More than a century ago, the term ‘virus’ was introduced to describe infectious agents that are invisible by light microscopy and capable of passing through sterilizing filters. In addition to their extremely small size, most viruses have minimal genomes and gene contents, and rely almost entirely on host cell-encoded functions to multiply. Unexpectedly, four different families of eukaryotic ‘giant viruses’ have been discovered over the past 10 years with genome sizes, gene contents and particle dimensions overlapping with that of cellular microbes. Their ongoing analyses are challenging accepted ideas about the diversity, evolution and origin of DNA viruses.
HISTORICAL PERSPECTIVE
Since the historical experiment by Ivanovsky (1892) revealing that the causative agent of tobacco mosaic disease could pass through the Chamberland sterilizing filter, the notion that viruses are by definition much smaller than bacteria has remained central in virology. From 1898, when Beijerinck introduced the word ‘virus’ (Beijerinck 1898) to 1939 when they were first visualized using the newly invented electron microscope (Kausche, Pfankuch and Ruska 1939), their nature remained mysterious and their relationship with the living/cellular world was hotly debated. Lwoff (1957) proposed the first set of formal criteria to discriminate between viruses and cells. These criteria that were further refined in a subsequent article (Lwoff and Tournier 1966) can be summarized as follows: viruses possess a DNA (or RNA) genome but, in contrast to cells (including obligatory intracellular parasites), they do not divide, do not encode a protein translation apparatus and do not synthetize the ATP they consume for their replication. Thus, despite its historical importance, small size was not part of Lwoff's criteria, even though it was still mentioned as an important characteristic ‘correlating with some essential properties which are responsible for some fundamental differences’ (Lwoff 1957). Yet, the unique property of viruses to pass through sterilizing filters with a porosity of 0.2–0.3 μm remained central to most isolation procedures until today, especially in most studies seeking to isolate viruses from the environment (Breitbart et al.2002; Mizuno et al.2013). Thus, during the 20th century, a microorganism that was retained by a sterilizing filter or that could be seen under a light microscope could not possibly be a ‘virus’. This epistemological barrier delayed the discovery of the viral nature of Mimivirus until 2003, 10 years after it was first isolated (La Scola et al.2003). ‘Giant virus’, defined as those exhibiting viral particles easily visible by light microscopy (i.e. smallest dimension >0.3 μm), now encompass four distinct families: the Mimiviridae (with a dozen different isolates/strains) (Raoult et al.2004; La Scola et al.2010; Arslan et al.2011; Colson et al.2011b; Legendre et al.2012; Yoosuf et al.2012, 2014a,b; Boughalmi et al.2013a,b; Saadi et al.2013a,b), the Pandoraviridae (three known members) (Philippe et al.2013; Antwerpen et al.2015), Pithovirus (possibly two isolates) (Michel et al.2003b; Legendre et al.2014; Claverie and Abergel 2015) and Mollivirus (Legendre et al.2015). All of them infect species of the amoebozoan genus Acanthamoeba, one of the most common protozoa in soil (Geisen et al.2014), ubiquitous in natural aquatic environments and sediments (Sawyer, Visvesvara and Harke 1977), as well as in man-made water systems (domestic networks, cooling towers or wastewater treatment plants) (Thomas et al.2008). Although Mimivirus was isolated in the context of a pneumonia outbreak (Raoult, La Scola and Birtles 2007), their pathogenicity to humans remains uncertain (Vanspauwen et al.2013).
Originally, nucleocytoplasmic large DNA viruses (NCLDVs) were regrouping the four known families of large DNA viruses infecting eukaryotes, the Poxviruses, Asfarviruses, Iridoviruses (including Ascoviruses) and Phycodnaviruses, postulated to have a common origin (Iyer, Aravind and Koonin 2001). They have dsDNA genomes with sizes ranging from 100 to ∼400 kb and can either be exclusively cytoplasmic or recruit functions from the cell nucleus for the earliest phases of their infectious cycles. The list of NCLDVs expanded with the discovery of Mimivirus (Raoult et al.2004; Iyer et al.2006), then of close relatives infecting Acanthamoeba (Raoult et al.2004; La Scola et al.2010; Arslan et al.2011; Colson et al.2011b; Legendre et al.2012; Yoosuf et al.2012, 2014a,b; Boughalmi et al.2013a,b; Saadi et al.2013a,b) and related large DNA viruses infecting other unicellular protists (Fischer et al.2010; Yau et al.2011; Santini et al.2013). Together they form the newly defined family ‘Mimiviridae’. In parallel, the ‘Marseilleviridae’, a distinct family (although related to the Iridoviridae) of Acanthamoeba-infecting dsDNA viruses is rapidly growing (Boyer et al.2009; Thomas et al.2011; Boughalmi et al.2013b; Colson et al.2013b; Doutre et al.2014). Comparison of the Mimiviridae and Marseilleviridae with the previous families constituting the NCLDVs places them within this group, that was proposed to form the new order ‘Megavirales’ (Colson et al.2013a) (not yet accepted by ICTV). A set of 47 (almost) conserved genes was proposed to originate from their postulated common ancestor (Yutin et al.2009). The proposed inclusion of the Pandoraviruses in the NCLDV on the basis of even less shared conserved genes (Yutin and Koonin 2013) is much more questionable.
THE STRUCTURES OF GIANT VIRUS PARTICLES
Mimiviridae: hairy pseudo-icosahedrons
Historical context
In 1992, Rowbotham (1983) investigated the origin of a pneumonia outbreak in Bradford (UK) using an Acanthamoeba coculture protocol. From the water of a cooling tower, he isolated what light microscopy observations suggested to be a Gram-positive coccoid bacterium that he called ‘Bradford coccus’. However, this microorganism resisted cultivation attempts in various axenic media and characterization using universal 16S rDNA bacterial primers (La Scola et al.2003; Raoult, La Scola and Birtles 2007). Ten years later, the viral nature of this mysterious microorganism was finally elucidated through an electron microscopy (EM) study of its replication cycle and partial genome sequencing (La Scola et al.2003). This was the first example of a giant virus, large enough to be confused with a bacterium. It was officially renamed Acanthamoeba polyphaga Mimivirus (short for ‘MIcrobe MImicking Virus’) (La Scola et al.2003; Raoult et al.2004)
Mimiviridae virions morphologies
Mimivirus virions exhibit an icosahedral protein capsid 440 nm in diameter, making them easily visible without staining under the light microscope (Fig. 1A). The perfect icosahedral symmetry is broken by a five-pronged star structure, called the ‘stargate’ (Zauberman et al.2008) that is present at a single vertex of the particle. Except above the stargate, the capsid surface is covered by a 150-nm-thick fibril layer with a composition thought to be related to peptidoglycan (Piacente et al.2012, 2014b) (Fig. 2A). This reticulated layer, probably responsible for the Gram colouration of Mimivirus particles, also makes them extremely difficult to break open, except using a powerful chemical lysis protocol (Renesto et al.2006). The stargate is the portal through which the internal nucleoid of the particle is unloaded (Zauberman et al.2008). Its opening enables the internal virion lipid membrane layering the protein capsid to deploy and fuse with the phagosomal membrane of Acanthamoeba, resulting in the delivery of the virion nucleoid into the cytoplasm (Figs 2A2 and 3A). The nucleoid is a membrane-bound spherical compartment approximately 320 nm in diameter containing a linear double-stranded (ds) DNA genome of up to 1.26 Mb (Arslan et al.2011) as well as associated proteins necessary to initiate the infection cycle, such as the machinery responsible for early gene transcription. The DNA genome of Mimivirus is not tightly compacted as confirmed by the diffraction pattern of individual virions produced by a hard X-ray free electron laser beam (Seibert et al.2011) and the consecutive 3D reconstruction (Ekeberg et al.2015). A cryo-EM 3D reconstruction of a Mimivirus particle (lacking fibres) has been achieved at 65 Å resolution, which confirmed the large interior volume enclosed by two 70-Å-thick layers (Xiao et al.2009). The overall structure of the particle is well conserved among the various known species of Acanthamoeba-infecting Mimiviridae, except for a reduced thickness of the fibre layer in Megavirus chilensis (75 ± 5 nm) (Arslan et al.2011) and Moumouvirus (100 ± 5 nm) (Yoosuf et al.2012). These differences may be linked to the presence of different acetamido sugars in their composition (Piacente et al.2014b). Morphogenesis of the particle inside infected amoeba has been analysed using a variety of sophisticated techniques, which have shown that lipid membranes are recruited from the endoplasmic reticulum and served as scaffolds to initiate and guide the assembly of the protein capsid, the vertex bearing the stargate budding first from the virion factory (Klose et al.2010; Kuznetsov et al.2010, 2013; Mutsafi et al.2013; Suárez et al.2013).
Four distinct virions. Light microscopy images (scale bar = 2 μm) and scanning EM images (scale bar = 100 nm) of Mimivirus (A, C), Mollivirus (B, D), Pandoravirus (E, G) and Pithovirus (F, H).
