movie 1 for the complete sequence). Contrast has been reversed to aid comparison with EM, so that the brightest fluorescence is printed most darkly. The white asterisk marks the cell portrayed in B–F. Time is in hrs:min. (Scale bar: 20 μm.) (B–D) The same area as in A before (B and C) and after (D) photoconversion. GFP (B, green) and the ReAsH (C, red) signals after fixation and before photoconversion are shown, followed by the transmission LM image (D) of the same area after FRET-mediated photoconversion, osmication, dehydration, and epoxy embedding. (E) A low-magnification electron micrograph of the same area shows increased electron-density specifically at the Golgi apparatus but not at other cellular regions. (Scale bars: 10 μm.) (F) Higher magnification shows electron densities in the Golgi stacks, but not in other organelles such as nucleus (Nuc), mitochondria (black arrow), and ER (white arrow). (Scale bar: 2 μm.) (G–J) MannII-GFP-4C distribution in prophase (G and H) and prometaphase (I and J) cells. The reaction product was present in small mitotic Golgi clusters as well as in larger fragmentation units. (Scale bar: 1 μm.) (K–M) MannII-GFP-4C distribution in late metaphase cells. Chromosomes and ER membranes are marked with black arrowheads and white arrows, respectively. Movie 2 shows the dynamic behavior of the Golgi apparatus in this cell before fixation. (Scale bar: K, 5 μm; L and M, 1 μm.)

Appearance of Four Golgi Clusters at Telophase.

According to descriptions in refs. 22 and 23, the Golgi haze coalesces into one ribbon-like structure after the daughter cells have undergone abscission. In time-lapse recordings of HeLa cells expressing mCherry-α-tubulin and MannII-GFP-4C (Fig. 3A; see Movie 3, which is published as supporting information on the PNAS web site) or high-resolution imaging of HeLa cells stably expressing MannII-GFP-4C (Fig. 3 B and C; see Movies 4 and 5, which are published as supporting information on the PNAS web site), we saw that twin clusters of vesicles and tubular structures containing MannII-GFP-4C consistently formed at the edge of the midbody and distal to the midbody, one pair in each daughter cell. For simplicity, we will refer to the large cluster, distal to the midbody, as the “major twin,” and the small cluster, which first appears adjacent to the midbody, as the “minor twin.” Although the minor Golgi twins appeared at the same time, their migration to the centrosphere and coalescence with the major twins was not synchronized in the separating cell pairs (Fig. 3C). One cell kept the twins separated from each other throughout most of telophase, whereas the other cell gradually fused them together and formed a single Golgi apparatus in a pericentrosomal location. A single Z section observed every 15 seconds at higher magnification showed the appearance and coalescence of the Golgi twins in more dynamic detail (see Movie 6, which is published as supporting information on the PNAS web site). The reassembly kinetics of the Golgi apparatus was scored in 15 daughter pairs (Fig. 3D). The coalescence took from 0.5 h to almost 2 h for those containing the faster twins, and from 1 to >3 h for those containing the slower twins. Independently from the variability in total time, the coalescence differs by 2-fold between daughter cells.

Fig. 3.

Four Golgi twins appear in telophase. (A) Golgi twins assemble at the midbody. HeLa cells expressing mCherry-α-tubulin (red) and MannII-GFP-4C (green) were imaged every 10 min. Major (“T”) and minor (“t”) Golgi twins appeared at the midbody (marked by a white asterisk) at late telophase cytokinesis (Movie 3). (Scale bar: 10 μm.) (B and C) Seventeen (B) and 7 h (C) zoomed kymograph (where the sequential regions of interests from the time-lapse recording are fused together to show dynamic changes in a 2D representation) of the sum intensity projection (inverted contrast) of the GFP fluorescence in a dividing cell. Cells were imaged every 8 min for 17 h every 0.25 μm (the total thickness is 34 μm; Movie 5; a higher spatiotemporal resolution imaging in one Z plane can be found in Movie 6). (Scale bar: 20 μm.) (D) Kinetics of Golgi twin coalescence. t = 0 was defined as the time when four Golgi clusters were clearly visible, and reconstitution was considered complete when 95% of the fluorescence in each daughter cell (as defined by 3D isosurface renderings) was contiguous. In 15 of 15 cell pairs observed, the rate of reconstitution was twice as fast in one daughter cell as compared with the other (mean fast = 66 ± 22 min; mean slow = 135 ± 34 min; P < 0.001, t test). The solid black line refers to values collected for the cell displayed in the kymograph in B and C.

EM Analysis of Golgi Twins.

