Genome of Phaeocystis globosa virus PgV-16T highlights the common ancestry of the largest known DNA viruses infecting eukaryotes

General Genome Features.

The genome of PgV-16T was fully assembled as a linear DNA sequence of 459,984 bp, making it the third-largest described genome among marine viruses behind CroV and Megavirus. Its overall nucleotide composition (A+T content: 68%) is less A+T-biased than that of CroV and Megavirus (A+T: 77% and 75%, respectively), although it reaches 87 ± 8% A+T between genes. Using a conservative annotation protocol (SI Materials and Methods), we identified 434 putative protein-coding sequences with an average length of 320 aa (ranging from 45 to 3,739 aa) and eight tRNAs (three tRNA_Leu, two tRNA_Asn, one tRNA_Arg, one tRNA_Gln, one tRNA_Ile). With an average intergenic distance of 104 nt (summing to 42,717 noncoding base pairs) the PgV-16T genome reaches a coding density of 90.7%. Unexpectedly, the partially assembled genome sequences of two other PgV strains deposited directly in the GenBank database (PgV 12-T, accession no. HQ634147.1; PgV 14-T, accession no. HQ634144.1) were found to be nearly identical to PgV-16T in terms of gene content and percentage of identity (98.1% of identity; 449 and 444 genes). We searched the final genome sequence assembly for the presence of large (direct or inverted) repeated regions of >5 kb. No evidence of block duplication involving a string of several genes was found (Fig. S1), indicating that the large genome size of PgV is not the result of segmental duplications. A finer look at these data also showed that unchecked short-repeat expansions do not contribute to the large size of the PgV genome. As with previously characterized dsDNA viruses infecting unicellular eukaryotes, most of the genes within the PgV-16T genome were vertically inherited from their common ancestor, and recent horizontal gene transfers do not account for a significant fraction of the gene content. Finally, unlike the Poxviruses, Asfarviruses, or Chlorovirus, there is no sequence similarity (direct or inverted) between the two extremities (telomeres) of the PgV genome.

Applying a BlastP E-value conservative cutoff of 10⁻⁵ (and disregarding the matches with the nearly identical partial PgV-12T and PgV-14T genome sequences), 245 coding DNA sequences (CDSs) (56.5%) exhibited a similarity to protein sequences in the National Center for Biotechnology Information (NCBI) nonredundant (NR) databank (14). Among those, 178 (73%) had their closest homolog in a large dsDNA virus, 35 in bacteria, 25 in eukaryotes, and 2 in archaebacteria. Unexpectedly, among the viral best matches, 138 (77.5%) were found in OLPV1 and/or OLPV2 (the DNA polymerase sequences of PgV-16T and the OLPVs share about 48% identical residues) and the only four described genes of P. pouchetii virus (sharing 85% identical residues in average) (Fig. 1). Only four PgV-predicted proteins (PgV140, PgV143, PgV411, PgV423) were found to be most similar to their homolog in Emiliania huxleyi virus (EhV)86 (15) (or related strains), the only other previously sequenced giant viruses known to infect a haptophyte.

Identification of a Virophage-Like Sequence.