Four distinct giant virus particles and replication cycles. Thin-section EM images of Acanthamoeba infected with purified virions. (A) Mimivirus replication cycle: (1) phagocytosis (bacterial mimicry); (2) opening of the ‘stargate’ and delivery of the nucleoid into the cytoplasm; (3) early transcription and protein translation leading to the building of a large electron-dense ‘virion factory’ from the periphery of which a large number of new particles will emerge first empty, then filled up with the nucleoid, then covered with a thick fibre layer. The nucleus remains intact during the whole replication cycle. (B) Pandoravirus replication cycle: (1) phagocytosis; (2) opening of the apical pore and fusion of the virus internal membrane with the vacuole membrane; during the ‘eclipse phase’ the virus genome is transferred to the host's nucleus for transcription and replication; (3) the nucleus membrane is progressively recycled into virus membranes and multiple new virions are synthesized at its periphery. (C) Pithovirus replication cycle: (1) phagocytosis; (2) removal of the ‘cork’ and fusion of the virus internal membrane with the vacuole membrane; there is an ‘eclipse phase’ while the virus genome is transferred to the host's cytoplasm and presumably transcribed by the virion-imported machinery; (3) buildup of a faint cytoplasmic virion factory recognizable by the exclusion of the cell organelles at its periphery; translation of the viral transcripts in the cytoplasm; synthesis of new particles from membrane vesicles of unknown origin; early exit of neo-formed particles through exocytosis prior to massive release following the complete lysis of the host cell. The nucleus (N) remains intact during the whole replication cycle. (D) Mollivirus replication cycle: (1) phagocytosis; (2) opening of the apical pore and fusion of the virus internal membrane with the vacuole membrane; there is an ‘eclipse phase’ while the virus genome is transferred to the host's nucleus and transcribed by the cellular machinery; (3) the nucleus membrane is progressively recycled into virus membranes and multiple new virions are synthesized; early exit of neo-formed particles through exocytosis prior lysis of the host cell. Inset: enlarged view of the virion factory where fibres of unknown composition accumulate and seems to contribute to the virion synthesis.
Schematic representation of the four giant virus infectious cycle. (A) Mimivirus, (B) Pithovirus cytoplasmic infectious cycles, (C) Pandoraviruses and (D) Mollivirus nucleocytoplasmic infectious cycle. While all virions keep their integrity, Mollivirus particles lose their spherical morphologies when in vacuoles.
Pandoravirus: amphora-shaped particles
Historical context
The systematic sampling of a variety of aquatic environments (and their sediments) in search for additional Acanthamoeba-infecting Mimiviridae recently led to the unexpected discovery of the Pandoraviridae family of giant viruses, of which two representatives have been studied in detail (Philippe et al.2013). Like for Mimivirus, they were identified in samples exhibiting strong lytic activity during cocultivation with Acanthamoeba. The first one named Pandoravirus salinus originated from a sample of superficial marine sediment layer from the coast of central Chile, whereas P. dulcis was isolated from mud at the bottom of a freshwater pond near Melbourne, Australia.
Pandoraviridae virions morphologies
After several rounds of amplification in Acanthamoeba, the two Pandoraviruses were readily observable by optical microscopy as a lawn of ovoid particles 0.8–1.2 μm in length and 0.5 μm in diameter (Figs 1E and 2B). Another Pandoravirus sharing the Pandoravirions (Pandoravirus virions) morphology, P. inopinatum, was isolated from a patient infected by Acanthamoeba keratitis and recently sequenced (Scheid, Hauröder and Michel 2010; Scheid, Balczun and Schaub 2014; Antwerpen et al.2015). Electron microscopy imaging revealed ultrastructural features unique among previously described viruses. Thin sections of mature virions revealed a membrane-bound empty-looking compartment encased in a ∼70-nm-thick tegument-like envelope consisting of three layers: an internal layer of light density (∼20 nm), a dark layer composed of a dense mesh of parallel fibrils (∼25 nm) and an external layer of medium density (∼25 nm) (Fig. 2B2 and 4B). One extremity of the particle exhibits an apical pore, the opening of which allows the uncharacterized internal content of the particle to be delivered into the host's cytoplasm, through a channel formed by the fusion of the internal membrane with that of the vacuole. In contrast with the Mimiviridae, the Pandoravirions lack an electron-dense central core, which typically corresponds to the location of the compacted genome. The DNA extracted from the purified particles was a linear double-stranded molecule of 2.77 Mbp for P. salinus, 1.93 Mbp for P. dulcis and 2.24 Mbp for P. inopinatum.
Open membranes intermediates involved in virion assembly. (A) In Mimivirus early virion factory, (B) Pandoravirus, (C) Pithovirus, (D) Mollivirus virion factories. The scale bar corresponds to 500 nm. Black arrows point on open membrane intermediates, the white arrow points on the external layer of Mollivirus particles. The black arrow head points on the 20 nm spaced rings corresponding to the layers of variable lengths fibres surrounding the particles. The white arrow head points to one of the fibres accumulating in the virion factory reproducibly seen during the virion assembly process.
Like for Mimivirus, the astonishing size of these particles (also combined with their unconventional asymmetrical shape) prevented their recognition as viruses 6 years earlier when they were first spotted in Acanthamoeba isolated from the inflamed eye of a patient with keratitis (Scheid et al.2008; Scheid, Hauröder and Michel 2010; Claverie and Abergel 2015).
Pithovirus: a delusive pandoravirus look
Historical context
The identification of two distinct families of giant viruses immediately suggested that more remained to be discovered. Accordingly, the first representative of a third family of giant virus was isolated a year after the Pandoraviruses using the same Acanthamoeba coculture protocol, this time starting from a 30 000-year-old sample of Siberian permafrost (Legendre et al.2014). Initially, this new virus, Pithovirus sibericum, seemed to resemble Pandoraviruses by its dimensions (1.5 μm in length and 0.5 μm in diameter) and particle shape. However, further analyses of its gene content, replication cycle and particle fine structure indicated that it was markedly different from the two previously characterized families of giant viruses.
Pithovirus virion morphology
Imaging of thin sections by TEM revealed that Pithovirus particles are enclosed by a 60-nm-thick electron-dense envelope underlined by a lipid membrane. This internal membrane delimits a compartment mostly devoid of discernible substructures (Figs 1F and 2C), with the exception of an electron-dense sphere (50 nm in diameter) and a tubular structure (observed episodically but in reproducible fashion) parallel to the long axis of the particle. In contrast with Pandoraviruses, the envelope consists of 10-nm-spaced stripes perpendicular to the surface. Mature Pithovirus particles also have an apical orifice that is plugged by a ‘cork’, composed of a highly regular hexagonal honeycomb-like grid. At the initial stage of infection, this cork is expelled in the amoeba vacuole, allowing the uncharacterized content of the particle to be delivered to the host cytoplasm after fusion of the virion internal membrane with that of the phagosome. Following this initial step, similar to that of Pandoraviruses and Mimiviruses, Pithovirus exhibits a markedly different replication cycle (Fig. 3B, see below). However, as for the Pandoraviruses, there is no recognizable structure suggesting the location of the 610-kb dsDNA genome within the Pithovirus virion.
Interestingly, as for the two other giant virus families, a modern relative of Pithovirus was first spotted in Acanthamoeba more than 15 years ago, but misinterpreted as an archaeal endocytobiont (Michel et al.2003a,b).
Mollivirus: a soft mollicute-like virion
Mollivirus virion morphology
Mollivirus was initially spotted using light microscopy as rounded particles multiplying in a culture of A. castellanii inoculated with the same sample of 30 000-year-old Siberian permafrost that produced Pi. sibericum. The particles are roughly 600 nm in diameter and enclosed in a hairy tegument (Fig. 1D and 2D). At least one internal membrane is layering the particle interior. The virions appear crowned by two to four 20-nm-spaced rings corresponding to layers of fibres of different lengths surrounding the particles (Fig. 2D2). The tegument is made of two layers of different densities, including an external layer 10–12 nm thick made of 30–40 nm interspaced parallel strips only visible on EM sections tangent to the tegument (Fig. 4D). A 12 to 14-nm-thick intermediate layer is made of a mesh of fibrils resembling those composing the central layer of the Pandoravirus tegument. At variance with the particles of Pandoraviruses and Pithovirus, the apex aperture of Mollivirus consists of a funnel, 160–200 nm in diameter. Once internalized in host vacuoles, the virions seem to lose their spherical appearance and take a ‘soft’ shape evoking giant clams (Fig. 2D1). As for the other non-icosahedral virions, there is no electron-dense structure suggesting how and where the 650 kb dsDNA Mollivirus genome is packed in the particle.
REPLICATION CYCLE OF GIANT VIRUSES
Despite the considerable differences in the structure of their particles, the four families of giant viruses use the same phagocytosis-dependent entry mechanism to infect their common host (Figs 2A1–D1 and 3). This strategy is directly linked to the heterotrophic nature of Acanthamoeba, for which giant viral particles mimic the bacteria on which they naturally feed. It has long been known that individual particles must be larger than 0.6 μm in diameter to efficiently trigger phagocytosis; otherwise, they must aggregate prior to internalization (Korn and Weisman 1967; Claverie and Abergel 2009). There is thus an evolutionary incentive for viruses that infect Acanthamoeba to maintain their size above this threshold, particularly in diluted aquatic environments where the probability of an encounter between viruses and hosts is low.
After their capture into phagocytic vacuoles, all four giant virus types use the same mechanism to reach the cytoplasm. In response to an unknown biochemical signal, the delivery portal of the particle opens, allowing an internal lipid membrane to unfold, protrude and fuse with that of the vacuole membrane (Figs 2A2–D2 and 3). This creates a channel through which the inner content of the particle is released into the cytoplasm, where the replication cycle can begin. Following this initial step, the fates of the four giant viruses differ (Fig. 2A3–D3 and 3).
