Ultrastructural Characterization of the Giant Volcano-like Virus Factory of Acanthamoeba polyphaga Mimivirus

Acanthamoeba polyphaga Mimivirus is a giant double-stranded DNA virus defining a new genus, the Mimiviridae, among the Nucleo-Cytoplasmic Large DNA Viruses (NCLDV). We used utrastructural studies to shed light on the different steps of the Mimivirus replication cycle: entry via phagocytosis, release of viral DNA into the cell cytoplasm through fusion of viral and vacuolar membranes, and finally viral morphogenesis in an extraordinary giant cytoplasmic virus factory (VF). Fluorescent staining of the AT-rich Mimivirus DNA showed that it enters the host nucleus prior to the generation of a cytoplasmic independent replication centre that forms the core of the VF. Assembly and filling of viral capsids were observed within the replication centre, before release into the cell cytoplasm where progeny virions accumulated. 3D reconstruction from fluorescent and differential contrast interference images revealed the VF emerging from the cell surface as a volcano-like structure. Its size dramatically grew during the 24 h infectious lytic cycle. Our results showed that Mimivirus replication is an extremely efficient process that results from a rapid takeover of cellular machinery, and takes place in a unique and autonomous giant assembly centre, leading to the release of a large number of complex virions through amoebal lysis.


INTRODUCTION
During the environmental study of an outbreak of pneumonia, a giant icosahedral DNA virus was discovered growing in amoebae. This virus was named Mimivirus (for mimicking microbe virus) [1]. With a diameter of about 650 nm, Mimivirus is the largest virus known to date. Morphologically, Mimivirus resembles Nucleo-Cytoplasmic Large DNA Viruses (NCLDV), such as the Iridoviruses, Asfarviruses and Phycodnaviruses [2]. Mimivirus comprises a central dense core that is surrounded by two lipid membrane layers inside a capsid protein shell covered by fibrils [3]. The sequence of its 1.2 Mb genome revealed 1262 putative open reading frames (Genbank accession number NC_006450; [4]). A phylogenetic study based on concatenated sequences of the eight class I genes [5] common to Mimivirus and to all NCLDVs revealed that Mimivirus belonged to this lineage, but stood apart from Phycodnaviridae, Iridoviridae, Asfarviridae and Poxviridae on the phylogenetic tree [4]. In A. polyphaga, Mimivirus replicative cycle was described as starting with a 4 h eclipse phase, followed by cytoplasmic accumulation of newly synthesized viruses, and ending with cell lysis and virus release 24 h post-infection (p.i.) [1,2]. Transmission electron microscopy (TEM) analysis of infected A. polyphaga suggested that viral replication, including DNA synthesis and particle assembly, might occur in and near the cell nucleus [1] and the existence of a virus factory was proposed [2]. As already described for a large variety of unrelated viruses, virus factories are perinuclear or cytoplasmic structures where virus replication and assembly take place. Their formation is the result of complex interactions between viral and cellular components and they induce profound alteration of the infected cell structure like recruitment of organelles and organisation of cellular compartments [6]. This paper describes for the first time the morphological characteristics of Mimivirus volcano-like giant virus factory, as determined by an extensive ultrastructural study and by tracking fluorescently-labelled viral DNA and viral proteins during the 24 h time course of infection. Our results reinforce the emerging picture of Mimivirus as a very complex and unique amoebal pathogen.

Ultrastructural aspects of the Mimivirus replication cycle
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 electronlucent 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.

Morphological description of the Mimivirus factory
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  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.

Specific aspects of the volcano-like giant Mimivirus factory
The height of the cells during the late phase of infection, and the limited depth of focus of the 636 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 mm 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 singlelayer 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 volcanolike 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 (mm 2 ) of the VF increased from 110 mm 2 at 4 h p.i. to 250 mm 2 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).

Fluorescence intensity quantification
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.

