Endoplasmic Reticulum-Golgi Intermediate Compartment Membranes and Vimentin Filaments Participate in Vaccinia Virus Assembly

Cells, viruses, and antisera.

HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% newborn bovine serum. VV wild-type strain Western Reserve (WR) was propagated and titrated in BSC40 cells (46). The conditional lethal mutants VVindA17L and VVindA14L, which inducibly express the VVp21 (encoded by the A17L gene) and p15 (encoded by the A14L gene) envelope proteins, respectively, had been previously generated and were grown in BSC40 cells in the presence of 2 mM IPTG, as described previously (45, 48). The rabbit polyclonal antisera against VV proteins p21, p15, and p39 (encoded by the A5L gene) have been previously described (46, 47). Antibodies specific for cytoskeletal proteins (vimentin, tubulin, actin, and cytokeratin) were purchased from Sigma: monoclonal antivimentin antibodies (clones V9 and VIM 13.2), a whole antivimentin antiserum raised in goat (V4630), a monoclonal anti-cytokeratin antibody (clone K8.13), an anti-keratin polyclonal antiserum raised in guinea pig (K4252), a rabbit anti-actin antiserum (A2668), and an anti-tubulin antiserum (T3526). The mouse monoclonal antibody G1193 against human ERGIC-53 protein was kindly provided by Hans P. Hauri (Biozentrum of the University of Basel). Secondary antibodies conjugated with Texas Red or fluorescein were purchased from Molecular Probes. Secondary antibodies-gold conjugates were provided by BioCell (Cardiff, United Kingdom).

Immunofluorescence microscopy.

HeLa cells grown on coverslips were infected at a multiplicity of infection (MOI) of 5 PFU/cell with VV strain WR. At 8 h p.i. cells were washed with phosphate-buffered saline (PBS) and fixed in methanol at −20°C for 5 min. After being washed with PBS, coverslips were blocked for 30 min with a solution of PBS containing 2% bovine serum albumin. Cells were then incubated for 1 h at 37°C with antibodies directed to the VV p21 protein together with antivimentin or anti-ERGIC-53 antibodies. The coverslips were extensively washed with PBS and incubated for 1 h at 37°C with secondary anti-mouse immunoglobulin G conjugated with fluorescein isothiocyanate (FITC) and anti-rabbit immunoglobulin G antibodies conjugated with Texas Red. The DNA staining reagent To-Pro (Molecular Probes) was included in this incubation. After several washings with PBS the coverslips were mounted with Fluoromount-G (Southern Biotecnology Associates, Inc.) on glass slides. Images were obtained using a Bio-Rad Radiance 2000 Confocal Laser microscope.

Fixation of cell cultures in situ for EM studies.

Monolayers of HeLa cells were infected at an MOI of 5 PFU/cell with the WR strain of VV in the absence or presence of 100 μg of rifampin/ml. HeLa cells were also infected at a similar MOI with VVindA17L or VVindA14L in the absence or presence of the inducer IPTG. For ultrastructural studies cells were fixed in situ with a mixture of 2% glutaraldehyde and 1% tannic acid in 0.4 M HEPES buffer (pH 7.5) for 1 h at room temperature. Fixed monolayers were removed from the culture dishes in the fixative and were transferred to Eppendorf tubes. After centrifugation and being washed with HEPES buffer, the cells were stored at 4°C until use.

For specific detection of proteins, monolayers of infected HeLa cells were submitted to a mild fixation with a solution of 4% paraformaldehyde containing 0.1% glutaraldehyde in PBS (pH 7.4). Fixed cells were removed from the dishes in the fixative, transferred to Eppendorf tubes, washed with PBS, and stored at 4°C until use.

Conventional processing for EM.

For ultrastructural studies, fixed cells were processed for embedding in the epoxy-resin EML-812 (Taab laboratories, Berkshire, United Kingdom) by methods previously described (42). Postfixation of cells was done with a mixture of 1% osmium tetroxide and 0.8% potassium ferricyanide in distilled water for 1 h at 4°C. After four washes with HEPES buffer, samples were treated with 2% uranyl acetate, washed again, and dehydrated in increasing concentrations of acetone (50, 70, 90, and 100%) for 10 min each at 4°C. Infiltration in resin was done at room temperature for 1 day. Polymerization of infiltrated samples was done at 60°C for 3 days. Ultrathin (20- to 30-nm-thick) sections of the samples were stained with saturated uranyl acetate and lead citrate by standard procedures.