Mimiviridae: a giant cytoplasmic virion factory
Initial ultrastructural studies of the replication cycle of Mimivirus focused either on the delivery of the nucleoid to the cytoplasm (to document the opening of the particle and subsequent membrane fusion) (Zauberman et al.2008; Claverie and Abergel 2009) or on the late replication phase (6–12 h post-infection (P.I.)), which involves the formation of an electron-dense virion factory up to several micrometers in diameter surrounded by a large number of neo-synthetized particles budding from its periphery, stargate first (Suzan-Monti et al.2007; Kuznetsov et al.2013; Mutsafi et al.2013). Open membrane intermediates, derived from endoplasmic reticulum membranes, form crescents structures (Fig. 4A) initiating the formation of the virions internal membranes. The four capsid proteins (one major capsid protein and three conserved homologues) are then seeding the protein array by shaping the icosahedral structure through a complex process where the major capsid protein fills the icosahedron faces (Kuznetsov et al.2010; Mutsafi et al.2013; Suárez et al.2013). The first 4–5 h of the cycle were initially thought to correspond to an ‘eclipse phase’, during which initial rounds of viral genome replication were thought to occur in the host nucleus (Suzan-Monti, La Scola and Raoult 2006; Suzan-Monti et al.2007). It is now clear that there is no eclipse phase (Claverie and Abergel 2009; Mutsafi et al.2010, 2014) and that the replication cycle of Mimiviridae entirely proceeds outside of the host nucleus, like for the Poxviruses (Carter et al.2005). Indeed, EM studies have indicated that the inner dense nucleoid of the Mimivirus (or Megavirus) particle remains detectable in the cytoplasm following its extrusion from the particle. This nucleoid (enclosing the viral genome and many associated proteins) initiates early viral gene transcription and DNA replication, behaving as a seed developing into a full blown virion factory (Fig. 2A3) isolated from the cytoplasm by a filamentous mesh (Fig. 4A) (Claverie and Abergel 2009; Mutsafi et al.2010; Arslan et al.2011).
In contrast with other known icosahedral double-stranded DNA viruses, packaging of the genome into Mimivirus particles proceeds through a transient aperture in the centre of an icosahedral face, which is distinct from the vertex portal (the stargate) used for genome delivery to the cytoplasm (Zauberman et al.2008). The final stage of maturation in the cell cytoplasm is the assembly of the peripheral layer of fibrils (Fig. 2A3). The replication cycle terminates 8–12 h P.I. (depending on the species of Acanthamoeba and virus tested), resulting in complete host cell lysis and the release of approximately 1000 mature particles.
Pandoraviridae: eclipse phase, then nucleus disruption
In the case of the Pandoraviruses, the delivery of the empty-looking inner compartment of the particle leads to a bona fide eclipse phase without noticeable change in the infected host cell during the subsequent 2–4 h. Following this initial phase, the host nucleus undergoes a major reorganization, involving a progressive loss of its normal spherical shape and the progressive fading of the electron-dense nucleolus. The nuclear membrane then develops multiple invaginations, resulting in the formation of numerous vesicles (Fig. 4B). Uncharacterized crystalline structures appear at the periphery of the deliquescent nucleus and progressively disappear as the cycle progresses. Approximately 8–10 h P.I., the infected cells lose their adherence and become rounded while new virions start to appear at the periphery of the region formerly occupied by the nucleus (Philippe et al.2013) (Figs 2B3, 3C and 4B).
Unlike Mimiviridae, for which empty particles are first formed then packed with internal components, the peripheral tegument and the inner compartment of Pandoravirus particles seem to be synthesized simultaneously in a continuous process initiated at the apical pore region and ending with mature virions fully closed at the opposite apex. Open membrane intermediates, probably derived from the nuclear membrane, form crescents structures (Fig. 3B) that seem to contribute to the virion formation via a process reminiscent of the assembly of the Poxvirus particles (Suárez et al.2013). The replication cycles of P. salinus and P. dulcis exhibit the same stages and characteristics (Philippe et al.2013) (Figs 2B, 3C and 4B) and lead to cell lysis and the release of about thousand particles 15–18 h P.I. At variance with the Mimiviridae, only 1% of the produced virions appear to be infectious.
Pithovirus: electron-lucent virion factories
Like for the pandoraviruses, the delivery of the uncharacterized inner content of the Pithovirus particle into Acanthamoeba is followed by an eclipse phase lasting from 4 to 6 h P.I. The progression into the infectious cycle coincides with the appearance of a roughly circular electron-lucent cytoplasmic area up to 4 μm in diameter. This presumed cytoplasmic virion factory is made visible by its lack of cytosolic substructures and the exclusion of cellular organelles at its periphery. Numerous electron-dense vesicles then appear at the virion factory boundary, later on followed by newly synthesized virus particles. Again, membrane intermediates of unknown origin accumulate in the virion factory (Fig. 4C). The process of virion formation is reminiscent of that of Pandoravirus, with the envelope and the interior of the Pithovirus virions being ‘knitted’ simultaneously. However, two assembly phases can be distinguished: first rectangular-shaped ‘corked’ particles with a 25-nm-thick wall are made before maturing into ovoid particles enclosed in a 60-nm-thick striated electron-dense tegument (Fig. 3B) (Legendre et al.2014).
Complete lysis of infected cultures occurs approximately 15 h P.I. releasing hundreds of particles. The final burst is preceded by the continuous release of mature particles via exocytosis, starting 8 h P.I (Legendre et al.2014). In contrast to the Pandoraviruses, the host nucleus and nucleolus appear unaffected throughout the whole Pithovirus replication cycle (Fig. 2C3 and 3B). The differences in gene contents, virus-encoded functions and particle proteomes between Pithovirus and Pandoraviruses are consistent with their intracytoplasmic versus nucleus-dependent replication modes, as discussed below.
Mollivirus: eclipse phase and fibrous virion factory
During the Mollivirus infection, the cells remain adherent and keep their trophozoite shape all along the cycle, but there is a decrease in the number of visible vacuoles after 4–5 h. Like for all giant viruses infecting Acanthamoeba, the replication cycle begins with the phagocytosis of Mollivirus particles with up to 10 virions per cell, either scattered in individual vacuoles or gathered in a same vacuole (Fig. 2D1). The opening of the particle was never visualized probably due to the fact that the depth of the funnel is larger than the thickness (∼70 nm) of the ultrathin section. However, the fusion between the underlying lipid membrane and the cellular vacuole membrane and the exchange of material between the virion interior and the cell cytoplasm are obvious (Fig. 2D2). Like for the Pandoraviruses, the cell nucleus becomes disorganized after 3–4 h exhibiting numerous invaginations of the nuclear membrane. Neo-synthesized virions appear soon after at the periphery of the nascent virion factory incorporating the modified nucleus with no visible nucleolus. Cell organelles are excluded from the virion factory. This virion factory (and the nucleus when still visible) is uniquely filled by a unique mesh of fibrillary structures (Fig. 2D3, inset, Fig. 3D). They may correspond to viral proteins involved in the particle assembly as they seem to be progressively incorporated in the virions. The process of virion formation is reminiscent of that of Pandoravirus, the envelope and the interior of the Mollivirus particles being assembled simultaneously except that individual fibres are seen entering the newly formed virions (Figs 3D and 4D). Crescent of lipid membrane also accumulate and may be involved in the assembly of the internal membrane (Figs 3D and 4D). After 6–8 h, particles at various stages of maturation coexist in the same virion factory. New virions start to appear in the growth medium after 6 h in the absence of visible cell lysis and mature virions are often seen in vacuoles suggesting that most of them are released through exocytosis.
GENE CONTENT OF GIANT VIRUSES
The discovery of the giant viruses filled the gap between viruses and cellular life forms in term of physical size and genome complexity (Claverie et al.2006; Raoult and Forterre 2008; Claverie and Abergel 2009, 2013). The four giant virus families also exhibit the largest genomes (with size in excess of 500 kb), a threshold that is occasionally used instead of particle dimension to distinguish giant viruses from other large DNA viruses (Yutin, Wolf and Koonin 2014). Genome sizes range from 0.99 Mbp (Boyer et al.2011) to 1.26 Mbp (Arslan et al.2011) for the Mimiviridae, 2.77, 2.24 and 1.93 Mbp for P. salinus, P. inopinatum and P. dulcis, respectively (Philippe et al.2013), 0.61 Mbp for Pithovirus (Legendre et al.2014) and 0.65 Mbp for Mollivirus. Although the viruses possessing the largest genomes are also the ones exhibiting the largest particles, there is no proportionality between the volume theoretically available for DNA packaging and the actual genome size. Interestingly, despite the ample space that would be available for the accumulation of ‘junk’ DNA, the gene density of the four giant virus genomes is comparable to that of regular-sized viruses and prokaryotes, with about one predicted protein-coding gene per kb and short intergenic regions of about 200 nt in average. This compact organization has been confirmed by transcriptome studies of Mimivirus (Legendre et al.2010, 2011) and Mollivirus (Legendre et al.2015), and in less details for Pithovirus (Legendre et al.2014).
A striking feature shared by all four families of giant viruses is the high proportion of ORFans (i.e. predicted proteins without significant database match) encoded by their genomes. At the time of publication of the first representative of each family, the fraction of ORFans was 76% for Mimivirus (Raoult et al.2004), 67% for Pithovirus (Legendre et al.2014), 84% for P. salinus (Philippe et al.2013) and 65% for Mollivirus (Fig. 5). Similar proportions of ORFans were found in the corresponding virion proteomes, suggesting that they are not the results of bioinformatic errors in gene predictions. This high proportion of ORFans in giant viruses is central to the ongoing debate on the origin and evolution of these viruses. With such divergent virus families, the number of ‘core genes’ defined as those strictly shared by all viruses decreases and tends to zero as we keep expanding our knowledge of virus diversity (Claverie and Abergel 2013). This trend is best illustrated by Pandoraviruses, for which even the most abundant virion protein is an ORFan. It lacks any similarity with the major capsid protein, until now one of the few hallmarks shared by all large dsDNA viruses infecting eukaryotes. Accordingly, the analysis of the ever increasing amount of metagenomic data might only reveal the presence of viruses sharing a substantial gene content and sufficient sequence similarity with members of previously described families.