Virophages are a class of satellite virus that have now been found in association with giant viruses such Mimivirus/Megavirus (16, 17), as well as with CroV (18) and OLPV (13). Virophages presumably use the characteristic replicative cytoplasmic virion factories of these viruses for their own replication (19). In addition to the single-scaffold PgV-16T genome sequence, the assembly process generated a 19,527-bp-long sequence scaffold, indicating the presence of an unsuspected virophage (denoted PgVV) associated with the PgV-16T particles of Phaeocystis globosa. However, no small viral particle (typically less than 75 nm in diameter) characteristic of previously described virophages were observed in infected P. globosa cells. Upon assembly, the PgVV genome sequence exhibited a much higher coverage (e.g., density of reads mapped to the sequence) than the PgV-16T genome (33,151 ± 14,153 vs. 4,477 ± 1,624), despite a similar G+C content (36% vs. 32%). This suggests that the PgVV genome was in a higher copy number (by a factor of ∼7.5) than the PgV-16T virus in the initial DNA sample. The central PgVV genome sequence is flanked by two 1-kb-long, nearly identical telomeric sequences complementary to each other (Fig. 2). Such a feature, not previously described for virophages, is common among poxviruses, chloroviruses, and asfarviruses and was also encountered in the plasmid prophage N15 (20). In all these cases, this genome feature was associated with covalently closed hairpin ends (telomeres). The PgVV genome was predicted to encode 16 proteins (numbered pgvv01 to pgvv16), all of them transcribed from the same DNA strand (Fig. 2), another unique characteristic. However, as for the previously described virophages, most of these protein sequences (10/16; 62.5%) had no significant matches or were remotely similar to hypothetical proteins of unknown function (2/16; 12.5%). Three of them, including an endonuclease (pgvv02) and the predicted DNAprimase/polymerase (pgvv04) have a homolog in the virophage infecting CroV (Mavirus) and the one associated with OLPVs (OLV1) (Fig. 2). Because no PgVV genes appeared to encode a product even remotely similar to a capsid protein, we suspected that the PgVV genome could either be carried as a linear plasmid within the PgV-16T particle or could integrate within the PgV genome, as described previously for the Sputnik virophage (17). We explored this eventuality by generating a map of the “broken” paired-end reads perfectly matching on the PgVV genome on one end and on the PgV-16T genome on the other end (Fig. S2). A total of 796 pair-end reads were found to meet these criteria, suggesting a multiplicity of integration sites. A possible integration “hot spot” defined by 28 read pairs was found between positions 47,500 and 50,000 in the PgV genome sequence (ORF no. 42), and positions 3200 and 4200 in the PgVV genome (ORF nos. 3 and 4). The presence of PgVV DNA packed in the PgV-16T virions was directly validated by PCR on DNA extracted from particles produced from an independent culture and purified by centrifugation on a cesium chloride gradient (SI Materials and Methods). Finally, pgvv14 is similar to a protein found in several copies in the PgV genome, supporting the possible role of virophages as gene shuttles between large DNA viruses.

Gene Content.

Using a BlastP with a conservative E-value threshold of 10⁻⁵, we found 48 genes shared by Megavirus, Mimivirus, CroV, and PgV, the four largest fully sequenced, viral genomes known (Fig. 3 A and B). Among those, only 18 correspond to genes defined previously as “core genes,” that is, genes found in a majority of large dsDNA viruses and presumed to encode functions essential for the replication of these viruses (Table S1) (2, 21). This finding suggests that these four viruses share these 48 genes as a result of a common ancestry, rather than functional requirements imposed on all large DNA viruses, in particular, those replicating in the cytoplasm of their host cell. Similar three-way comparisons with the Emiliania huxleyi virus (EhV) (having the fifth largest viral genome at 407 kb) (15) resulted in a very different picture, with only 14 genes in common, 11 of which are bona fide core genes (Fig. 3 C and D) encoding the fundamental components of the DNA replication and transcription machineries and a major capsid protein. Thus, even though EhV and PgV are two large DNA viruses infecting two plankton species belonging to the haptophytes, they do not show extensive evidence of common ancestry and clearly belong to different virus families.