For immunolabeling studies, cells submitted to mild fixation were processed for embedding in the acrylic-resin Lowicryl K4M (Taab Laboratories) as described previously (41). After dehydration at −20°C in increasing concentrations of ethanol, samples were infiltrated in the resin at −30°C for 1 day and polymerized with UV light, 2 days at −20°C and 2 more days at room temperature. Ultrathin sections were processed for immunogold detection of VV proteins or cytoskeletal components.

Immunogold labeling.

Immunogold localization of VV proteins and cytoskeletal components was done by placing the ultrathin sections on drops of different solutions. After a 30-min incubation with Tris buffer-gelatin (TBG) (30 mM Tris-HCl, pH 8.0, containing 150 mM NaCl, 0.1% bovine serum albumin, and 1% gelatin), sections were floated for 1 h on a drop of the specific primary antibody diluted in TBG. After jet washing with PBS, grids were floated on 4 drops of TBG and incubated 5 min on the last drop before a 45-min incubation with the secondary antibody conjugated with colloidal gold of 5 or 10 nm. Grids were then jet washed in PBS and distilled water before being stained with a solution of saturated uranyl acetate for 30 min (for Lowicryl sections) followed by 1 min with lead citrate (for EML-812 sections). For double-labeling experiments, representative signals corresponding to both primary antibodies were obtained after testing different combinations of labeling steps, as described previously (44).

Quick freezing and freeze-substitution.

Small pellets of chemically fixed cells were cryoprotected with glycerol, applied to small pieces of filter paper, blotted for 15 s, and quick frozen in liquid propane at an approximate speed of 10⁴°C/s. Frozen specimens were transferred to a Reichert-Jung AFS freeze-substitution unit (Leica, Vienna, Austria) and maintained for 24 h at −90°C in a mixture of pure acetone and 0.5% (wt/vol) osmium tetroxide. This incubation allows a complete substitution of the water of the sample by the solvent (42). Samples were then subjected to a controlled increase of temperature before being embedded in EML-812.

Freeze fracture and freeze-etching.

Frozen samples were processed in a BAF 060 freeze fracture unit (BAL-TEC; Liechtenstein). Regular freeze fracture was performed at −150°C following procedures previously described in detail (43). When freeze-etching was carried out after fracturing, the temperature of samples was switched from −150 to −100°C and was maintained at a pressure of 10⁻⁷ mbar for 5 min to sublimate the surface layer of ice. Metal replicas of the exposed surfaces were obtained by evaporating 2 nm of platinum with an electron gun at an angle of 45° and 20 nm of carbon with an electron gun at an angle of 90°. Replicas were floated in commercial bleach and incubated overnight for the digestion of cellular material. After being intensively washed in distilled water, the replicas were picked up in Formvar-coated EM grids and studied by EM.

EM: imaging and measurements.

Regular thin sections were collected on uncoated copper grids of 400 mesh, stained, and studied by EM. Serial sections were collected on nickel grids of 50 mesh or parallel bars coated by formvar films. Ultrathin sections of the samples were either stained by standard procedures, stained with saturated uranyl acetate in 70% ethanol (procedure that improves contrast), or processed for immunogold labeling. Metal replicas were picked up on copper 400-mesh grids. Collection of images and measurements were done with two different microscopes: a JEOL 1200-EX II electron microscope operating at 100 kV and a LEO TEM 812 (Zeiss) operating at 120 kV and equipped with an LaB6 filament, Omega in-column energy filter, and slow-scan CCD camera.

VV assembly areas contain numerous cellular elements.