DISCUSSION
Mimivirus is a pathogen resistant to phagocytic destruction in amoebae, and as such should be considered as a possible new causative agent of human pneumonia. Indeed, links with human pneumonia were recently reported [9][10][11]. Among the different amoebae-resistant pathogens studied to date, Mimivirus appears to be the only one with such a rapid lytic effect on amoebae [12]. Until now, little was known about the different steps of the Mimivirus replication cycle. Our initial electron microscopy observations of in vitro Mimivirus-infected A. polyphaga showed the intra-cytoplasmic production and accumulation of newly synthesised viruses within a 24 h lytic cycle [1,2]. In these papers, we initially speculated that Mimivirus multiplied in the nuclei of infected cells. Indeed, we mistakenly identified the host nucleus as the VF because of its size and aspect. In the present study, in addition to ultrastructural characterization, the unusual size of Mimivirus allowed us to follow the different stages of its replication cycle using fluorescence and DIC microscopy. This enabled us to characterize the formation and growth of the giant Mimivirus VF, and to describe how progeny virions are synthesised, assembled and released from the replication centre to invade the cytoplasmic space. We propose the following replication cycle (Figure 8), composed of an early phase between 0-3 h p.i. (steps 1-4) and of a late phase thereafter (steps 5-8).
Electron microscopy images indicated that Mimivirus entry into the amoebae was most likely due to a phagocytic process, followed by fusion of phagocytic vacuoles. Mimivirus genetic material was delivered into the cytoplasm at this stage after fusion of the viral and vacuole membranes, most probably through the virus vertex. Whether the Mimivirus vertex plays also a role in the attachment to the cell surface, as described for Phycodnaviruses [13,14], remains to be established. Quantification from fluorescent nucleic acid labelling studies between 0 h and 3 h p.i. showed an increase in intensity in the amoebae nucleus, reflecting an increase in ATcontent that might be the result of Mimivirus DNA acquisition. We hypothesise that Mimivirus DNA first enters the amoebae nucleus, shortly after infection, probably for a first round of replication. At 3 h p.i., AT-rich Mimivirus DNA becomes localised to the cell cytoplasm in a structure distinct from the cell nucleus. This might be interpreted as the exit of the Mimivirus genetic material from the nucleus to form the replication centre of the VF in the cell cytoplasm. This structure has also been observed in ultrastructural studies, and it may be the major site of Mimivirus DNA production, independent of the cell nucleus machinery. At 8 h p.i., transmission electron microscopy, direct and indirect fluorescence labelling and quantification of Mimivirus DNA or protein allowed us to clearly distinguish the Mimivirus factory from the cell nucleus. This stage was characterized by an increase in viral DNA production within the cytoplasmic VF, while DNA staining in the host nucleus decreased. The Mimivirus factory size increased with time and with the production of progeny virions. This datasuit may be illustrated by a 3D model of the Mimivirus factory, composed of a replication centre made of unpackaged DNA, all around which viruses are formed in an assembly zone before being released in the cell cytoplasm ( Figure S2B).
The present results indicate that pre-formed capsids are filled with viral DNA, since all of the successive steps of capsid formation could be observed. As viral capsids were shown to assemble at the periphery of the replication centre, it might be envisaged that viral proteins are partially or fully concentrated or synthesised in the replications centre; this idea is supported by results obtained with a Mimivirus-specific mAb recognizing the late virion-associated R710 protein (Supporting Information Text S1, Figure S3). Furthermore, proteomic data analysis showed that no cellular host proteins seem to be incorporated within the virus particles [15] which might be indicative of an active mechanism of cell protein exclusion. It is not known whether gene extinction and cell machinery hijacking occurs in the Mimivirus factory to allow its replication and production, as has been described for other VFs [6].
A large variety of virus factories have been described for unrelated viruses [6]. It has previously been demonstrated that the replication site is predominantly cytoplasmic for Poxviridae [16], nuclear and cytoplasmic for Asfarviridae [17,18] and nuclear for Iridoviridae and Phycodnaviridae [13,14,19,20], whereas the assembly sites are all cytoplasmic. The main characteristics of these viruses, as well as their replication and assembly sites, are summarized in Supporting Information, Table S1. Here, we described a new VF,  which might be specific to the Mimiviridae, with a still undescribed replication centre which may insure a high degree of replication autonomy for this virus family regarding the host cell machinery. Similarities to the Asfarviridae could be observed, particularly in terms of early nuclear viral DNA replication, DNA insertion/ encapsidation into pre-formed capsids, and number of VF per infected cell [17,21]. However major differences are noticeable such as the weak detection of membranes within the VF [22], or the larger Mimivirus factory area compared to ASFV [17]. This is also true when comparison is made with other large DNA viruses factories which may occupy a large region of the infected cell [6]. Another difference is the absence of membrane surrounding the Mimivirus factory. Several questions are raised by our results. First, what is the source of nucleotides for building such a large DNA structure? Second, is there an exploitation of the aggresome pathway by cytoplasmic Mimivirus DNA to concentrate viral proteins at the assembly site, as previously reported for African swine fever virus [23]? Third, how are host proteins and organelles excluded from the VF region, and how is the cellular cytoskeleton reorganized?
In conclusion, the Mimivirus particle, composed of RNA transcripts combined with more than 100 viral proteins, appears to be particularly complex. Very specific mechanisms and complex interactions between viral and cellular factors must be involved to build this remarkably large and efficient VF, which can rapidly generate such a sophisticated microorganism.

Viral infection
A. polyphaga were seeded at 4610 5 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 1,0006 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.

Electron microscopy and immunofluorescence
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 4uC 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 ml of cell suspension at 4610 5 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 (49,69-diamidino-2-phenylindole) staining, cells were covered with 5 mM 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 ml of mAb P4C8G2, raised against  (3), where the first round of DNA replication might begin (4). At 3 h p.i. Mimivirus DNA came out the host nucleus to form the VF replication centre (5). At 5 h p.i. the VF size showed a 50% increase and viral proteins began to be detected. Proviral capsid assembly and viral capsids budding from the VF central core could be observed (6). Empty or DNA filled capsids accumulated nearby the central core, resulting in a growing VF with viral particles free in the cytoplasm (7). Complete viral capsids surrounded by fibrils might be released through cell lysis (8). doi:10.1371/journal.pone.0000328.g008 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 37uC for 30 min. After three washes in PBS, the slides were incubated for 30 min at 37uC with 100 ml 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 406, 636 or 1006 lenses. DIC images were acquired using the Axio Imager microscope. All images were acquired with a cooled (230uC) DS1-QM (Nikon) black and white camera driven by ''Lucia G'' software (Nikon & LIM Ltd. Prague, Czech Republic). DAPIfluorescence 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 mm. 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 2mean X )/(X max 2X 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.Rproject.org.)