Predicted proteins encoded in the genomes of giant viruses. Pie charts of the best match against the non-redundant database of all known proteins sequences at the time of the first publication for Mimivirus (Raoult et al.2004), Pandoraviruses (Philippe et al.2013), Pithovirus (Legendre et al.2014) and Mollivirus (Legendre et al.2015). Proteins are classified according to the broad taxonomic assignation of the best matches. It is worth noticing that the majority (83 out of 93) of the virus-like Mollivirus proteins have their best match in Pandoraviridae. In all cases, more than two third of the predicted proteins do not have a BlastP match with an E-value <10−5. There is no consensual theory to explain the large proportion of ORFans although they might be central to the understanding of the origin and evolution of giant (and large) dsDNA viruses.
Finally, despite their distinct properties, it is important to note that all of the known giant viruses still obey Lwoff's original criteria discriminating viruses from cellular organisms: they do not encode ribosomal proteins or ribosomal RNAs, they do not encode an ATP-producing pathway and they do not multiply by division. It is quite remarkable that these rules, which were proposed more than 50 years ago, at a time when only a handful of viruses had been characterized, are still of operational value for determining the nature of giant viruses (Lwoff 1957; Lwoff and Tournier 1966).
Genomic features of the Mimiviruses
All Mimiviruses have a linear A+T rich DNA genome (Table 1). They have been classified in three groups based on their phylogenetic affinity. Group A gathers Mimivirus (Raoult et al.2004), Mamavirus (Colson et al.2011b), Terra2 (Yoosuf et al.2014b) and Samba virus (Campos et al.2014), the largest of this group. The group B corresponds to Moumouvirus (Yoosuf et al.2012) and Moumouvirus monve, and group C corresponds to M. chilensis (Arslan et al.2011), M. Iba (Saadi et al.2013a), M. courdo7, M. courdo11 (Yoosuf et al.2014a) and M. Terra1 (Yoosuf et al.2014b) (Table 1). They were isolated from various environments including fresh water, sea water, sediments, soil, human faeces (Saadi et al.2013b) and even from a leech (Boughalmi et al.2013a). They were propagated using either A. polyphaga or A. castellanii. The more distant members between the three groups are Mimivirus and M. chilensis sharing about 50% of their genes with an average sequence similarity of 50% identical residues at the protein level. About 80% of those not shared by the two viruses correspond to ORFans (Arslan et al.2011; Legendre et al.2012).
Characteristics of giant viruses (grey rows) compared to other large eukaryotic viruses.
| Particle largest | Genome max. | Nuclear phase | ||||||
|---|---|---|---|---|---|---|---|---|
| Family or genus | Name | Virion shape | dimension (nm) | size (kb) | G+C% | Genes | aaRS | |
| Pandoraviridaea | Pandoravirus salinus | Ovoid | 1200 × 500 Ø | 2770 | 64 | 2556 | + | Tyr |
| Trp | ||||||||
| Pithovirusb | Pithovirus sibericum | Ovoid | 1500 × 500 Ø | 610 | 36 | 467 | – | |
| Mollivirusc | Mollivirus sibericum | Ovoid | ∼600 Ø | 651 | 60 | 523 | + | |
| *Mimiviridae group Cd | Megavirus chilensis | Icosahedral | 610 Ø (440 Ø) | 1259 | 25 | 1123 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met Asn | ||||||||
| Trp | ||||||||
| Ile | ||||||||
| *Mimiviridae group Be | Moumouvirus | Icosahedral | 600 Ø (420 Ø) | 1021 | 25 | 915 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Asn | ||||||||
| Ile | ||||||||
| *Mimiviridae group Af | Mimivirus | Icosahedral | 630 Ø (390 Ø) | 1182 | 28 | 1018 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met | ||||||||
| *CroVg | CroV | Icosahedral | 300 Ø | 730 kb | 23 | 544 | – | IleRS |
| *PgVh | PgV | Icosahedral | 460 kb | 32 | 442 | – | ||
| Coccolithoviruses | EhV 86i | Icosahedral | 180 Ø | 407 kb | 40 | 478 | + | |
| Marseilleviridaej | Marseillevirus | Icosahedral | 220 Ø | 368 kb | 45 | 457 | – | |
| Poxviridae | Canarypoxk virus | Ovoid enveloped | 330 × 280 × 200 nm | 365 kb | 30 | 328 | – | |
| Chloroviruses | PBCV-NY2Al | Icosahedral | 200 Ø | 370 kb | 41 | 416 | + |
| Particle largest | Genome max. | Nuclear phase | ||||||
|---|---|---|---|---|---|---|---|---|
| Family or genus | Name | Virion shape | dimension (nm) | size (kb) | G+C% | Genes | aaRS | |
| Pandoraviridaea | Pandoravirus salinus | Ovoid | 1200 × 500 Ø | 2770 | 64 | 2556 | + | Tyr |
| Trp | ||||||||
| Pithovirusb | Pithovirus sibericum | Ovoid | 1500 × 500 Ø | 610 | 36 | 467 | – | |
| Mollivirusc | Mollivirus sibericum | Ovoid | ∼600 Ø | 651 | 60 | 523 | + | |
| *Mimiviridae group Cd | Megavirus chilensis | Icosahedral | 610 Ø (440 Ø) | 1259 | 25 | 1123 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met Asn | ||||||||
| Trp | ||||||||
| Ile | ||||||||
| *Mimiviridae group Be | Moumouvirus | Icosahedral | 600 Ø (420 Ø) | 1021 | 25 | 915 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Asn | ||||||||
| Ile | ||||||||
| *Mimiviridae group Af | Mimivirus | Icosahedral | 630 Ø (390 Ø) | 1182 | 28 | 1018 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met | ||||||||
| *CroVg | CroV | Icosahedral | 300 Ø | 730 kb | 23 | 544 | – | IleRS |
| *PgVh | PgV | Icosahedral | 460 kb | 32 | 442 | – | ||
| Coccolithoviruses | EhV 86i | Icosahedral | 180 Ø | 407 kb | 40 | 478 | + | |
| Marseilleviridaej | Marseillevirus | Icosahedral | 220 Ø | 368 kb | 45 | 457 | – | |
| Poxviridae | Canarypoxk virus | Ovoid enveloped | 330 × 280 × 200 nm | 365 kb | 30 | 328 | – | |
| Chloroviruses | PBCV-NY2Al | Icosahedral | 200 Ø | 370 kb | 41 | 416 | + |
Note: aPandoraviridae (Philippe et al.2013; Antwerpen et al.2015); bPithovirus sibericum (Legendre et al.2014); cMollivirus (Legendre et al.2015), Mimiviridae (dgroup C (Arslan et al.2011; Saadi et al.2013b; Yoosuf et al.2014a,b), egroup B (Yoosuf et al.2012), fgroup A (Raoult et al.2004; Colson et al.2011a; Legendre et al.2011; Campos et al.2014; Yoosuf et al.2014b); gCafeteria roenbergensis virus (Fischer et al.2010; Fischer and Suttle 2011); hPhaeocystis globosa virus (Santini et al.2013); iEmiliania huxleyi virus 86 (Wilson et al.2005); jMarseilleviridae (Boyer et al.2009; Thomas et al.2011; Boughalmi et al.2013b; Aherfi et al.2014; Doutre et al.2014); kCanarypox (Tulman et al.2004; Roberts and Smith 2008); lParamecium bursaria chlorella virus NYA (Fitzgerald et al.2007; Van Etten and Dunigan 2012; Jeanniard et al.2013). Giant viruses visible by light microscopy are marked in grey. Viruses with a known virophage are marked by a red asterisk.
Characteristics of giant viruses (grey rows) compared to other large eukaryotic viruses.
