Distant Mimivirus relative with a larger genome highlights the fundamental features of Megaviridae

Megavirus and Mimivirus virion particles exhibit a very similar overall morphology, with a dense core nucleocapsid encased into an icosasedral-like capsid, itself covered by a layer of fibers. However, details make them readily recognizable, even in mixed culture (Fig. 1A): The Megavirus fibers are noticeably shorter (75 ± 5 nm compared with 120 ± 5 nm for Mimivirus) and cover a capsid slightly larger in diameter (440 ± 10 nm, compared with 390 ± 10 nm for Mimivirus). These dimensions refer to dehydrated virions as prepared for thin section transmission electronic microscopy (TEM), which induce an ≈20% shrinking of the capsids (17). Native Megavirus icosasedral capsids are thus 520 nm in diameter, corresponding to a total particle diameter of 680 nm. In addition, the “hair” of Megavirus virions often exhibits one or two patches of slightly longer and denser fibers (nicknamed “cowlicks”) (Fig. 1A Inset). The Megavirus particles exhibit a clearly visible special vertex (Fig. 1B), corresponding to the “Stargate” already described for Mimivirus, a five-pronged star structure the opening of which triggers the release of the nucleocapsid into the host cell cytoplasm (11, 14). Upon protease treatments, Megavirus particles appeared more fragile than the Mimivirus’ ones, showing a larger proportion of damaged particles in TEM preparation.

The Megavirus isolate was initially tested for growth on A. griffini, A. polyphaga, and A.castellanii, and found to replicate in these three species. For comparison purpose, the replication of Megavirus was then studied in detail on A. castellanii (Neff American Type Culture Collection 30010). The infectious cycle of Megavirus lasts 17 h from the initial phagocytosis event to the time of maximal release of particles from fully mature virion factories, compared with 12 h for Mimivirus (MOI ≈ 10). The difference is mostly due to a slower progression from the “seed stage” (Fig. 1 C and D) to the fully bloomed virion factories (Fig. 1 G and H). Another macroscopic phenomenon specific of Megavirus is that ≈35% of the A. castellanii cells appear to die without undergoing productive infections. The reasons for this cytotoxicity is unknown but suggest that the laboratory A. castellanii strain behaves as a nonoptimal substitute for the unknown natural environmental host of Megavirus. We previously observed that Mimivirus induced a rounding of the infected A. castellanii cells, ≈6 h after infection. The same phenomenon, albeit slightly delayed, was observed with Megavirus. However, at variance with Mimivirus where rounded cells remained adherent, Megavirus infection caused most of them to come off their support, making the cultures more difficult to monitor by regular microscopy. A. castellanii cells infected by Megavirus exhibited three distinctive ultrastructural features in succession, as described for Mimivirus (11, 14, 18). First, the “seed,” corresponding to the Megavirus core nucleocapsid extracted from the external particle layers, appeared clearly separated from the cell cytoplasm by a well-delineated lipid membrane (Fig. 1 C and D). This membrane likely derives from the most internal of the two membranes visible inside the virus particles, as expected from the infection process (opening of the stargate, followed by the fusion of the first virus membrane with the vacuole membrane). These seeds then progressively turn into early virion factories recognizable as electron-dense nucleic acid-rich regions (18) isolated from the surrounding cytoplasm by an exclusion zone constituted of a mesh of fibrils 16 nm in diameter (Fig. 1 E and F). The biochemical nature of these fibrils remains to be characterized. Finally, fully mature virion factories release a large number of particles from their periphery where three stages of maturation are seen: empty assembled particles “budding” from the factory edges, “bald” particles filled with the nucleocapsid, and fiber-covered mature particles ready to exit the host cell (Fig. 1 G and H). The burst size of Megavirus-infected A. castellanii cells is approximately one-half the thousand virions released by those infected by Mimivirus.

The genome of Megavirus is a linear double-stranded DNA molecule with a size of 1,259,197 bp (74.76% A+T), making it the largest described viral genome. The whole sequence was assembled and corrected at once from a dataset of 278,663 454-titanium reads combined with 42,288,396 Illumina Hiseq (paired-end) reads (SI Materials and Methods). We annotated 1,120 putative protein-coding sequences (CDSs) and 3 tRNAs (1 Trp, and 2 Leu). Megavirus CDSs range from 29 aa to 2,908 aa in length, for an average of 338.4 aa (median = 281.5 aa). The distance separating consecutive CDSs was short (114.5 nt in average), resulting into a very high coding density (90.14%). Eight hundred sixty-two of the 1120 (77%) predicted Megavirus CDS have homologs in Mimivirus, whereas 793 of the 979 (81%) Mimivirus CDSs (19) have homologs in Megavirus. Using the best reciprocal match criterion (SI Materials and Methods), Megavirus and Mimivirus were found to share 594 orthologous proteins exhibiting a broad distribution of similarity centered on an average of 50% identical residues (Table S1A and Figs. S1 and S2). Most likely inherited from a Megavirus/Mimivirus common ancestor, the corresponding gene set provides a minimal estimate of the core genome of ancestral Megaviridae.

