A. polyphaga were infected with a cell-free Mimivirus supernatant at a multiplicity of infection of 10, and processed for TEM at different times p.i.. At 30 min after infection, defined as the 0 h p.i. time point, Mimivirus appears to enter the amoebae by phagocytosis (Figure 1A) and was next observed within the phagocytic vacuoles of the amoebae (Figure 1B). Empty particles could be seen with an open vertex (Figure 1C). At 4 h p.i., several viruses could be found within the same vacuole either as fully closed or as empty open particles (Figure 1D and G). The most interesting phenomenon observed was the internal Mimivirus membrane extruding from the particle to fuse with the vacuole membrane (Figure 1E and F), and the apparent pouring out of electron dense material, most likely the viral DNA, into the cell cytoplasm (Figure 1G and H). Moreover, these events might occur through the vertex described by Xiao et al. [3], since this structure appeared to be open on empty particles (Figure 1C and G). It should be noted that the external structures, the outer layers and fibrils, remained intact on the empty particles at this stage. Next, condensed genetic Mimivirus material appears to enter the cell nucleus (Figure 1I and J). These structures were never seen in uninfected amoebae (data not shown). At later p.i. times, we previously described viral particles at the periphery of what we originally thought to be the cell nucleus [1], [2]. Further detailed examination of other series of ultrathin sections of A. polyphaga at 4 h p.i. revealed the appearance of an electron-dense structure, clearly distinct from the nucleus that might represent a cytoplasmic viral replication centre surrounded by mitochondria (Figure 1K and L). Contrary to the cell nucleus, this structure did not appear to be surrounded by a membrane. The size of this structure increased rapidly between 5 h and 8 h p.i. At 8 h p.i., newly synthesized viral particles were observed at the periphery of the putative replication centre (Figure 2A), surrounded by an electron-lucent zone, forming a virus factory (VF). At 12 h p.i., almost all the cytoplasmic space was occupied by the VF, and the cell nucleus could still be observed at the periphery (Figure 2B). These observations indicated that Mimivirus replication and assembly took place in a very specific cytoplasmic structure composed of a dense central core from which newly formed particles appeared.

The Mimivirus factory could be divided into three zones: the inner replication centre, the intermediate assembly zone and the peripheral zone where the newly formed particles acquired their fibrils. This later zone appeared electron-lucent, probably due to exclusion of cellular material and organelles by the expanding VF (Figure 2C, 3A). Closer examination of the VF replication centre suggested a possible sequence of events from assembly of the capsid shell, to the release of complete viral particles with a condensed core surrounded with fibrils. The replication centre of the VF showed a heterogeneous structure with dense inclusions (Figure 2C, 3A). The hexagonal shape of the capsid appeared as assembling progressed (Figure 2C–G), and empty capsids were then filled with electron-dense material before being released (Figure 2C). The vertex was clearly visible on the virus particle, opposite to the side linked to the replication centre (Figure 2C–G). Membranes underlining the capsid layer were observed in growing (Figure 2D–G) and released (Figure 2G) viral particles. These membranes did not encircle the replication centre and were always observed at its periphery. Their origin is still unknown.

Figure 3 illustrated how viral DNA might be encapsidated into nascent Mimivirus particles. The encapsidation process occurred at the replication centre periphery once capsid assembly is almost complete (Figure 3A). Viral DNA condensation seemed to begin within the replication centre (Figure 3D) before being inserted into viral capsids through an open portal located on the opposite side to the vertex (Figure 3B–D). Figure 3 C and D showed how viral DNA condensation progressed within the viral capsid to form the core centre of the viral particle.

Altogether these observations compelled us to modify our original interpretation and to further characterize the different stages of the Mimivirus assembly pathway.

The formation kinetics of the VF, and its viral DNA content in particular, were studied by direct fluorescent staining with the blue fluorescent stain DAPI. The choice of this molecule was based on the fact that DAPI preferentially stains dsDNA by association with AT clusters in the minor groove [7]. The Mimivirus genome has a high AT proportion (72%, [4]) compared to A. polyphaga (genomic AT content estimated to be 49%, determined using 96 shotgun sequences of amoebae genomic DNA; data not shown). Mimivirus-infected A. polyphaga were consequently stained at different time points p.i. Fluorescence and differential interference contrast (DIC) images of the same field are presented in Figure 4. Representative images are also shown in the Supporting Information (Text S1, Figure S1). In uninfected amoebae, cell nuclei showed a characteristic ring-like staining pattern with unlabeled nucleoli surrounded with labelled chromatin, similar to the nucleus morphology observed with TEM (Figure 4A and B). At 0 h p.i. Mimivirus nucleic acid staining appeared as bright single or clustered dots within the cell cytoplasm, contrasting with the cytoplasmic background DAPI labelling (Figure 4C and D). At 1 h p.i., the Mimivirus DAPI-stained DNA dots reached the cell nuclei, where size and staining intensity increased until 3 h p.i. At 4 h p.i. strongly stained clusters showing a heterogeneous structure appeared outside of the cell nuclei (Figure 4E and F). In most of the cells showing these structures, only one cluster could be seen per infected amoeba. Similar observations were made using standard DNA-staining histological dyes such as carbolic toluidine blue (data not shown). Such structures were not observed in uninfected amoebae. The size of these clusters peaked between 8–12 h p.i. and sustained their maximal size and staining intensity until the end of the replication cycle. However at 8 h p.i. the clusters exhibited a homogeneous structure (Figure 4G and H), whereas at 18 h p.i., they showed a heterogeneous less organised morphology. The time course of the development of these structures and their morphology clearly showed that they corresponded to the replication centre of the VFs observed by electron microscopy.