Transmitted light, GFP, and ReAsH images were recorded in time-lapse from live cells (Fig. 4A Top, Middle, and Bottom, respectively; see Movie 7, which is published as supporting information on the PNAS web site). When the Golgi twins were clearly distinguishable by fluorescence, cells were fixed, photoconverted, and processed for EM. Thin sections of the epoxy-embedded samples were used to examine the contact region between two daughter cells (Fig. 4 B–D). MannII-GFP-4C localized in vesicles and vesicle-tubular structures closely apposed to the microtubules of the midbody (Fig. 4C) and to the microtubules stretching between the midbody and the centrosphere (Fig. 4D; see, which are published as supporting information on the PNAS web site). Similar observations were made at lower resolution in HeLa cells expressing MannII-GFP-4C fixed and stained for α-tubulin (Fig. 4 E and F).

Fig. 4.

Correlated microscopy of microtubule-associated Golgi twins. (A) Time-lapse imaging (20-min intervals, 3 h and 40 min total) of Mann II-GFP-4C during cell division (Movie 7). Transmitted light (Top), GFP (Middle), and ReAsH (Bottom) fluorescence are shown. The contrast of the fluorescence images has been inverted to maximize visibility of the vesicle clusters. (Scale bar: 10 μm.) (B) Transmitted light image of the daughter cells shown in A after FRET-based photoconversion of DAB. (Scale bar: 10 μm.) (C and D) EM analysis of thin sections of the same cell shown in B. (C and D Insets) The same area at lower magnification. (Scale bar: 10 μm.) (C) Photoconverted vesicles of the Golgi twin at the midbody are closely associated with microtubules. Elongated vesicle/stack structure and larger structures appear alongside of the microtubules (arrows). (D) A more apical section of the same cell is shown. Here, a nuclear furrow is enriched with an elongated trail of round and elongated photoconverted vesicles (arrows) alongside of microtubules. Note that by using thin section EM, the four colinear clusters are not always found in the same section. Nuc, nucleus. (Scale bars: C and D, 0.5 μm.) (E and F) Proliferating HeLa cells expressing MannII-GFP-4C (green) were fixed and counterstained for α-tubulin (red) and DNA (blue). Note the close association of the Golgi fragments with microtubules (arrows) through the nuclear furrow (two different single Z planes). (Scale bar: 5 μm.)

Discussion

The ability to precisely localize target proteins with high spatial and temporal resolution is fundamental in understanding dynamic events in cell biology. The combination of multiple molecular tags is advantageous for this purpose, as we showed in our study on Golgi apparatus behavior during mitosis. Golgi-resident proteins tagged with GFP have been extensively used to study the disassembly and reassembly of the Golgi apparatus by live imaging. Although this approach is relatively easy to set up and can be extended to functional studies, it lacks the resolution required to resolve the mitotic haze, when Golgi residents are in the thousands of 50- to 70-nm vesicles scattered throughout the cytoplasm of the dividing cell. Because GFP is not directly visible in the electron microscope, traditional immunocytochemistry approaches for correlative light/electron microscopy studies must be used. By appending a tetracysteine motif to the carboxyl terminus of GFP and using this combinatorial tag in a fusion with a Golgi resident protein, we generated a multiresolution marker that allowed us to merge fluorescence live-imaging studies and ultrastructural analysis. The sequence FLNCCPGCCMEP belongs to the latest series of enhanced tetracysteine motifs, which have higher fluorescence quantum yields (ReAsH quantum yield on GFP = 0.42–0.47), improved dithiol resistance, and higher contrast. We further increased the absolute contrast of ReAsH fluorescence and photoconversion by fusing the tetracysteine motif to GFP and exciting the biarsenical by FRET from the donor fluorescent protein (15). Because the distance between acceptor and donor must be within ≈8 nm, only the biarsenical bound to GFP-4C is excited by FRET, whereas the nonspecific biarsenicals are disfavored. Because cells were fixed after labeling with ReAsH and no antibodies were needed to paint the studied structures, we were able to use high concentrations (2%) of a strong cross-linking fixative (glutaraldehyde), which allowed optimal preservation of the ultrastructure.

Recovery of Binding in Oxidizing Compartments by Brief Reduction.

Biarsenical binding to an oxidized tetracysteine motif is recovered by acute application of a reducing agent (16). We showed that recovery is possible even if the tetracysteine motif is located intraluminally in an oxidizing cellular compartment, such as the Golgi apparatus or the ER. The only limiting factor in this case is the use of a membrane-permeant reducing agent, such as TEP or TBP. The requirement for reduction of the thiol groups in the cysteines actually can be used to our advantage to sequentially label two tetracysteine-tagged proteins localized in cellular compartments with different redox values. As shown in Fig. 1E, tetracysteine-tagged actin and MannII-GFP-4C are labeled with FlAsH and ReAsH, respectively, without cross interference.

Noncoalescence of Golgi Apparatus and ER During Mitosis.