The cytoplasmic regions of VV assembly have been traditionally defined as low-electron-dense areas of organelle exclusion (19, 32). This description is based on the apparent absence of structures around assembling viruses. In fact, this is one of the arguments used to support the hypothesis of a de novo synthesis of membranes originating the viral crescents: the viruses would not have any cellular elements around to be used for their assembly (11, 19). We have addressed this point in studying VV-infected cells (at different p.i. times) by high-preservation procedures, such as freeze-substitution. At low magnification, general structural features are similar to those of the well-known images of VV-infected cells processed by conventional methods (Fig. 1). At 10 h p.i. immature and mature viruses are equally represented (Fig. 1A). Mature virions in the process of taking the additional membrane characteristic of the IEVs are frequent (Fig. 1B). At 24 h p.i. the accumulation of large amounts of IMVs is a general feature (Fig. 1C and D). Apparently, IEVs are scarce at this p.i. time (for a more detailed quantification see below). Higher-magnification fields show clear differences between conventionally processed and freeze-substituted samples (Fig. 1E and F). When freeze-substitution is applied, numerous cellular structures (identified by their characteristic morphology) are distinguished around assembling viruses: membranes, microtubules, cytoskeletal IFs, and ribosomes are seen around the viroplasm foci of the factories and forming IVs (Fig. 1E). An equivalent field from a conventionally processed sample is shown in Fig. 1F: although some membranous elements can still be distinguished around IVs, the cytoplasm has lost part of its content and structural definition. Close proximity between forming IVs and cellular membranes or mitochondria is best appreciated in freeze fracture images (Fig. 1G). IVs are also seen close to rough endoplasmic reticulum (RER) cisternae with peripheral dense spots (Fig. 1H). Some unidentified structures, such as rigid tubules of around 50 to 60 nm in diameter, are frequently seen around assembling IVs (Fig. 1I). These structures are strongly reminiscent of a recently described ER subdomain (2). The presence of structures in the areas of assembly was seen at every time p.i. tested, although their amount clearly decreases at long times p.i.

As previously described (45, 77), VV morphogenesis is blocked at an early stage when expression of the VV p21 protein is prevented by infection with the conditional lethal mutant VVindA17L under nonpermissive conditions (absence of the inducer IPTG). The electron-dense masses representing truncated viroplasm foci, observed in the perinuclear areas under these conditions, are surrounded by numerous tubulovesicular membranes derived from the ERGIC (47). In the present study we have detected many other structural elements around these truncated viroplasm foci: ribosomes, IFs, aggregates similar to nuclear chromatin, and tubular structures of a variety of diameters (Fig. 1J to L).

The envelope of VV IV particles is a double bilayer.

The improved structural preservation obtained in frozen and freeze-substituted infected cells has provided several new data about the organization of the different VV assemblies. A double membrane bilayer is solved in viral crescents and IVs for the first time (Fig. 2A). Depending on the plane of the section, the trilamelar structure of the external membrane can be either clearly distinguished or partially masked by the presence of protrusions (arrows in Fig. 2A), termed spikes or spicules (12). Freeze-substitution shows that the internal membrane of the viral crescents is similar to a conventional cellular membrane, while the external one is deeply modified by the spikes. Examples of cellular membranes within the same cells are shown in Fig. 2B (mitochondrial double membrane) and C (intercellular junction). Thickness and morphology of individual membranes within these structures are similar to the corresponding viral crescents and IVs. These two bilayers are not resolved in viral crescents and IVs from conventionally processed samples, since the external one looks like a fuzzy layer, with spikes in some locations (Fig. 2D). When the section plane goes through the surface of the forming IV, the envelope shows a close-packing-like organization of the spikes (asterisk in Fig. 2E). These images show that when having an equatorial section plane of the envelope, the section can go along the line of spikes or through the space between two lines of spikes. In this case, the profile of the bilayer is seen. For well-preserved VV crescents, the thickness of the whole structure corresponds to almost three times the thickness of a standard plasma membrane: 5 nm for the internal bilayer, 5 nm for the external one, and 3 to 4 nm for the space between them (Fig. 2A).

Viral crescents seem to acquire their final described structure before attaching to the viroplasm foci of the viral factories (Fig. 2F and G). The crescents were confirmed as individual elongated cytoplasmic pieces with one or two visible bilayers at the crescent end, as seen in serial sections (data not shown). The arrow in Fig. 2F points to the external bilayer at one crescent end.

Finally, freeze fracture also reveals a double membrane in viral crescents and IVs: double lines are clearly distinguished in cross-fractured IV particles (Fig. 2H). The double mitochondrial membrane produces very similar images by freeze fracture (Fig. 2I).

Tubular membranes associate with assembling IVs.

