For several years, the investigation of novel dsDNA viruses has excited the imagination of virologists and evolutionists, prompting the development of several original theories about the evolution of certain dsDNA viruses and their role in shaping the genome of the organisms of the tree of life. It has been proposed that dsDNA viruses were the evolutionary origin of the eukaryotic nucleus or at least that they have contributed to its creation by transferring genetic information to the nucleus 1234567.

The recent discovery and characterization of several large dsDNA viruses from aquatic environments belonging to the phycodnavirus and mimivirus families led to the development of new hypotheses 8910. A comparative analysis of the gene content of these viruses with poxviruses, iridoviruses and asfarviruses indicated that they have nine gene products in common, and 33 more gene products are present in at least two of these families. It follows that they might have a common evolutionary ancestor, a nucleocytoplasmic large dsDNA virus (NCLDV) 1112. Although most evolutionists seem to accept the hypothesis of a common ancestor, there is no consensus concerning the general morphology of the ancestral NCLDV and how it evolved to give rise to the different classes of viruses. On the one hand, the ancestral NCLDV is believed to have been a dsDNA virus which has evolved by acquiring genes from the host and bacterial endosymbionts and gene duplication 13. On the other hand, the ancestral NCLDV could have been a huge virus or even a cellular organism that evolved through genome regression 9141516.

A well-studied NCLDV is Ectocarpus siliculosus virus-1 (EsV-1), a phaeovirus that infects a small marine filamentous alga, Ectocarpus siliculosus. Phaeoviruses belong to the Phycodnavirus family. They possess icosahedral morphologies with internal lipid membranes and large double-stranded DNA genomes 17. EsV-1 and other phaeoviruses only infect free-swimming, wall-less gametes or spores. One copy of the DNA is integrated into the cellular genome and then transmitted via mitosis through all cell generations of the developing host 181920. The viral genome remains latent in vegetative cells and is expressed in cells of the reproductive algal organs, sporangia and gametangia only when stimulated by, for example, changes in light composition and temperature 1721. Infected algae show no apparent growth or developmental defects other than partial or total inhibition of reproduction and phaeoviruses are pandemic in several brown algal species examined 17. In addition to its lysogenic life cycle, the genome structure is the major feature which distinguishes EsV-1 from other NCLDVs. The genome is circular, but sequencing the genome resulted in a linear molecule of 335,593 bp ending in inverted repeats. Such a molecule might be able to assemble and form a circular genome 2223. It also contains numerous single-stranded regions randomly distributed over the genome whose functions remain unknown 24.

Another characteristic of the genome is its low gene density compared to other NCLDV genomes. The 231 genes occupy only 70 % of EsV-1 genome; these were assembled in islands of densely packed genes that are separated by large regions of DNA repeats and noncoding sequences 23. In fact, the EsV-1 genome resembles a small eukaryotic chromosome more than a typical viral genome. A comparative analysis of the EsV-1 genome with the genome of two other phycodnaviruses indicated that they could have evolved by loss of genes from a common ancestor, possibly an endosymbiotic organism, thereby supporting the theory of genome regression 25.

To elucidate how the EsV-1 genome is replicated, we first tried to determine where the viral genome is integrated into the host genome. We screened a cosmid library of Ectocarpus algae infected by EsV-1 using labelled EsV-1 DNA as a probe. Sequencing and analysis showed that the host chromosomal DNA was interspersed with single copies of small fragments of viral DNA. However, the fragments diverged in their DNA sequence from the corresponding EsV-1 DNA sequences. In the vicinity of these segments, we found other viral genes and pseudo-genes, some of which are present only in the genome of the mimivirus. This finding indicates that the Ectocarpus genome contains remnants of the genome of either one giant dsDNA virus or one unicellular organism; this organism could be the ancestor of the contemporary NCLDVs.

Furthermore, this hypothesis is supported by previous data, indicating that the viral integrated fragments might be reassembled to constitute the EsV-1 genome. A putative protein coded by the EsV-1 and algal genome might be a key enzyme of this mechanism due to its similarity to tyrosine recombinases.