These experiments were undertaken to obtain high-quality structural details of the Golgi fragmentation and reconstitution during mammalian cell division. The high-resolution analysis after photoconversion revealed that MannII-GFP-4C typically did not localize to the ER or other organelles at metaphase and anaphase, with MannII-bearing vesicular Golgi fragments appearing to maintain autonomy throughout mitosis. However, we found some photoconverted material in close proximity to ER membranes, indicating the possibility of an interaction between the two membranous systems. Additional studies using the labeling methods described here, with greater temporal resolution and employing multiple coexpressed ER-specific and Golgi-specific markers, could help resolve open questions regarding the dynamics of the relationship between ER and Golgi organelle systems during cytokinesis.

Golgi Twins.

We followed the formation of four clusters of Golgi membranes at telophase and their asynchronous coalescence during the later stages of cytokinesis before abscission of the daughter cells. Although we defined the Golgi twins on the basis of MannII-GFP-4C localization in HeLa cells, transfection of the same fusion in fibroblasts (Rat-1) or epithelial cells (NRK, MDCK), and analysis of endogenous Golgi proteins in telophase cells, revealed similar results (data not shown). The four Golgi twins appeared to contain an array of proteins similar to that present in the interphase Golgi apparatus, including Giantin and GM130 (see the nontransfected cell in Fig. 1D). The static labeling of similar structures also has been shown in other studies, but was either unnoticed or not fully appreciated due to incomplete capture of the entire 3D volume of the cells or limited temporal resolution. Thyberg and coworkers (24–27) reported that the Golgi apparatus reassembles near the midbody. They proposed a model in which Golgi stacks reorganize during telophase at the “proximal” side of the nucleus, near the midbody. During cytokinesis, the Golgi stacks were found to migrate toward the “distal” side of the nucleus. However, in more than one-third of the cases, the authors found more than two clusters in the daughter cells. More recently, GM130 was shown to be localized in four clusters during telophase (28).

Asynchronous Coalescence of Golgi Twins.

The two clusters of Golgi membranes clearly visible in each daughter at telophase fuse into one formation in an asynchronous fashion before abscission. Although the reason for this behavior is not clear, it is likely to be related to other asynchronous events taking place right before completion of cytokinesis. Amongst those, events affecting centrosomes and cytoskeleton are likely candidates. It has been shown that the centrosome splits into mother and daughter centrioles after formation of the cleavage furrow, and that the mother centriole often migrates close to the midbody in ≈70% of the observed HeLa cells (29). In this system studied by Piel et al. (29), abscission appeared to occur preferentially on the side where the mother centriole came closest to the midbody. This raises the interesting possibility that the movements of the mother centriole (together with the microtubular network to which it is connected) are correlated with the asynchronous fusion of the Golgi twins.

Conclusions and Future Perspectives.

Fluorescent proteins genetically fused to the protein of interest have become increasingly popular tags in the past decade, yet their use has been limited to fluorescence microscopy. Here, we demonstrate the application of a more versatile tagging system, extend its applicability to intracellular sites inside the lumen of oxidizing environments, and demonstrate that such capabilities have value for basic studies of the complex dynamics of intracellular organelle systems, such as the Golgi apparatus. The fusion of an intrinsically fluorescent protein and a tetracysteine motif is a relatively simple enhancement that will allow researchers to take advantage of a widely used tag with high fluorescent quantum yield and excellent contrast (GFP and derivatives) and the adaptability of the small tetracysteine tag system and the biarsenicals ligands that bind selectively to these domains.

Materials and Methods

In Vivo Labeling in Reducing Environments.

Cells stably expressing MannII-GFP-4C were labeled for 1 h at 37°C with 1.25 μM ReAsH-EDT₂/10 μM EDT/1 mM TEP in Hank's Balanced Salt Solution with D⁺glucose (1 g/liter). One molar solutions of TBP and TEP were prepared fresh in ethanol and diluted 1:1,000 in HBSS before the addition of ReAsH-EDT₂ and EDT. Free and nonspecifically bound ReAsH was removed by washing with EDT (600 μM).

Photoconversion of DAB and Electron Microscopy.

ReAsH-labeled cells were fixed in 2% glutaraldehyde in sodium cacodylate buffer (0.1 M, pH 7.4), and the microscope stage was switched from 37°C to 4°C. After 20 min, the cells were rinsed in cacodylate buffer and treated for 5 min with blocking buffer: 10 mM KCN/10 mM aminotriazole/0.01% hydrogen peroxide/50 mM glycine in 0.1 M cacodylate buffer (19). This buffer was replaced with blocking buffer, including 1 mg/ml DAB, and photoconversion was performed by using intense illumination (75 W xenon lamp without neutral density filters) focused through the microscope objective. Washing and further preparation for EM were performed as described in ref. 19.

Supporting Information.

Additional details can be found in Supporting Materials and Methods, which is published as supporting information on the PNAS web site.

Abbreviations

CFP

cyan fluorescent protein

DAB

diaminobenzidine

EM

electron microscopy

ER

endoplasmic reticulum

LM

light microscopy

MannII

α-mannosidase II

TBP

tributylphosphine

TEP

triethylphosphine.