Fig. 2 illustrates at one glance both the overall similarity and divergence of the Megavirus and Mimivirus genomes. They display a large central region of colinearity extending from mg210 to mg804 in Megavirus, and L192 to R730 in Mimivirus. This region is solely disrupted by the inversion of a central 338-kb genome segment and the translocation of a 76-kb distal segment. Interestingly, one of the boundary of the inversion coincides with the slope reversal of the A+C excess profile, a position associated to the origin of replication in bacteria (Fig. S3). However, this quasiperfect colinearity abruptly vanishes at the two extremities of the Megavirus chromosome, respectively corresponding to 193 kb and 327 kb. Although these regions still encompass several hundred of homologous genes, their location and orientation appear extensively shuffled (Fig. 2). The process leading to the total loss of colinearity at the genome extremities remains unknown, because it is not correlated with a local enrichment in transposases, or to more divergent orthologous sequences. For instance, the highest conservation (93% identical residues) was found for a predicted cholinesterase (Mimivirus L906/Megavirus mg981) located at the extremity of the chromosomes.

Interestingly, the same pattern is observed when comparing two poxviruses exhibiting a level of sequence divergence comparable to the one between Megavirus and Mimivirus (e.g., DNA polymerases exhibiting 65% identical residues) (Fig. S4). This observation suggests that Poxviruses and Megaviridae, despite their considerable differences, might share a genome replication strategy (e.g., coupling replication with recombination) (20, 21) that favor the rearrangement, gain, or loss of genes at the extremities of viral chromosomes. Again, the dramatic 190-kb genome reduction exhibited by a recently described spontaneous Mimivirus mutant is mainly due to large deletions occurring at both ends of the genome (22).

Previous analyses of the Mimivirus genome uncovered two distinctive features within its noncoding moiety. The first one was the perfect conservation of the octameric motif “AAAATTGA” in front of 45% of Mimivirus CDSs (23). The second was the presence of unrelated palindromic sequences (capable of generating hairpins with a minimal stem length of 15 bp) at the 3′ end of 72% of all Mimivirus mapped transcripts (12). The AAAATTGA motif was later shown to be strongly correlated to early expressed transcripts (24), and the predicted hairpins were demonstrated to serve as polyadenylation signals (12, 24). We examined the 100-nt region upstream of the predicted start codon of the 1,120 Megavirus CDSs. Overall 446 (40%) were found to contain the exact AAAATTGA motif. This proportion was 33.8% (201/594) among Megavirus genes with orthologs in Mimivirus. For 170 of these (85%), the AAAATTGA sequence was also found upstream of Mimivirus orthologs. This ratio shows that (i) Megavirus and Mimivirus are using the same motif to specify early gene expression, and (ii) that the expression pattern of orthologous genes is globally well conserved. Detailed statistics on the distribution of early and late promoter elements are presented in Table S1B.

Similarly, we searched the Megavirus genome 3′ intergenic regions for palindromic sequences obeying the same constraints used to identify them in Mimivirus. Nine hundred fifty-four Megavirus CDSs (85%) genes were found to be followed by a suitable predicted hairpin. The proportion was also 85% for Megavirus genes orthologous to Mimivirus genes exhibiting a hairpin. This correlation again suggests that these hairpins are the termination signal of Megavirus transcripts and that the transcript structures are well conserved between the two viruses. This prediction was experimentally verified by sequencing the 3′ end of the cDNA of Megavirus mg464, orthologous to Mimivirus major capsid (MCP) gene, L425. These two CDSs, 78% identical at the nucleotide level, encode two proteins sharing 79% of identical residues. As shown in Fig. S5, the polyadenylation of the Megavirus MCP mRNA occurs within the predicted hairpin, albeit 14 nucleotides upstream of the site used in Mimivirus. Given the low level of sequence similarity between these 3′ UTRs (<49% identical), this result provides the demonstration that the polyadenylation of Megaviridae transcripts is uniquely guided by secondary structure information, rather than a sequence signal.