Because our experimental conditions used non-synchronized infected cells, we quantified the proportion of each of these morphologically distinct types of replication centre at different time points p.i. in order to determine whether there might be a progression from one type to the other across the time course of infection. Results are shown in Figure 5. Four different core centre morphologies were characterised and quantified (Figure 5A) : i) type I with a clustered morphology (see also Figure 4E), predominant in the first 4 hours of infection; ii) type II in which the core centre appeared as a completely homogeneous structure with a blurry aspect in microscopy images (see Figure 4G), most likely resulting from the fusion of the clusters seen earlier. This form was predominant from 6 to 12 h p.i.; iii) type III in which the whole Mimivirus DNA cluster was surrounded by more and more small bright dots, similar to virus particles, quickly filling the cytoplasmic volume. This form was detected from 7 h p.i. to the end of the infection; and iv) type IV in which the clusters had a heterogeneous morphology with a disorganized appearance with holes and fiber-like patterns (see also Figure 4I). This form was detected in the latest times of infection. These results allowed us to propose a progression of the different characteristic Mimivirus production stages from an early heterogeneous stage (I) corresponding to the appearance and formation of the VF core centre, followed by a homogeneous “mature” stage (II) corresponding to the growing core centre, then by a “productive” stage of the VF (III), and finally by a heterogeneous “degenerative” stage (IV) which most probably signed the exhaustion of the Mimivirus factory. Confocal data were used to build 3D volumic reconstruction of the three main types of Mimivirus factory replication centre during the time course of infection (see also Supporting Information Text S1, Figure S2A). Variation of the DAPI intensity staining is indicative of the variation of the DNA content in the core centre. A quantitative analysis is presented below.

The height of the cells during the late phase of infection, and the limited depth of focus of the 63× objective, allowed us to explore their 3D organisation based on DIC images. Details from one cell are shown in the upper part of Figure 5B. The maximum height of the cell was estimated to be about 10 µm using the difference between the uppermost and the lowest focused images that could be obtained (Figure 5B, sections a and d, respectively); this measurement was confirmed using confocal optical slides. The uppermost focused image always corresponded to the top of the VF as identified by DAPI staining (section a). Intermediate images (sections b and c) showed the newly synthesized virus particles spreading all around the VF and finally organizing into a single-layer crystal-like structure at the bottom of the cell (section d, see also Figure 4J). The Extended Depth of Focus (EDF) technique [8] allowed us to generate a topological view of all regions on the same focal plane. EDF allowed us to build a 3D image of the volcano-like structure found in Mimivirus-infected cells as shown in Figure 5B, lower part. The position of the VF replication centre in the EDF image was obtained by the overlay of the DAPI fluorescence image. The area (µm²) of the VF increased from 110 µm² at 4 h p.i. to 250 µm² at 12 h p.i. which represented about 42% of the cell surface (data not shown). These characteristics classified the Mimivirus factory among the largest described until now.

All the results obtained by transmission electron microscopy, widefield fluorescence, confocal and volume reconstruction analyses allowed us to propose a 3D model of the morphology of Mimivirus factory (Supporting Information, Figure S2 B).

During the infection cycle, nuclear DAPI staining peaked from 0 to 3 h p.i. and then decreased. After 8 h p.i., nuclei showed weaker labelling and a modified appearance: newly synthesized viral particles could be observed as single dots in the cytoplasm. The number of dots increased dramatically by the end of the replication cycle, at which time they completely filled the intracellular space. One interpretation of these results is that during the 0–3 h p.i. period, the increased nuclear fluorescence intensity was the consequence of the transient nuclear localization of Mimivirus DNA, which then moved into the cytoplasm to form the highly fluorescent VF. The brightness of the VF fluorescence indicated the accumulation of AT-rich DNA (Figures 4 and S1). Fluorescence intensity was analyzed and quantified as described in the Materials and Methods section, and results are shown in Figure 6. The fluorescence attributes (mean intensity and area) of nuclei and Mimivirus factory showed a concomitant and inverse evolution compared to each other, with the most remarkable point around 5 h p.i.: at this time, the cell area occupied by nuclei showed a 50% drop, whereas the cell area occupied by the Mimivirus factory showed a 50% increase (Figure 6 A and B). Statistical analysis revealed a significant increase of the mean nuclear fluorescence intensity between 0 h and 3 h p.i. (p<0.01) and a significant decrease between 0 h and 8 h p.i. (p<0.01; Figure 6 C). Conversely, there was a significant increase in the mean VF fluorescence intensity between 4 h and 8 h p.i. (p<0.01; Figure 6 D). Taken together, these observations favour a model in which the major site of Mimivirus DNA replication is the cytoplasmic VF, and further suggest that there is a relationship between the two different structures during the replication cycle. Quantification of total fluorescence intensity at different time points p.i. showed an 7-fold increase in total DNA in the cell between 0 h and 8 h p.i., which is exponential growth (e^0.2568x, Figure 7), equivalent to a doubling time of 2.7 h. In comparison, the total fluorescence intensity in uninfected amoebae varied from 1 at 0 h to 1.2 at 8 h (Figure 7, hatched bars). These results complemented the microscopy results, and allowed the first insights into Mimivirus replication cycle.