| Particle largest | Genome max. | Nuclear phase | ||||||
|---|---|---|---|---|---|---|---|---|
| Family or genus | Name | Virion shape | dimension (nm) | size (kb) | G+C% | Genes | aaRS | |
| Pandoraviridaea | Pandoravirus salinus | Ovoid | 1200 × 500 Ø | 2770 | 64 | 2556 | + | Tyr |
| Trp | ||||||||
| Pithovirusb | Pithovirus sibericum | Ovoid | 1500 × 500 Ø | 610 | 36 | 467 | – | |
| Mollivirusc | Mollivirus sibericum | Ovoid | ∼600 Ø | 651 | 60 | 523 | + | |
| *Mimiviridae group Cd | Megavirus chilensis | Icosahedral | 610 Ø (440 Ø) | 1259 | 25 | 1123 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met Asn | ||||||||
| Trp | ||||||||
| Ile | ||||||||
| *Mimiviridae group Be | Moumouvirus | Icosahedral | 600 Ø (420 Ø) | 1021 | 25 | 915 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Asn | ||||||||
| Ile | ||||||||
| *Mimiviridae group Af | Mimivirus | Icosahedral | 630 Ø (390 Ø) | 1182 | 28 | 1018 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met | ||||||||
| *CroVg | CroV | Icosahedral | 300 Ø | 730 kb | 23 | 544 | – | IleRS |
| *PgVh | PgV | Icosahedral | 460 kb | 32 | 442 | – | ||
| Coccolithoviruses | EhV 86i | Icosahedral | 180 Ø | 407 kb | 40 | 478 | + | |
| Marseilleviridaej | Marseillevirus | Icosahedral | 220 Ø | 368 kb | 45 | 457 | – | |
| Poxviridae | Canarypoxk virus | Ovoid enveloped | 330 × 280 × 200 nm | 365 kb | 30 | 328 | – | |
| Chloroviruses | PBCV-NY2Al | Icosahedral | 200 Ø | 370 kb | 41 | 416 | + |
| Particle largest | Genome max. | Nuclear phase | ||||||
|---|---|---|---|---|---|---|---|---|
| Family or genus | Name | Virion shape | dimension (nm) | size (kb) | G+C% | Genes | aaRS | |
| Pandoraviridaea | Pandoravirus salinus | Ovoid | 1200 × 500 Ø | 2770 | 64 | 2556 | + | Tyr |
| Trp | ||||||||
| Pithovirusb | Pithovirus sibericum | Ovoid | 1500 × 500 Ø | 610 | 36 | 467 | – | |
| Mollivirusc | Mollivirus sibericum | Ovoid | ∼600 Ø | 651 | 60 | 523 | + | |
| *Mimiviridae group Cd | Megavirus chilensis | Icosahedral | 610 Ø (440 Ø) | 1259 | 25 | 1123 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met Asn | ||||||||
| Trp | ||||||||
| Ile | ||||||||
| *Mimiviridae group Be | Moumouvirus | Icosahedral | 600 Ø (420 Ø) | 1021 | 25 | 915 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Asn | ||||||||
| Ile | ||||||||
| *Mimiviridae group Af | Mimivirus | Icosahedral | 630 Ø (390 Ø) | 1182 | 28 | 1018 | – | Tyr |
| Cys | ||||||||
| Arg | ||||||||
| Met | ||||||||
| *CroVg | CroV | Icosahedral | 300 Ø | 730 kb | 23 | 544 | – | IleRS |
| *PgVh | PgV | Icosahedral | 460 kb | 32 | 442 | – | ||
| Coccolithoviruses | EhV 86i | Icosahedral | 180 Ø | 407 kb | 40 | 478 | + | |
| Marseilleviridaej | Marseillevirus | Icosahedral | 220 Ø | 368 kb | 45 | 457 | – | |
| Poxviridae | Canarypoxk virus | Ovoid enveloped | 330 × 280 × 200 nm | 365 kb | 30 | 328 | – | |
| Chloroviruses | PBCV-NY2Al | Icosahedral | 200 Ø | 370 kb | 41 | 416 | + |
Note: aPandoraviridae (Philippe et al.2013; Antwerpen et al.2015); bPithovirus sibericum (Legendre et al.2014); cMollivirus (Legendre et al.2015), Mimiviridae (dgroup C (Arslan et al.2011; Saadi et al.2013b; Yoosuf et al.2014a,b), egroup B (Yoosuf et al.2012), fgroup A (Raoult et al.2004; Colson et al.2011a; Legendre et al.2011; Campos et al.2014; Yoosuf et al.2014b); gCafeteria roenbergensis virus (Fischer et al.2010; Fischer and Suttle 2011); hPhaeocystis globosa virus (Santini et al.2013); iEmiliania huxleyi virus 86 (Wilson et al.2005); jMarseilleviridae (Boyer et al.2009; Thomas et al.2011; Boughalmi et al.2013b; Aherfi et al.2014; Doutre et al.2014); kCanarypox (Tulman et al.2004; Roberts and Smith 2008); lParamecium bursaria chlorella virus NYA (Fitzgerald et al.2007; Van Etten and Dunigan 2012; Jeanniard et al.2013). Giant viruses visible by light microscopy are marked in grey. Viruses with a known virophage are marked by a red asterisk.
It is becoming clear that besides the truly giant viruses comprising the Mimiviruses, the Mimiviridae family also encompasses an increasing number of smaller (genome and particle wise) representatives infecting various unicellular protists such as the marine phagotrophic flagellate Cafeteria roenbergensis (CroV) (Fischer et al.2010), several unicellular algae (Larsen et al.2008; Johannessen et al.2014) including Phaeocystis globosa (PgV) (Santini et al.2013) and even unknown hosts living in hypersaline cold environments (Yau et al.2011). These viruses (with genomes ranging from 730 to 450 kb) are clearly phylogenetically distinct from other known algal viruses making their default classification as ‘phycodnaviruses’ inappropriate (Santini et al.2013) (Fig. 6).
Gene-content cladistic clustering of the large and giant DNA viruses infecting eukaryotes (from Legendre et al.2015).
Regulatory elements and transcriptional landscape of Mimiviruses
Currently, Mimivirus is the only giant virus for which detailed analyses of the genome, transcriptome and particle proteome have been published (Raoult et al.2004; Renesto et al.2006; Legendre et al.2010, 2011). Initially predicted to encode 911 proteins, Mimivirus was subsequently found to express 26 polyadenylated non-coding mRNAs and 46 additional protein-coding transcripts (Legendre et al.2010). Ultradeep transcriptome sequencing throughout the entire replication cycle of Mimivirus indicated that its 1.18-Mb genome generates 1018 transcripts, among which 979 encode proteins, 6 encode tRNAs (as sporadically found in other large DNA viruses) and 33 correspond to non-coding mRNAs of unknown functions (Legendre et al.2011). The transcriptomic study of infected A. castellanii cells by Mimivirus revealed a number of features which are shared by the member of the A, B and C Mimivirus groups. All predicted genes are transcribed at same point during the infectious cycle and belong to three classes of expression: early, intermediate and late. Regulatory elements have been identified; a strictly conserved promoter element (AAATTGA) is found 5′ of nearly all early expressed genes (Suhre, Audic and Claverie 2005), and a less conserved promoter was identified upstream of the late genes (Legendre et al.2010). Transcription arrest and polyadenylation are governed by what was named the ‘hairpin rule’ (Byrne et al.2009; Arslan et al.2011) along which polyadenylation occurs in a palindromic sequence made of a stem of at least 13 paired nucleotides and a loop of at most 5 nucleotides. This structure, not conserved at the sequence level, is located at the 3′ end of the vast majority of the Mimiviruses genes. The molecular machinery responsible for transcript maturation and polyadenylation corresponds to two conserved clusters of genes responsible for the 3′ end and 5′ end maturation of the viral transcripts (Priet et al.2015). One cluster encodes a protein with tandem RNAse III-like domains recognizing and cleaving inside the hairpin structure, as well as a self-processive polyA polymerase able to add up to 700 nt long polyA tails. The second cluster encodes a trifunctional mRNA capping enzyme, capping and N7-methylating the caps in the 5′ end of the transcripts (Benarroch, Smith and Shuman 2008), as well as a methyltransferase performing the 2′-O cap methylation. The two clusters belong to the class of late expressed genes. The corresponding enzymes are loaded in the virions and readily available for action after the nucleoid is released in the cytoplasm, together with the transcription machinery and DNA repair enzymes initiating transcription. A cap-specific guanine-N2 methyltransferase is forming a 2,7-dimethylguanosine DMG cap which could also favour viral protein synthesis (Benarroch et al.2009).
Mimiviruses are cytoplasmic viruses
All the sequenced Mimiviridae genomes (small and large) encode transcription machinery (Raoult et al.2004; Fischer et al.2010; Colson et al.2011a; Santini et al.2013) consistent with the use of promoter sequences that differ from those of Acanthamoeba (Suhre, Audic and Claverie 2005; Legendre et al.2010). In the case of Mimivirus, this machinery has been shown to be packaged in the particles as proteins (Renesto et al.2006; Claverie, Abergel and Ogata 2009), allowing the infectious cycle to be initiated from within the cytoplasm. As the cycle proceeds, the Mimivirus/Megavirus virion factories grow from the size of ∼0.4 μm (the size of the nucleoid) to 10 μm in diameter (Raoult et al.2004; Suzan-Monti et al.2007; Claverie and Abergel 2009; Mutsafi et al.2010; Arslan et al.2011). The functional resemblance of these virion factories with (transitory) parasitic microorganisms is such that they can be infected by their own viruses, called virophages (La Scola et al.2008; Claverie and Abergel 2009) (Box 2). Accordingly, the virophages use the same regulatory elements than their Mimivirus hosts (a late promoter and the terminal mRNA hairpin structure), relying on the giant virus transcription apparatus to express their genes (Claverie and Abergel 2009; Ruiz-Saenz and Rodas 2010; Desnues, Boyer and Raoult 2012; Tiwari et al.2014).
Mimiviruses specific features
In addition to their own DNA replication and transcription machineries, the Mimiviridae encode a large complement of DNA repair enzymes including remote homologues of Escherichia coli MutS, a component of the mismatch repair system. The sequences of the two viral versions of MutS (MutS7 and MutS8) are sufficiently distinct from their cellular counterparts to serve as specific markers in environmental studies (Ogata et al.2011; Hingamp et al.2013; Wilson et al.2014). For no clear reason, the Mimiviridae also encode their own distinct version of the glutamine-hydrolyzing asparagine synthetase, the sequence of which can serve the same purpose (Mozar and Claverie 2014).
By far the most unexpected finding in the genomes of the largest Mimiviridae is the recurrent presence of up to seven different amino-acyl tRNA synthetases (aaRS); these enzymes are involved in the charging of tRNAs with their cognate amino acids. Four aaRS (ArgRS, CysRS, MetRS, TyrRS) are encoded by Mimivirus (Raoult et al.2004) to which IleRS is added in Moumouvirus (Yoosuf et al.2012), and a further two enzymes (TrpRS and AsnRS) in M. chilensis (Arslan et al.2011) (Table 1). The Mimivirus aaRS are active and the structure of the TyrRS suggested that the enzyme recognizes a two-letter codon due to a shorter anticodon-binding loop (Abergel et al.2007), conserved in all Mimiviridae endowed with this aaRS. There are also several other enzymes involved in translation such as translation initiation and elongation factors and even an autoregulated termination factor the translation of which proceeds through two stop codons via two distinct recoding events, a frameshift and a readthrough (Jeudy et al.2012). The presence of these central components of the translation machinery, normally a trademark of cellular organisms, is central to the ongoing debate about the origin and evolution of giant viruses (see below).
As expected from a viral genome entirely expressed outside of the host nucleus, the genes are not interrupted by spliceosomal introns. A small fraction of genes are interrupted by inteins (Ogata, Raoult and Claverie 2005) and group I self-excising introns including a major capsid protein presenting two type I introns (Azza et al.2009).