In keeping with the cytoplasmic localization of its replication cycle, Megavirus possesses orthologs of all of the genes previously predicted to encode components of Mimivirus’ transcription machinery: the two largest RNA polymerase subunits (mg373, mg339); 11 additional transcription factors: mg307, mg332, mg344, mg577, mg519, mg563, mg552, mg544, mg462, mg438, mg414, and one TATA-box binding like protein (mg435); and one polyA polymerase (mg561). All those genes are located in the middle of the colinear Megavirus/Mimivirus genome segments and exhibit a higher-than-average sequence similarity (64 ± 10% of identical residues).

A recurrent feature of viral genomes is the small proportion of genes for which a function can be predicted (e.g., Megavirus possesses 610 anonymous genes, 54.5% of its genome). It is then tempting to use the intensity of expression of a given gene as an indicator of its importance (i.e., essentiality), eventually to prioritize functional studies. Unexpectedly, this rationale appears to be false, because the expression level of Mimivirus genes exhibited no correlation with their conservation in Megavirus (Fig. S6).

Mimivirus was found to possess many genes never before identified outside of cellular genomes (Table S1C). The most unexpected were those related to protein translation, a trademark of cellular organisms. Determining whether these oddities were anecdotal (e.g., due to random horizontal gene transfers) or fundamentally linked to the origin and evolution of all Megaviridae required the identification of a Mimivirus relative at a suitable evolutionary distance. The Mimivirus genome exhibits eight components central to protein translation, including the first four AARS (ArgRS, CysRS, MetRS, and TyrRS) ever found in a virus (2). Megavirus orthologs were found for all of them (respectively encoded by mg804, mg807, mg771, mg907), strongly suggesting that they were present in the Megavirus/Mimivirus common ancestor. More unexpectedly, three additional AARS are found in the Megavirus genome. The AsnRS (mg743) is a member of the class-II AARS, whereas both the TrpRS (mg844) and the IleRS (mg358) are class-I AARS, like all of the ones found in Mimivirus. The significance of this finding is twofold: first, it demonstrates that viral AARS are not limited to class-I enzymes, and second, it makes the scenario of independent acquisition of these genes by HGT increasingly unlikely. Interestingly, the 730-kb genome of the Cafeteria roenbergensis virus, a very distant relative of Mimivirus, also encodes an IleRS (25). Both of these viral IleRS sequences are from the cytoplasmic archaeal/eukaryotic type, but do not exhibit phylogenetic affinity with a known eukaryotic clade (Fig. 3). The Megavirus AsnRS is branching at the origin of the mitochondrial enzyme type, before the radiation of the eukaryotes (Fig. S7A). The Megavirus TrpRS is of the archeal/eukaryotic type, again branching out before the radiation of the eukaryotes (Fig. S7B). In our opinion, these three additional Megavirus AARS were part of an ancestral Megaviridae genome and were subsequently lost in the Mimivirus lineage. These seven AARS were most likely the remnant of a complete set of 20 AARS inherited from an ancestral cellular genome.

The Mimivirus genome was also found unusually packed with DNA repair enzymes, capable of correcting damages caused by UV light, ionizing radiation, or chemical mutagens (2). Megavirus exhibits orthologs to all these previously identified genes, including a specific type of mismatch repair enzyme MutS (mg543) common to large DNA viruses and specifically abundant in the marine environment (26). In addition, Megavirus exhibits a DNA photolyase (mg779), an enzyme using the energy of light to repair thymidine dimers. A remnant of this gene, interrupted by a transposase, can be found in Mimivirus (R853-R855). The presence of this functional photolyase in Megavirus, among its many other DNA repair enzymes, might participate to its increased resistance to UV irradiation: At a level sufficient to totally inactivate Mimivirus (30 min, 20 cm under a 30 W UV lamp, 253 nm), close to 100% of Megavirus particles remained infectious. One hour of irradiation at the same intensity was required to cause a 90% decrease in Megavirus infectivity.

The three types of topoisomerases found in Mimivirus: one of type II (R480), a bacterial-like type I (L221), and a pox-like type I (R194) are also found in Megavirus (mg403, mg323, mg859). However, Megavirus mg859 is a much closer homolog (50% identical) to the topoisomerase 1b found in CroV (crov152), suggesting that the Mimivirus R194 gene has a distinct origin.

Mimivirus exhibited a number of enzymes related to protein folding, and various sugar and amino acid manipulating enzymes not usually found in viruses. With the exception of two enzymes of the cholesterol biosynthetic pathways, all these genes have well conserved homologs in Megavirus (Table S1C), showing that they are part of the Megaviridae core gene set.

Finally, Megavirus mg431 encodes a uridine monophosphate kinase, an enzyme previously undescribed in a DNA virus. This enzyme is 44% identical to its closest homologs in bacteria where it is the rate-limiting enzyme of the pyrimidine salvage pathway (Table S1C).