The 48 genes (Table S1) shared between PgV and the three other largest sequenced viral genomes [but also found in the draft genome sequences of other Megavirus relatives (17) and, for most of them, within the partial OLPV1 and OLPV2 sequences] are devoted to essential functions such as deoxynucleotide synthesis (ribonucleotide reductase, deoxycytidine deaminase, thymidine kinase, thymidylate synthase, Nudix-like hydrolase); DNA replication [primase, ribonuclease H1, DNA polymerase B, PCNA clamp, three clamp loaders (replication factor C), NAD-dependent DNA ligase, topoisomerase IA and II]; DNA repair (Lambda-type exonuclease, MutS7, Holliday junction resolvase, XRN 5′–3′ exonuclease); DNA transcription (Ribonuclease III, VV A18-like and VV D11-like helicases, TATA-box–binding protein, polyadenylate polymerase, transcription factors TFIIB-like, VLTF3-like, and TFIIS-like); protein synthesis and processing (mRNA-capping enzyme, translation initiation factor eIF4E-like, DNAJ, thiol oxidoreductase, cysteine protease); and virion structure (capsid protein and DNA packaging ATPase). However, seven of these genes uniquely conserved among these giant viruses are also predicted to encode proteins of unknown function (“hypothetical proteins”: PgV50, -58, -134, -183, -259, -260, -298) or were similar to enzymes performing seemingly insignificant roles such as a phospholipase (PgV117) or a methyltransferase (PgV234). Most curious among these well-conserved homologs was asparagine synthetase (>38% identical residues), which occurred in PgV-16T, Megavirus, and Mimivirus isolates, in OLPV1 and OLPV2, and Micromonas and Ostreococcus viruses. A phylogenetic tree built from these viral sequences was amazingly congruent with the global similarity of these viruses and clearly suggests that all known viral asparagine synthetase genes derived from a common ancestor that diverged from the cellular version quite a long time ago and were since vertically inherited within two distinct viral lineages (Fig. S3). The rationale behind the conserved occurrence of the asparagine synthetase gene in the genome of this subset of dsDNA viruses, or the role that asparagine might specifically play in their lifecycle, is unknown.

Of the 434 putative CDS mapped in the PgV genome, only 150 (34.6%) could be assigned a putative function based on homology, a percentage typical of large eukaryotic viruses. This small fraction of known genes was nevertheless sufficient to draw a number of conclusions.

For instance, the PgV genome encodes a high number of RNA polymerase II subunit homologs (RPB1, RPB2, RPB5, RPB6, RPB7, RPB9, and RPB10), a complete set of transcription initiation (PgV113), elongation (PgV146, PgV158), and termination factors (PgV321), together with a polyA polymerase (PgV 73), a TATA-box-like–binding protein (PgV48), and a mRNA-capping enzyme (PgV235). This impressive complement of transcription-related genes suggests that PgV most likely replicates in its host’s cytoplasm, even though its infectious cycle has not yet been studied at the cellular level.

In addition to the previously cited DNA manipulating enzymes common to the four largest known viral genomes, PgV possesses DNA-repair enzymes encountered in a more sporadic manner in a variety of other viruses: an ERCC4-type nuclease (PgV318, with homologs in OLPV and Asfarviruses), an apurinic-apyrimidinic exoDNase (PgV94, with a homolog in CroV), and a mismatch repair enzyme (PgV283) of a type (MutS8) uniquely found in OLPV1 and -2, Phaeocystis pouchetii virus, and Pyramimonas orientalis virus (22). The PgV genome is also rich in putative ribonucleases such as a bacterial ribonuclease HII-like (PgV121, with homologs in CroV and EhV), a RNase T-like (PgV123, with homologs found in many viruses and phage but very divergent in Megavirus and Mimivirus), and a PIN-domain containing RNase H (PgV303, with homologs in CroV and phages).

The PgV genome also contains a large number of putative endonucleases of the HNH-type (PgV60, -72, -243, -438) and GIY-YIG type (PgV69, -216, -359, -429), but none of them was found associated to an intein, or a type I intron (although such introns cannot be reliably identified when they occur within a protein without a significant homolog in the databases). A single intein was identified within the coding region of the VV A18-like helicase (PgV190). The closest relative of the predicted mature PgV190 gene product was found in OLPV1, but this homolog exhibits no trace of intein. The intein was likely inserted into the PgV190-coding sequence after PgV diverged from the OLPV lineage. Two predicted transposases with a strong similarity (77% identical residues) were also identified (PgV137, -217), the sole viral homologs of which could be found in OLPV1 (also in two copies). The presence of several DNA methylases and type II restriction endonucleases (with their best viral homologs in CroV or Ostreococcus viruses) suggests that the PgV genome contains modified nucleotides (such as 5mC or 6mA), as already well documented for Paramecium bursaria Chlorella virus 1 (23).