Among the 10% of Mimiviridae proteins that are similar to proteins with known functions, in addition to the proteins involved in transcription and translation, the following enzymes have also been characterized in detail: a DNA topoisomerase IB (Benarroch et al.2006), an NAD+-dependent DNA ligase (Benarroch and Shuman 2006), a DNA glycosylase (Imamura et al.2012), the first virally encoded nucleoside diphosphate kinase with a unique affinity for pyrimidine nucleotides due to a shorter Kpn loop conserved in all Mimiviruses (Jeudy et al.2009), a cyclophilin (Thai et al.2008), a lysyl hydroxylase (Luther et al.2011), a sulfhydryl oxidase (Hakim and Fass 2009) and a copper chaperone-independent superoxide dismutase (Lartigue et al.2015). Surprisingly, they encode pathways for biosynthesis of the unusual sugar composing their fibres capsids combining enzymes encountered either in bacteria or eukaryotes and a number of carbohydrate-processing enzymes (Parakkottil Chothi et al.2010; Piacente et al.2012, 2014a,b). Recently, the structure of one of the major protein components of Mimivirus fibres suggested that it could be involved in the viral entry by enzymatically digesting its host cell wall (Klose et al.2015). In most cases, the Mimivirus-encoded enzymes have structural and/or functional features similar to that of their cellular homologues. The puzzling number of additional ORFan genes in each newly sequenced genome, which cannot be explained by lateral gene transfer from the host, suggests that the ancestor of Mimiviridae may have been more complex than its descendants. The presence of protein translation-related enzymes also suggested that the viruses may originate from an ancestral cell belonging to an extinct fourth domain of the tree of life (Raoult et al.2004; Claverie et al.2006; Colson et al.2011a; Desnues, Boyer and Raoult 2012; Legendre et al.2012). The hot debate on alternative evolutionary scenarios (Yutin, Wolf and Koonin 2014) will be presented in a separate section. As additional Mimiviridae were isolated with genomes in the same Mbp-size range, the feeling was growing that an asymptote of maximal genetic complexity had been reached for giant DNA viruses. This belief was soon to be proven wrong by the discovery of the Pandoraviruses.
Genomic features of the Pandoraviridae
In contrast to most large DNA viruses, the nucleotide composition of Pandoraviruses genomes is G+C rich (≈61%). There are at least six copies of a terminal fragment of 50 kb for P. salinus and 1 repeat of a 20-kb terminal fragment for dulcis. A total of 2556 putative protein-coding sequences (CDSs) were identified in the P. salinus 2.47-Mb unique genome sequence (considering a single terminal repeat), 1502 for P. dulcis and 1839 for P. inopinatum. However, the sequence comparison of these predicted genes with those in public databases revealed that 84% were ORFans. The three Pandoraviruses share around 650 genes with an average of 60% sequence identity at the protein level, but surprisingly, one half of their genes are unique to one or the other. The vast majority of Pandoravirus-encoded proteins do not resemble anything even in metagenomics databases, thus ruling out that they might originate from lateral gene transfers from their cellular host (Fig. 5). More than half of the remaining genes with homologues in databases correspond to non-informative structural motifs (such as Morn, ankyrin and F-box) or are homologous to uncharacterized proteins. Despite lacking a recognizable major capsid protein gene, as well as many of the other core genes presumed to be conserved among large dsDNA viruses (Iyer et al.2006), their genomes remain typically virus-like, with the largest fraction of the genes with a predicted function (54 out of the 136) devoted to DNA replication, transcription, repair and nucleotide synthesis (Philippe et al.2013).
Pandoraviruses are nucleocytoplasmic viruses
The Pandoraviruses encode four RNA polymerase subunits (RPB1, RPB2, RPB5, RPB10), together with their own mRNA capping enzyme and at least three other transcription factors. However, the virally encoded transcription machinery is not a component of mature virions, implying that the host nucleus must be actively involved in the early stage of the Pandoravirus infection before it decays, as observed by EM. Furthermore, although they encode their own DNA polymerase, the Pandoraviruses lack other essential components of the DNA replication machinery such as a DNA ligase, topoisomerases and a DNA sliding clamp. This strongly suggests that these viruses could not replicate without the help of host enzymes that are normally found in the nucleus. Finally, the presence of spliceosomal introns in about 10% of the viral genes further suggests that at least part of the Pandoravirus genome is transcribed in the host nucleus.
Another remarkable feature of the Pandoravirus genome is the presence of two amino-acyl-tRNA synthetase (TyrRS and TrpRS) genes, which are also found in M. chilensis. However, these TyrRS and TrpRS are more similar to their Acanthamoeba homologues (57 and 58% identical residues, respectively) than to their Mimiviridae counterparts, suggesting that they might have been acquired from an Acanthamoeba-related host.
The proteomic study of the purified P. salinus particles confirmed that the ORFans corresponded to bona fide proteins, 80% of which without homologues in the public databases. Given this unusual proportion of ORFans, the absence of a major capsid protein homologue, and their unique particle morphology, the Pandoraviruses appear to define a new dsDNA virus family, the Pandoviridae.
Genomic features of Pithovirus
In contrast to Pandoravirus, Pithovirus has a G+C poor (36%) genome that is only 610 kb in size. This genome size is surprisingly small given a particle internal volume that is larger than that of Pandoraviruses (Table 1). Furthermore, the Pithovirus 610-kb genome ‘only’ encodes 467 proteins. Compared with other large DNA viruses, the low coding density (68%) of its genome is due to the presence of a large number of regularly interspersed palindromic motifs. These motifs (about 150 nt in length) are most often found in 2-kb long arrays of tandem repeats. They are not transcribed and do no resemble previously described mobile elements or repeats found in other viral genomes (Allen, Schroeder and Wilson 2006; Delhon et al.2006; Legendre et al.2014). More Pithovirus relatives will need to be studied to determine if this is a general characteristic of the family and how it might be linked to its evolution. The topology of the Pithovirus genome, a terminally redundant circularly permutated linear dsDNA molecule (or a closed circle), is also unique among giant viruses. Such topology is encountered in the Iridoviruses, a family of large icosahedral DNA viruses.
Pithovirus exhibits some affinity with Iridoviruses and Marseilleviruses
Compared to other giant viruses, the gene content of Pithovirus is globally more similar to those of previously described large icosahedral eukaryotic DNA viruses (Legendre et al.2014). This is illustrated by the phylogenetic clustering of the Pithovirus DNA polymerase (Fig. 6) within a clade comprising the Iridoviruses and Marseilleviruses, two distantly related families of eukaryotic large dsDNA viruses. The Pithovirus RNA polymerase subunits RPB1 and RPB2 display a similar clustering pattern, although their sequences remain fairly distant from those of their closest homologues in Marseilleviruses (about 30% identical residues). However, only a third of the 467 predicted Pithovirus proteins (i.e. 152, 32.5%) have recognizable databases homologues, including only 19, 15 and 10 best matching their counterparts in Marseilleviridae, Mimiviridae and Iridoviridae, respectively (Fig. 5). Given the very partial overlap of its gene content with these established DNA virus families together with the unique morphology of its particle, we proposed to classify Pithovirus sibericum as the first member of a new family, the Pithoviridae.
Pithovirus is a cytoplasmic virus
Among the above 152 predicted proteins with a database match, only 125 (26.7% of the total gene content) have previously recognized motifs, most of which are poorly informative such as ubiquitous protein–protein interaction motifs or motifs involved in a large variety of signalling/regulatory pathways. Typical of large DNA viruses, the dominant functional categories were DNA transcription (17 genes), DNA repair (11 genes, including an ATP-dependent DNA ligase), nucleotide synthesis (7 enzymes, including a ThyX alternative thymidylate synthase) and DNA replication (5 genes). In contrast with Pandoraviruses, a proteomic analysis of purified Pi. sibericum particles revealed that the virally encoded transcription machinery is packaged in the mature virion. This includes four RNA polymerase subunits (RPB1, RPB2, RPB5, RPB10), three transcription factors, three helicase-topoisomerases and a mRNA capping enzyme. The presence of these proteins is consistent with the replication cycle of Pithovirus occurring outside of the nucleus, as suggested by EM. The only identified intron corresponds to a self-splicing intron in the gene encoding the DNA-dependent RNA polymerase large subunit. The dsRNA-directed RNAse III homologue found in the Pithovirus virions might be involved in the processing of the viral transcripts, most of which are terminated by a hairpin as found in Mimiviridae (Byrne et al.2009; Legendre et al.2014). As in many icosahedral viruses, but at odds with the Pandoraviruses, the Pithovirus particle contained glycosylated proteins.
Genomic features of Mollivirus
Mollivirus possesses a 60% G+C rich dsDNA linear genome of 651 kb with a 10-kb tandem repeat at each extremity. It encodes 523 predicted proteins out of which 65% are ORFans. Mollivirus thus belongs to an additional family of giant viruses. The rest of its genome can be distributed in three categories: 18% of genes homologous to viral genes (89% of which have their closest relatives in Pandoraviruses), 14% homologous to eukaryotic genes (67% of which have their closest relative in A. castellanii) and 3% homologous to bacterial genes. Overall, 80% of Mollivirus predicted proteins are of unknown function. The weak phylogenetic affinity of Mollivirus sibericum with Pandoraviruses is confirmed by the fact that only 28 of the 136 viral proteins found to constitute Mollivirus particles are homologous to Pandoravirus virion proteins, all of them with vastly different relative abundances. For instance, the first shared protein ranked 9th in Mollivirus virions is ranked 39th in P. salinus virions. Reciprocally, the best ranked (14th) shared protein in P. salinus virions is 55th in Mollivirus virions. Unexpectedly, one of the most abundant proteins in the Mollivirus virion is homologous to the major capsid protein of icosahedral large DNA viruses. Like the Poxviruses D13 protein, this protein may participate in the particle assembly (Bahar et al.2011) despite its lack of final symmetry. The closest relative of this major capsid-like protein is curiously found in A. castellanii that appears to have acquired it from an icosahedral virus infecting unicellular algae (Legendre et al.2015). This Mollivirus gene might thus originate from two successive horizontal transfers first from virus to host, and then from host to Mollivirus.