Studying the areas where IVs are forming reveals that many of them have tubular or vesicular membranes associated (Fig. 3A, arrows). It is frequently observed that incomplete IVs with open pores have vesicles associated with them (Fig. 3B and C). Figures 3D to F are three serial sections of the same forming IV, showing tubules and vesicles associated to the structure in different planes. The vesicles are similar to the heads of the 30-nm-thick tubulovesicular elements seen in the areas of VV assembly (Fig. 3G, arrow). These elements are seen in groups around the viroplasm foci, and according to immunolabeling procedures they carry VV envelope proteins (Fig. 3H) and react with a monoclonal antibody specific for the well-characterized marker ERGIC-53 (Fig. 3I). In cells infected with VVindA17L under restrictive conditions (in the absence of VV p21 expression), these tubules attach to the surface of the truncated viroplasm foci, forming a palisade (Fig. 3J). Previous works have also reported the presence of ERGIC-like tubules in the areas of VV assembly. Mohandas and Dales (31) showed images with very prominent tubules, which were described as being continuous with the spherical virion envelopes. The abundance of ERGIC elements carrying VV envelope proteins around and in association with assembling IVs strongly supports the idea that viral crescents form from ERGIC tubules. We then studied the general organization of this compartment in infected cells by confocal microscopy (Fig. 3K to N). In VV-infected cells the areas of assembly are placed in the vicinity of the nucleus and can be clearly visualized by immunofluorescence with specific antibodies against VV proteins (Fig. 3K). Moreover, since these are areas of intensive DNA synthesis, they can also be defined by labeling them with a DNA marker (Fig. 3M). Similar to what is described for other cell types, the ERGIC-53-specific signal mainly concentrates in an area adjacent to the nucleus of HeLa cells, with some peripheral elements, both in uninfected (not shown) and VV-infected cells (Fig. 3L). The position of the viral factory is coincident with the areas of accumulation of ERGIC elements, which are precisely localized within the cytoplasmic mass of VV DNA and proteins (Fig. 3N). Although membranes carrying VV p21 or ERGIC-53 get together near the nucleus, it is clear that strict colocalization is not detected, since green and red colors remain separated in the merge image (Fig. 3N). This is confirmed by analysis of the different planes that compose the whole merge image (data not shown). We think that accumulation of viral proteins could exclude cellular proteins and consequently displace ERGIC-53 to particular subdomains within the membranous compartment.

Other new findings in VV-related structures.

Reexamination of other VV-related structures has shown additional new details. Analysis of cells infected with VVindA14L constitutes a good example. Conventional processing had shown that in the absence of the A14L gene product, p15, viral crescents are unable to interact with the surface of the viroplasm foci within the viral factories (48, 68). These curved membranes were undistinguishable from the normal crescents formed in WR VV-infected cells simultaneously processed by conventional methods. They were seen forming sphere IV-like particles full of holes or interrupted areas (Fig. 4A). However, freeze-substitution and freeze fracture shows that VVindA14L forms IV-like particles with apparently normal crescents together with 30- to 40-nm-diameter tubular structures (Fig. 4B and C). In this mutant the absence of the envelope-associated p15 protein would interfere with the transformation of tubular membranes into normal crescents. Our results strongly suggest that during conventional processing these tubular structures collapse and are no longer distinguishable. Serial sections of HeLa cells infected with the VVindA14L mutant in the absence of p15 protein show that the spherical IV-like particles are formed by individual tubular and crescent-like pieces of membranes in many different orientations (Fig. 4D to F). Some membranous pieces in IV-like particles are able to incorporate the characteristic spikes, which seem to be absent from the tubular structures (Fig. 4G).

Rifampin bodies (RBs) are structures formed in VV-infected cells in the presence of the drug rifampin. This drug sequesters the VV p65 membrane protein in cytoplasmic deposits close to the RBs (63). RBs are large structures similar to the viroplasm foci of the viral factories but with peripheral membranes that are not able to form viral crescents. Removal of the drug results in formation of normal crescents within minutes (35). Under these conditions crescents most probably originate from the previously recruited peripheral membranes, whose organization has not been characterized in detail. In our study no single-unit membranes are detected on the surface of RBs. Instead, different types of membranous structures are seen: dense membranes (around 18 nm thick), twisted double membranes with vesicular ends, and tubular elements (30 nm thick), some of them with vesicles at one end (Fig. 4H). Similar 30-nm-thick tubules have been detected before in association with RBs (31).