Surprisingly, given the high A+T content of their genomes, both Megavirus and Mimivirus lack deoxyuridine 5′-triphosphate nucleotidohydrolase (dUTPase), an enzyme crucial to preventing the misincorporation of deoxyuridine instead of thymidine. This enzyme is present in most large eukaryotic DNA viruses (including CroV and EhV), as well as in phages. Thus, the most parsimonious scenario is that the dUTPase gene, present in the common ancestor of all these viruses, was uniquely lost in the branch leading to Megavirus and Mimivirus. The presence of a dUTPase homolog in PgV (PgV389, most similar to the one of OLPV1) adds support to this hypothesis.

Large dsDNA Virus Core Gene Duplication.

After the first round of gene annotation, a number of redundancies were found within the subset of 150 PgV-16T genes with putative functional assignments. Multiple copies of genes encoding a similar function could arise in three ways: (i) vertical ancestral inheritance; (ii) acquisition of additional copies by horizontal transfer; and (iii) in situ gene duplication. Molecular phylogenetic approaches were of little help in deciphering these mechanisms, given the small number of available viral homologs for these genes. Therefore, we used the following simple heuristics to determine which mechanisms were most likely responsible for the duplications: when the gene copies were found to be most similar to each other (relative to a whole NR database search), they were classified as “recent duplication”; if they were separately found to be most similar to their homologs in OLPV (currently the closest known relative of PgV-16T), the duplication was classified as “vertically inherited”; all other cases were presumed to indicate an acquisition by horizontal gene transfer. Table 1 summarizes the results of this analysis, pointing out eight cases of probable duplication since the divergence of the OLPV and PgV lineages, including that of a transposase that might itself be involved in the duplication process. This analysis led us to unravel two unprecedented cases of “core gene” duplication within giant viruses. The first one involves the A32-like packaging ATPase (NCVOG0249) that is uniquely found in two tandem copies in PgV-16T (PgV64, PgV65), most likely the result of a recent event. The second case involves the DNA-directed RNA polymerase second largest subunit (PgV208, PgV228) (NCVOG0271). In this case, the duplication was most likely inherited from the PgV-OLPV common ancestor, because two copies of this RNA polymerase subunit are also present in OLPV. Phylogenetic analyses are consistent with these interpretations (Fig. S4 A and B).The packaging ATPase and the RNA polymerase are both “viral hallmark genes” encoding key functions that are shared by a diverse subset of viruses (24) and are usually found in a single copy per viral genome. However, nothing in these sequences suggests that one of the two copies is a defective, nonfunctioning gene.

Discovery of a Putative Mobile Element: PgV-MIGE.

The above systematic search for genes individually duplicated in the PgV-16T genome led us to identify 12 copies of a PgV-specific protein (e.g., without significant similarity in the NR database) for which no function could be predicted. These repeated ORFans are evenly dispersed throughout the genome, as indicated by their gene numbers (9, 29, 45, 78, 139, 220, 273, 297, 322, 333, 370, 433) and transcribed from both strands (six from each). These ORFans averaged 320 aa in length (range 270 to 355 aa), with an average pairwise similarity of about 58% (range, 42–65%). The N-terminal halves of the predicted proteins exhibited lower conservation than the C-terminal halves. An exception was PgV333, the most divergent (and shortest) paralog from which the last 50 residues were missing. In a multiple alignment of the 11 full-length homologs, 95 residues are strictly conserved among the 170 C-terminal positions (56%), suggestive of a constrained functional signature. However, submitting this multiple alignment to the sensitive 3D structure homology search program Fugue (25) failed to return any significant match or functional clue. To further analyze the primary structure of the DNA segments encompassing and flanking these multiple ORF copies, we focused on the eight most similar ones (average pairwise identity of 63%). A local alignment of their nucleotide sequences identified a highly conserved DNA segment of about 300 nt, beginning immediately upstream of the ORF N terminus (Fig. S5). The nucleotide conservation of this noncoding segment contrasts with the flanking N-terminal part of the ORFan coding region that is not well-conserved (Fig. S5). A huge diversity of mobile elements has been described in the three domains of life (26–28), and the list is still growing (29). However, the PgV-16T repeats do not resemble any of them. For instance, examining the sequences and putative secondary structures flanking the 5′ and 3′ boundaries of these repeated segments failed to reveal any direct or inverted repeats, or putative target sequences shared by the various repeated elements (Fig. S5). The conserved predicted protein associated with these repeats is also puzzling. To our knowledge, the coding regions associated with large mobile genetic elements (even those using a rolling-circle replication process) always encode proteins bearing a recognizable similarity with DNA-processing enzymes (e.g., homing endonucleases, transposases, integrases, reverse transcriptase, etc.). This is not the case here. The presence of a highly conserved noncoding region within the repeat is also a remarkable feature (Fig. S6). We can only hypothesize that these two conserved elements are involved in the molecular mechanism leading to their duplication. Given their lack of resemblance to any previously described mobile genetic element, we named these PgV-specific repeats: “PgV major interspersed genomic element” (PgV-MIGE).