A. polyphaga were seeded at 4×10⁵ cells/ml in Page's amoebal saline (PAS) [24], infected with titrated Mimivirus at an amoeba cell:virus ratio of 1∶10 and centrifuged at 1,000× g for 30 min. Amoebae viability was estimated by counting the cells immediately after centrifugation and every two hours after that for the next 32 h.

Mimivirus-infected A. polyphaga were prepared for TEM as follows. Cells were washed three times in PBS, resuspended in 5% glutaraldehyde (Sigma) in PBS for 1 h at 4°C and then washed again three times in PBS. The cell pellet was fixed in 1% osmic acid, washed twice in PBS, dehydrated in 50, 70, 95 and 100% alcohol and embedded in Epon.

For fluorescence labelling, 100 µl of cell suspension at 4×10⁵ cells/ml were put into a Cytospin chamber, centrifuged for 10 min at 800 rpm in a Shandon Cytospin 4 (Thermo Electron Corporation) and then fixed for 10 min in methanol. For direct fluorescence with DAPI (4′,6′-diamidino-2-phenylindole) staining, cells were covered with 5 µM DAPI from a ready-to-use solution, “ProLong Gold Antifade Reagent” (Molecular Probes) and stained for 10 min in the dark prior to observation. For indirect immunofluorescence, 100 µl of mAb P4C8G2, raised against purified Mimivirus (data not shown), was diluted 1∶100 in PBS with 3% (w/v) non-fat dry milk and added to the slides. Slides were incubated in a moist chamber at 37°C for 30 min. After three washes in PBS, the slides were incubated for 30 min at 37°C with 100 µl of a FITC-conjugated goat anti-mouse Ig (Jackson ImmunoResearch) diluted 1∶100 in PBS containing 0.2% Evans blue. After three washes with PBS, the slides were mounted using a phosphate-buffered glycerol medium, pH 8, prior to observation.

Cells were observed using upright microscopes (Olympus BX 51and Zeiss Axio Imager) equipped with 40×, 63× or 100× lenses. DIC images were acquired using the Axio Imager microscope. All images were acquired with a cooled (−30°C) DS1-QM (Nikon) black and white camera driven by “Lucia G” software (Nikon & LIM Ltd. Prague, Czech Republic). DAPI-fluorescence images were taken using a DAPI filter (360/55 nm; 460/50 nm). FITC-mAb images were taken using an FITC filter (480/20; 535/40). Confocal images were acquired with an LSM 510 Zeiss microscope, with DAPI staining observed using a UV diode (405 nm), z step = 0.3 µm. 3D volumic reconstruction was achieved using OsiriX Medical Imaging Software [25]. The topology of infected cells was obtained using limited depth focus DIC images. The 3D reconstruction was obtained with Lucia software's EDF algorithm. The position of the VF replication centre in the EDF image was obtained by overlay of the DAPI fluorescence image.

Image analysis was performed using “Lucia G” and ImageJ software (Rasband, W.S., ImageJ, National Institutes of Health, Bethesda, Maryland, USA, 1997–2006, http://rsb.info.nih.gov/ij/). Images were acquired in 12 bit depth with the same exposure parameters. Three images per field were recorded, and a total of 1030 cells were analyzed using the following protocol. The absolute value of intensity was measured in regions of interest (ROIs): nucleus, cytoplasm, DNA clusters, monoclonal Ab staining, background. The following parameters were measured: area = sum of ROIs area; area fraction = area/area of cells in the field; intensity = [mean of (ROIs intensity/ROI area)]–background. In order to compare in the same graph the variations of the parameters having different units (Intensity, Area Fraction) we calculated for each parameter the “centered and normalized” value using the formula: X_t = (X_t−mean_X)/(X_max−X_min).

To quantify the evolution of the mean fluorescence intensity, measurements were compared: 0 h p.i. versus 3 h p.i. and 0 h p.i. versus 8 h p.i. For this purpose, four parameters were measured in DAPI-stained Mimivirus infected cells: nuclear area, mean nuclear fluorescence intensity, VF area and mean VF fluorescence intensity. For fluorescence quantification, acquisition time was 4 msec for the VFs and 100 msec for the nuclei. Statistical analyses were performed with the Wilcoxon test using R software (R Development Core Team (2006). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.)