The progressive accumulation of point mutations (random drift) appears to be the main cause of the divergence between the Megavirus and Mimivirus genomes. The orthologous CDSs, as well as the “orthologous” intergenic regions, exhibit an average similarity of 65% identical nucleotides, higher than the 50% identical residues shared by orthologous proteins (Fig. S1).

Two hundred fifty-eight Megavirus CDSs exhibit no obvious homolog in Mimivirus and, reciprocally, 186 Mimivirus CDSs have no homolog in Megavirus. More than 85% of these lineage-specific CDSs correspond to proteins without functional predictions. They appear to cluster at the ends of the Megavirus chromosome (Fig. 2B). These CDSs might also correspond to fast evolving proteins pushed below our similarity threshold (e < 10⁻⁵). More likely, they correspond to lineage specific losses along the Mimivirus or Megaviruses branches. Genome reduction is a universal, fast, and irreversible process among intracellular parasitic microorganisms (27, 28) that might also apply to the evolution of Megaviridae from their more complex ancestors (3). An alternate scenario would attribute the 258 Megavirus “private” genes to horizontal gene transfers (HGT) that occurred after the divergence of the Megavirus/Mimivirus branches. However, when screened against the NR protein sequence database, these Megavirus private genes exhibited a much lower percentage of matches (17% vs. 52%) than the genes with orthologs in Mimivirus. Moreover, the few matching genes (44/258) exhibited no peculiar affinity with potential gene donors. This result argues against recent HGTs (at least from known viruses or cellular organisms) as the major cause in the difference in gene content between Megavirus and Mimivirus.

The difference in the gene content of the two Megaviridae is also due to the differential expansion or reduction of large paralogous families. Six hundred fifty-two of the Megavirus predicted proteins are in one copy (not matching elsewhere in the genome at e < 10⁻⁵), representing 58.2% of the gene content. The corresponding number is 585 (59.8%) for Mimivirus. The genome of Megavirus is thus truly more complex than the one of Mimivirus and not simply repeat-rich or more redundant. Although the distributions of single- vs. multiple-copy genes are globally similar for the two viruses, specific protein families experienced lineage specific expansions. For instance, the third largest Mimivirus gene cluster (referred to as the N172 L cluster; ref. 29) corresponding to 14 paralogues is represented by a single copy in Megavirus. Conversely, the set of 10 FNIP repeat-containing proteins constituting the N165 cluster in Mimivirus (29) is inflated to 55 members (mg34 paralogues) in Megavirus.

Finally, only a few differences between the Megavirus and Mimivirus genomes resulted from lineage-specific movements of mobile elements. Megavirus exhibits five segmented CDSs. The major capsid protein (mg464, with two introns), the largest RNA polymerase subunit (mg373, with one intron), and the DNA polymerase (mg582, with one intein) exhibit the same exact topology than their Mimivirus orthologs. Orf mg500, encoding a HSP70 chaperonin-like protein, contains a type I intron, whereas its Mimivirus ortholog (L393) does not. The second largest RNA polymerase subunit (mg339 orthologous to Mimivirus L244) exhibited the most intricate rearrangements with changes in the numbers and locations of introns, and the insertion of an intein (30) (Fig. S8). The duplication and movement of transposases (five in Mimivirus; only two in Megavirus) also caused the disruption of a Mimivirus photolyase (R853-R855) and a second photolyase paralogue in Megavirus (mg400).

The purified Megavirus pellet was resuspended in 50 mM Tris⋅HCl at pH 7.5 and incubated for 30 min at 37 °C in the presence of 1.25 mg/mL lysozyme. Lysis was performed by adding 0.1 mg/mL proteinase K incubated 30 min at 55 °C, 1% laurylsulfate 30 min at room temperature, and 10 mM DTT overnight at room temperature. After phenol-chloroforme extraction and alcohol precipitation, the recovered DNA was resuspended in RNase-Dnase free water.

The A. castellanii-infected cells were washed in PBS and resuspended in 2% paraformaldehyde 0.5% glutaraldehyde in PBS for 1 h at room temperature. After three washes in PBS, cell pellet was fixed in 2% osmium tetroxyde, washed once in PBS, dehydrated in 70, 95, and 100% ethanol, and embedded in Epon-812. Ultrathin sections were poststained with 4% uranyl acetate and lead citrate and were observed by using FEI Tecnai TEM operating at 120 kV.

SI Materials and Methods contains the details of the procedures used for the virus isolation and its genome sequencing, assembly, and annotation.