Mollivirus is a nucleocytoplasmic virus
The absence of transcription machinery in the virion proteome, the aggressive remodelling of the cell nucleus during the infection and the presence of spliceosomal introns in 3% of its genes are all consistent with the necessity for Mollivirus to translocate its genome into the host nucleus to at least initiate its replication cycle. Unexpectedly, the analysis of the virion proteome also revealed the presence of several ribosomal proteins, a unique feature among large DNA viruses, the functional significance of which remains to be elucidated.
CURRENT VIEWS AND DEBATES ON THE EVOLUTION OF GIANT VIRUSES
The virion is not the virus
Until now, viruses were seen as little boxes made of a minimal number of self-associating proteins designed to transport just enough genetic information to encode the box components, seize the centre of command of the host cell (i.e. the nucleus) and order it to replicate the virus genome and translate it into more boxes. As these ‘boxes’ are metabolically inactive, viruses in this traditional sense are deemed not ‘alive', although they exhibit the two properties uniquely associated with living organisms: the capacity to self-replicate and to evolve. The discovery that viruses could have genomes as complex as cellular microorganisms such as bacteria or parasitic eukaryotes led to the realization that the intracellular virion factory, akin to a transient parasitic microorganism using the host cell as a growth medium, should be recognized as the bona fide ‘virus’, whereas the virion (the particle) should be reappraised as a mere vehicle by which the virus (i.e. the virion factory) is propagated from cells to cells. Similarly, the viral genome should not be seen as the mere blueprint of the particle, but as encoding the complex process leading to their multiplication. This shift in perspective was illustrated by noting that if the genome carried by a spermatozoid rightly appears disproportionate for a single cell, the paradox disappears as soon as one remember that its actual purpose is the making of a whole human being (eventually producing more spermatozoids) (Claverie and Abergel 2010). Similarly, the viral genome not only includes the blueprint of the virion, but all the information required to construct a transitory reproductive organ (elle) using the cell as a growth medium to produce more virions. The notion that the Mimiviridae virion factories should be considered as transient ‘living’ microorganisms received additional support with the discovery that they could be ‘infected’ by their own virus (La Scola et al.2008; Fischer and Suttle 2011; Yau et al.2011; Santini et al.2013; Campos et al.2014; Gaia et al.2014; Zhou et al.2015) (Box 2).
The first discovered virophage, Sputnik, is a dsDNA virus with a-20 Kb genome encoding 21 proteins. Sputnik sticks to Mimivirus virion to be internalized with it. Its genome delivery in the host cytoplasm was never visualized but since the Acanthamoeba cell nucleus remains intact all along its replicative cycle, it was assumed that it never enters the cell nucleus and instead uses the Mimivirus virion factory to replicate. The accumulation of neo-synthetized virions in the periphery of the virion factory was the first evidence that Sputnik replicative cycle takes place in Mimivirus virion factory. Another was provided by the analysis of Sputnik genome. Mimivirus late promoter elements are detected upstream of the majority of Sputnik genes and the hairpin structures governing the transcription termination of Mimivirus genes and mRNA polyadenylation are also present in the 3′ UTRs of Sputnik's genes (Claverie and Abergel 2009; Arslan et al.2011). Altogether, this strongly suggests that Sputnik needs a fully mature virion factory to express its genes and replicate. This new kind of parasitism might be a common feature among large DNA viruses infecting eukaryotes, as other virophages were recently discovered associated with more distant members of the Mimiviridae such as Mavirus coreplicating with CroV (Fischer and Suttle 2011) and the defective PgV virophage which does not encode a capsid protein, and is presumably propagated as an episome in PgV virions (Santini et al.2013).
Making a clear distinction between the virion and the virus, the inert particle stage and its intracellular active stage were already proposed in the early days of virology by Lwoff (1957), who used the term ‘virus’ (phage) to designate both the inert particle and the ‘vegetative’ stage corresponding to the actively replicating virus in its host. Such definition immediately reinstates all viruses (not only the giant ones) within the living world.
With the ‘virocell’ concept, Forterre (2013) also focused on the replicative stage of viruses by proposing that viral infection transforms the infected cells into a novel type of cellular organism, whose aim is to produce virions. Given the obligate dependence of viruses on the host translation machinery, he also proposed that viruses originated after ribosome encoding cells (ribocells) (Forterre 2013). Later on, it was proposed to divide biological entities into two groups, the ribosome-encoding organisms, the cells and the capsid-encoding organisms, the viruses (Raoult and Forterre 2008). However, the amphora-shaped viruses demonstrate that the capsid protein cannot be a discriminative criterion to recognize cells from viruses. Without changing the current definition of viruses, particles could come in any shape, and be made of any biochemical material (such as a simple lipid membrane) capable of enclosing nucleic acids. There is a large predominance of icosahedral particles among viruses and the general feeling was that their regular geometrical shape was a consequence of their simplicity, that is the constraint to build identical virions from the spontaneous symmetrical self-assembly of a minimal number of proteins. As a result, microbiologists were not prepared to recognize cell-like shaped virions or particles with heterogenous shapes and sizes. Interestingly, out of the four families of giant viruses known today only one corresponds to the icosahedral model, although with cellular dimensions. When seen in infected cells, members of these new viral families were not originally identified as possible viruses.
Parasitism and dependence toward the host
Since the very beginning of virology, the viruses were recognized as ultrafilterable obligatory parasites before being more formally defined by the Lwoff (1957) criteria. Later it was realized that viruses do not encode ribosomes and were not multiplying by binary fission (Lwoff and Tournier 1966) and that immediately after infection, preceding their multiplication, was an eclipse phase during which an infected cell could not be discriminated from a healthy one. Finally, Lwoff (1957) through his well-known concluding statement ‘viruses are viruses’ argued against any possible continuity between viruses and cells, excluding ipso facto viruses from the cellular world. Even if viruses were considered as intrinsically small and simple, large DNA viruses infecting eukaryotes were already known much before the discovery of the first giant viruses. They belonged to four known families of DNA viruses, Asfar-, Irido-, Pox- and Phycodna- viruses (Box 1). Their particle sizes culminated at 300 nm and their genomes encoded hundreds of proteins. They were mostly spotted by EM observation of infected cells without triggering the idea that there may be a continuum between the smallest and the largest viruses. Based on shared genes between all of them, including Asfar-, Irido-, Pox- and Phycodna- viruses, it was proposed that they evolved from a common ancestor, a nucleocytoplasmic icosahedral virus. As such they were grouped in the NCLDV, corresponding to nucleocytoplasmic large DNA viruses infecting eukaryotes (Iyer, Aravind and Koonin 2001) (Box 1).
With the discovery of the first giant virus, Mimivirus, it became clear that different viruses, even if all obligatory parasites, were showing different level of dependence toward their host, blurring the frontier between cellular parasites and viruses. In the case of Mimivirus, like for Poxviruses (Carter et al.2005), the infection can be followed by microscopy during the whole replicative cycle: there is no eclipse phase. The delivery of the particle inner core within the host cytoplasm is followed by its rapid development into a spectacular intracytoplasmic ‘virion factory’ (Claverie and Abergel 2009; Mutsafi et al.2010). This factory resembles an organelle created de novo to become the site of translation and transcription, as well as replication of the viral genome, using the host's pool of metabolic precursors as well as ribosomes. It is important to notice that this organelle is built from the genetic information provided by the virus and functions independently of the nucleus. At this stage, the resemblance with an intracellular bacterial infection is clear. The numerous cellular and biochemical functions involved in rebuilding from scratch this transient microorganism upon each infection cycle is what justifies the complexity of the giant virus genome. Then, the level of dependence of each virus on its host depends on the proteins packaged in the virion, excluding the structural proteins making the ‘box’. These proteins are the ones initiating the infectious cycle once delivered in the host together with the viral genome.
Viruses all need the host translation machinery but some can also rely on the host nuclear function, such as transcription and even for the most reduced the host replication machinery. Since there is no known example of DNA virus encoding a RNA polymerase but no DNA polymerase it seems that there is a strict evolutionary hierarchy between the two apparatus. However, neither of them is required to make highly successful and ubiquitous viruses such as the Polyomaviruses and the Papillomaviruses. We could thus expect that viruses with the simplest genomes are the most dependent on their host and with the increase in complexity should come a certain level of autonomy. Nuclear viruses such as the adenoviruses, Phycodnaviruses and Herpesviruses encode their own replication machinery but rely on the transcription machinery all along their replicative cycle. Given their genomic complexity, we could have expected that the Pandoravirus would be culminating on the ‘autonomy meter’. Instead, like Mollivirus, Pandoraviruses do not package a transcription apparatus in their virions and thus the largest giant viruses also rely on their host nuclear function. On the other hand, Asfar- (Tulman et al.2009), Pox-, Pithovirus and the Mimiviridae use their own transcription machinery which is unloaded together with their DNA upon infection. They are exclusively cytoplasmic viruses and only rely on their host translation apparatus to translate the virally synthetized mRNAs. Interestingly, this is this last category which can be infected by virophages. With its mere 470 genes, Pithovirus is more autonomous than the Pandoraviruses which are nucleocytoplasmic. So there is no clear correlation between the genome complexity of viruses and their level of dependency toward the host nuclear functions. The only rule is that to be independent from the host nucleus a virus must package an operational transcription apparatus or the cognate mRNAs in the virion.
The differences between the Mimiviridae giant viruses and more host-dependent viruses triggered a hot debate on the model best explaining viruses’ origin and evolution.