Freeze fracture gives us a three-dimensional view of the arrangement of membranous pieces in viral factories (Fig. 5). Tubular membranes (30 to 40 nm in diameter) can be occasionally seen interacting laterally on the periphery of viroplasm foci within the viral factories (Fig. 5A). When the surface layer of water is eliminated from the foci before making the metal replicas (freeze-etching), these show additional information: linear pieces, whose thickness (15 to 20 nm) is compatible with viral crescents as seen in metal replicas, are frequently seen interacting laterally on the surface of the viroplasm foci (Fig. 5B and C), and spikes organized in close-packing groups are also seen (double arrow in Fig. 5B). IVs fracture along their external surface, as confirmed in thin sections of fractured cells (data not shown). In this external surface, particles arranged in lines are distinguished (arrowheads in Fig. 5B and arrows in Fig. 5D) while on the internal surface no recognizable pattern is distinguished (asterisks in Fig. 5A and B).

Viral polymorphism: potential new VV maturation intermediates.

Figure 6 is a collection of images showing the different viral forms detected in freeze-substituted HeLa cells infected with wild-type VV. In Fig. 6A, an IV particle packaging DNA is shown. A large area of structured material (asterisk), which is reminiscent of the organization of nuclear chromatin, is seen nearby. The dense virus shown in Fig. 6B is identical to the structures previously proposed as intermediate maturation stages (64). Serial sections have confirmed that these are independent viral forms and not IMVs sectioned in particular orientations (data not shown). Anti-DNA antibodies confirmed that the dense fiber-like material occupying most of the interior of the particle contains DNA (data not shown). The characteristics of the particles in Fig. 6C and D suggest that they also could be maturation intermediates between IVs and IMVs. These assemblies are round- or ovoid-shaped particles with a variable internal structure, suggesting that the IMV core shell starts to form independently from the double membrane of the envelope (Fig. 6D). Mature VV forms, both intracellular and extracellular, show a complex internal organization (Fig. 6E to H). At least five independent layers (marked in Fig. 6E) are distinguished in IMVs sectioned along an equatorial plane. The interpretation of these profiles is rather complex, since maturation has produced a major reorganization of IV structure. The next viral form is shown in Fig. 6F, where an IEV exhibits a double membrane added around the basic structure of the IMV. Two different section planes of extracellular viral particles (Fig. 6G and H) show that they also have several internal layers and an external fuzzy coat (arrows). The different viral forms shown here are not equally represented at short and long p.i. times (Fig. 6I). At 10 h p.i. forming IVs, IMVs, and IEVs are all abundant. At 24 h p.i., however, although forming IVs are still detected, IMVs are very abundant and IEVs are scarce. Potential transitional maturation intermediates are, in both cases, a minor class and represent less than 5% of the total amount of viral structures. These quantitative data suggest that later in infection assembly and maturation processes are less dynamic and certain events could be taking place at different speeds.

Filaments around and inside the viroplasm foci react with antibodies specific for the cytoskeletal protein vimentin.

Filaments, whose thickness corresponds to the cytoskeletal IFs, are frequently detected around viroplasm foci of VV factories and forming IVs (Fig. 7A). A fibrous texture can sometimes be distinguished inside the foci, as previously described (44). IFs around viroplasm foci and IVs specifically reacted with three different (monoclonal and polyclonal) vimentin-specific antibodies (Fig. 7B). The three antibodies rendered similar results, with only minor differences in labeling intensity. Confocal microscopy shows that the vimentin-associated signal is mostly concentrated in the perinuclear area of infected cells, surrounds the macrostructures of the viral factories, and appears to be enclosing them (Fig. 7C to F). At the ultrastructural level, vimentin-labeled filaments are also seen inside forming and complete IVs, but IMVs were devoid of labeling (Fig. 7G and H). In cells infected with the recombinant virus VVindA17L, vimentin-associated signal is seen inside small foci formed early after induction of p21 expression by IPTG addition (data not shown) as well as in large foci assembled at long postinduction times (Fig. 7I). In these large foci labeling concentrates in particular areas, mainly where crescents (marked c in Fig. 7I) protrude. Antibodies specific for tubulin, actin, or cytokeratin provided disperse and weak labeling signals around the viroplasm foci or IVs (data not shown). Double-labeling experiments showed that vimentin and the VV core protein p39 (the product of the A4L gene) colocalize in the areas of the viroplasm foci where crescents are coming out (Fig. 7J and K).

These data support the hypothesis that ERGIC cellular membranes and vimentin IFs would act coordinately to build and organize the assembly foci of the viral factories in the first step of VV morphogenesis.