Estimating the Actual Genome Sizes of OLPV1 and OLPV2, the Closest Relatives to PgV.

The PgV-16T genome can be used to estimate the level of completion of the OLPV1 and OLPV2 partial genomes assembled from metagenomic data. Fig. S6 indicates that a total of 145 predicted proteins are shared between OLPV1 or OLPV2 and PgV. Given that OLPV1 and OLPV2 are much closer to each other (their shared predicted proteins are about 80% identical) than they are from PgV (OLPV-PgV orthologous protein share 50% identical residues in average), it is reasonable to assume that most genes shared between PgV and OLPV1 or OLPV2 should also be present in both OLPVs. Following this hypothesis, we can now perform a simple statistical computation: out of the 145 “conserved genes,” only 121 are retrieved in the available OLPV2 draft genome. This might be taken to indicate that this genome is only 121/145 (83.4%) complete. Reciprocally, only 125 of these “conserved” genes occur in the draft OLPV1 genome, which, in turn, suggests that it is 125/145 (86%) complete. Applying these proportions, we can now estimate the total number of genes for these two viruses as 401/0.86 = 466 genes for OLPV1, and 326/0.834 = 390 genes for OLPV2. According to this extrapolation, OLPV1, OLPV2, and PgV-16T would possess numbers of genes falling within the variation range exhibited by close members from the same virus family, a result consistent with the fact that PgV proteins have OLPV proteins as their closest homologs.

Phaeocystis Globosa Culture and Virus Purification.

Exponential axenic cultures of Phaeocystis globosa Pg-G (Groningen University Culture Collection, The Netherlands), cultured according to ref. 34 with 40 µmol quanta⋅m⁻²⋅s⁻¹ illumination, were infected with axenic PgV-16T at a virus to algal host ratio of 10 allowing a one-step virus growth cycle (Fig. S8). Upon full lysis of the host, viruses were concentrated by ultrafiltration on Vivaflow 200 (molecular mass cutoff, 30 kDa; Vivascience), after which the viral concentrate was clarified of cell debris by low speed centrifugation at 7,500 × g for 30 min at 4 °C. The supernatant was decanted, and viral particles were concentrated by ultracentrifugation at 141,000 × g for 2 h at 8 °C using a TFF55.38 rotor. Viral pellets were resuspended in TE buffer (pH 8.0) and stored frozen until further use.

Genome Sequencing and Assembly.

Around 1 µg of phenol-chloroform–extracted PgV-16T virus DNA was used to construct a paired-end library, with ∼280-bp insert size, as described by the manufacturer’s (Illumina) protocol. Around 60 million paired-end reads were produced on an Illumina HiSeq2000 corresponding to ∼6 Gb of raw nucleotide sequence. SI Materials and Methods contains further details and the bioinformatic procedures of genome assembly and annotation.