Reductive evolution
Reductive genome evolution, a hallmark of all parasitic life forms, corresponds to the irreversible loss of genes. It has been particularly well documented and studied for parasitic bacteria (Moran 1996; Mira, Ochman and Moran 2001; Dufresne, Garczarek and Partensky 2005; Morris, Lenski and Zinser 2012; Martínez-Cano et al.2014). When a microorganism enters into a parasitic association with another one, previously essential functions for a free-living organism can be lost provided they are compensated by the host's homologous function. In addition, intracellular microorganisms tend to exist as smaller and more isolated populations than free-living ones. These conditions accelerate the fixation of neutral of even slightly detrimental mutations through an irreversible evolutionary mechanism known as ‘Muller's ratchet’ (Moran 1996). This slow, cumulative and irreversible loss of functions is the reason why no parasitic microorganism could ever evolve back toward a more independent lifestyle. The rule is among the few biological ones that has no exceptions (Poulin 2007). In the case of symbiosis, gene losses may continue until the parasitic microorganisms become fully integrated in the host cell as an organelle (such as mitochondria), or vanishes altogether, sometimes following the transfer of its genes to the host's genome (Sloan et al.2014). However, for cells living inside another cell, the existence of various compartments delimited by membranes (the cell nucleus on one side and the parasite membrane on the other), as well as regulatory incompatibilities between bacteria and eukaryotic subsystems, makes some functions more challenging to lose than others throughout reductive evolution. The impossibility to import ribosomes across a periplasmic membrane may be why the protein translation apparatus, in particular the ribosomes, are the last to be lost in all cellular parasites. If we now consider that viruses are another example of obligatory parasites, genome reduction might also apply to DNA viruses and provide a simple explanation for the wide range of genome sizes and diversity of gene contents they exhibit (Claverie 2006; Claverie et al.2006). This modern view of virus evolution is a consequence of giant viruses’ discovery. Previous models were the exact opposite and large DNA viruses were considered as efficient ‘pick-pockets of cellular genes’, gaining functions and genes over time, rather than losing them (Iyer et al.2006; Filée, Siguier and Chandler 2007; Moreira and Brochier-Armanet 2008). It was even proposed that viruses were going through recursive phases of accretion and reduction, gaining genes through lateral gene transfer, while losing genes by reductive evolution (Filée 2013). The number of genes estimated to originate from the ancestral virus is around 50 (i.e. the ‘core’ genes with homologues in different virus families; Iyer et al.2006). Thus, all viruses with genomes encoding more than these 50 ‘core’ proteins must have acquired genes. This means that more than 90% of the genome of giant viruses should have been acquired, either through duplication event or by lateral gene transfer from their cellular host or other cellular organisms as proposed in the gene accretion model (viruses are gene pickpocket). However, very few of them correspond to families of paralogues that could be the result of duplication and only a very small percentage (less than 10% for Pandoravirus) of the genes in giant viruses share a common ancestry with genes from cellular organisms (prokaryote or eukaryotes). Thus, if the giant genome of giant viruses was mostly acquired from cellular organisms, these organisms are not around and available for comparison anymore. Since large DNA viruses do not evolve faster than cells (Doutre et al.2014), the absence of recognizable similarity cannot be explained by the fast divergence rate of these viruses. As cellular microorganisms which exhibit various levels of autonomy, from autotrophy to obligate intracellular parasitism, DNA viruses exhibit a gradation in their autonomy toward the host cell within which they replicate. The reductive evolution theory can explain this observed gradation in loss of function as well as why most viruses lack a complete translation machinery. Practically, the first stage of all virus infection is the unloading of the particle content in the host cytoplasm, which means that entirely cytoplasmic viruses have an immediate access to the ribosomes and their host translation machinery. This can explain why, in the hierarchy of loss of function, translation seems to have been lost first. Along the same line, binary fission implies a compartment limited by a physical barrier where chromosomes and organelles are confined, amplified and segregated, and thus have no direct access to the host cytoplasmic ribosomes. Viruses do not divide and are not compartmentalized. As a consequence, cytoplasmic viruses do not need the sophisticated machinery akin to nuclear pores to transport back and forth proteins and mRNA to and from the cytoplasm. While nucleocytoplasmic viruses can use the nuclear machinery to exchange with the cytoplasm, to be able to access the nucleus their genomes need to cross the only physical barrier left which is the nuclear membrane. Nucleic acids can cross the nuclear membrane provided they can be actively transported to the nucleus in the crowded cytoplasm of the host cell, which is especially challenging for giant viruses such as the Pandoraviruses endowed with millions base pair genomes. Once this limit is bypassed, they can lose their transcription machinery and can evolve to use the host transcription. The last tolerated loss is the ability to replicate as long as they can rely on the host replication machinery. Viruses able to transcribe can also replicate their genomes and can even be parasitized by their own virus. As a consequence, one can hypothesize that the incentive for viruses to become nucleocytoplasmic viruses could be to escape the virophage infection by entering the host nucleus. Viruses able to transport their DNA to the nucleus become protected from virophage infection. There, they can use the host transcription machinery at least for the early phase of their infectious cycle. They can either replicate their own genome or, in an ultimate phase of reduction, rely on the host DNA replication machinery.
In the four families of giant viruses reported here, two are cytoplasmic, the Mimiviridae and the Pithovirus, and two are nucleocytoplasmic, the Pandoraviridae and Mollivirus. Again, there is no relationship anymore between genomic complexity and the level of dependence of these viruses.
Giant viruses origin
At the time of Mimivirus discovery, the presence of genes encoding translation enzymes thought to be reserved to the cellular world had two implications. First, it became possible to build a phylogeny using sequences shared by cells and DNA viruses. This phylogeny clearly separated the cluster of all cellular enzymes from that of NCLDV. Second, the presence of translation enzymes in Mimivirus suggested that according to the reductive evolution theory, large DNA viruses could have been endowed with a complete translation machinery and thus could originate from an ancestral cell (Raoult et al.2004). As a result, it was proposed that large DNA viruses corresponded to a fourth domain in the tree of life (Raoult et al.2004). If there is a consensus that the giant Mimiviridae and NCLDV have a common ancestor, the fourth domain of life statement sparked a hot debate (Raoult et al.2004; Moreira and Brochier-Armanet 2008; Moreira and López-García 2009; Boyer et al.2010; Forterre 2010; Williams, Embley and Heinz 2011). It appeared that the diverse large DNA virus families most likely resulted from alternative reductive evolutionary pathways. However, the discovery of the Pandoraviridae, Pithovirus and Mollivirus, all infecting the same host as Mimivirus, is also questioning this possible common ancestry for all viruses. These giant viruses have even less genes in common with the other large DNA viruses and it becomes more and more difficult to consider the ‘core’ genes as a hallmark for all viruses as many of them appear to have been duplicated (Santini et al.2013) opening the opportunity for non-orthologous replacement between different large DNA viruses or even cellular organisms (Yutin and Koonin 2013). For example, the best match of Pandoravirus ‘core’ genes can either be a viral or a cellular gene. Moreover, the virion structures are different from one giant virus family to the other and their proteomes, mostly made of proteins of unknown function, is specific to each virus family. Pandoraviruses and Pithovirus also lack a homologue of the classical major capsid protein. The known families of giant viruses are as different from each other as they are different from extent cellular organisms. Yet, irrespective of their metabolic capacities, genome types, particle structures, sizes or morphologies, they all propagate their genome the same way. According to the reductive evolution scenario, we have to hypothesize that these various virus lineages have different origins, possibly in multiple ancestral protocell types that were once in competition with the one that gave rise to the Last Universal Cellular Ancestor (LUCA), the last universal common ancestor to the Eubacteria, Archaea and Eukarya. The hypothesis that DNA viruses might originate from a time where polyphyletic protocells coexisted was previously discussed (Villarreal and Witzany 2010 and references herein). Yutin, Wolf and Koonin (2014) recently concluded that the paucity of shared genes between different groups of giant viruses effectively rules out their origin from a common giant ancestor. The only alternative to the massive gene gain scenario is then their independent early emergence from multiple ancestral organisms followed by massive losses in the branches leading to the smaller extant viruses. Extending the above fourth domain of life hypothesis, we now have to postulate unknown domains of protocellular life at the roots of each giant viral lineage. These ancestral lineages could have evolved in competition with other life forms until they lost the race and became parasites of the winning cellular lineage that led to LUCA. At the end of a fierce evolutionary battle, the ‘loser’ protocells would have partially survived as parasites (or symbionts) of the ‘winner’ cellular lineages, giving rise to the diversity of DNA viruses we know today. DNA viruses would thus be the descendant of these vanished cell types following a billion year of coevolution in a variety of extant organisms derived from the Last Universal Common Ancestor. Paradoxically, virus-bearing ancestral cells might also have enjoyed a selective evolutionary advantage either through direct import of genes or the acceleration of genes exchanges and nucleic acid shuffling. As speculative as these ideas might seem, the recent discovery of 2.1 billion-year-old fossils likely to correspond to an aborted pre-metazoan lineage (El Albani et al.2010, 2014) suggests that there is still room for imagination and that our present conception of early cellular evolution might be too simple.
Clearly, any theories will remain highly speculative until much progress is made in characterizing the function and structure of the ORFan proteins making the large majority of the proteome of giant viruses. Together with the search for more monster viruses, this should become a priority for the years to come.
We thank all our colleagues from the IGS laboratory whose ten years of hard work provided the matter for this review article. We thank Dr A. Bernadac, Dr A. Kosta from the IMM, Dr J-P. Chauvin, F. Richard, A. Aoune, from the IBDM and S. Nitsche from the CINaM imagery platforms.
FUNDING
This work was partially supported by France Génomique Grant ANR-10-INSB-01–01, the French National Research Agency ANR-14-CE14–0023–01, by the Provence-Alpes-Côte-d'Azur région (2010 12 125), National Centre for Scientific Research, Aix-Marseille University.
Conflict of interest. None declared.
REFERENCES