To determine where the viral genome is integrated in the host chromosome, we screened a genomic library of Ectocarpus siliculosus sporophyte NZVicZ14 infected by EsV-1. The strain NZVicZ14 was created by mating a male gamete from the strain NZ88 15d2 m containing the EsV-1 genome with a non-infected female gamete from the strain Vic88 12–15 f 26. The DNA used for constructing the NZVicZ14 library was isolated from filaments prior to the appearance of symptoms and EsV-1 virions in order to limit the number of positive clones containing viral DNA from EsV-1 particles. The library was screened with radio-labeled EsV-1-DNA that had been digested with the DNA restriction enzyme Sau3A. The screening isolated 200 positive clones whose cloned fragments have an average size of 35 kb. To distinguish clones containing EsV-1 DNA linked to algal DNA from clones containing only EsV-1 DNA, the isolated positive clones were screened with labelled algal DNA from Vic88 12–15 f female gametophytes.

Surprisingly, all the clones were positive, indicating that they also have algal DNA. Consequently, no clone possessing only EsV-1 DNA was isolated. This result may be due either to the absence of viral particles in the algae or to the single-stranded DNA regions of the EsV-1 genome: these regions make the viral DNA fragile and therefore might render large DNA fragments of EsV-1 in the cosmid vector unclonable 22.

The terminal regions of the EsV-1 genome are thought to be recognized by a putative protelomerase coded by the EsV-1 genome which could in turn linearize the EsV-1 genome and form telomeres 27. So far, however, experiments to investigate this possibility have not succeeded. To investigate the hypothesis that the terminal repeats might also be the sequences involved in the integration of the viral DNA into the host genome, a second screen was performed using the probes A and A' corresponding to the terminal regions of the viral genome (Fig. 1). Of the 200 positive clones previously isolated, only four overlapping clones hybridized to both probes. Therefore, other clones have viral DNA that is not related to the terminal region. The DNA fragments of the positive clones can be assembled to form a 51-kb genomic region; such a region has been almost entirely sequenced and analyzed (Table 1 and Fig. 1, region A). The genomic region contains a ca. 5430 bp fragment that is 98 % identical to a segment overlapping the terminal region of the EsV-1 genome (Table 1, Fig. 1). The region of identity of the viral fragment and the EsV-1 genome is located just upstream of the promoter region of ORF 1 and ends approximately in the middle of ORF 231 (Fig. 1). The protein product of ORF 231 has no counterpart in the protein data banks 23. Moreover, a recent blast search of the region indicated that ORF 231 is probably not a gene, since pieces of another ORF situated in ORF 231 code for remnants of a transposase which is related to those coded by the EsV-1 genome (Fig. 1) 23.

The integrated viral fragment contains the same DNA inverted repeats in the same orientation as that which characterizes the EsV-1 genome. Moreover, region A contains additional similar repeats at one extremity (Fig. 1). Apparently, the terminal DNA repeats of EsV-1 genome are longer than previously reported 23 (unpublished results). The discrepancy in the DNA sequence between the proviral and EsV-1 DNA sequences is due to small mutations, deletions or insertions, indicating that the integrated fragments have either accumulated mutations during the evolution or come from an Ectocarpus virus other than EsV-1.

Several genes and remnants of genes were found in the vicinity of the viral fragments. Two homologues of ORF 213 of EsV-1 are present in tandem, upstream of ORF 1 (213h1 and 213h2, Fig. 1). ORF 213 is capable of encoding a large enzyme; this enzyme might be involved in DNA integration and recombination (Fig. 2). Protein 213 possesses at its C-terminus a tyrosine-recombinase domain resembling those found in the members of the integrase family whose hallmark is an invariant pattern of four amino acids (Arg, His, Arg, Tyr) forming the catalytic site 28. ORF 213h1 and 213h2 do not have any similarity in their DNA sequences with ORF 213, indicating that ORF 213h1 and 213h2 are either algal genes or viral genes that NZVic14 horizontally acquired from another virus. The second copy of the gene is truncated at its C terminus resulting in the absence of a part of the tyrosine recombinase domain (Fig. 1 and 2). A homologue of ORF 213 was also found in the genome of another phaeovirus, the Feldmannia irregularis virus (FirrV-1), signifying that protein 213 might be essential for the life cycle of the phaeoviruses 25.

A putative transposase, Tn1, which shares identity with eukaryotic transposases of the Fot 1 family, is also present. It is predicted to contain a putative intron with 5'-AG/GUGAGG and 3'UGCAG/GU splice site sequences as well as a branch point UCAC. The genes of EsV-1 genome do not contain introns, whereas these are common in the algal genes. This therefore supports the contention that region A is composed of viral DNA framed by algal DNA.

In order to know whether other clones contain viral DNA, several DNA inserts were partially sequenced and analyzed. Viral DNA identical to EsV-1 DNA was missing. Of these sequenced inserts, one contains viral DNA at one extremity (Table 1; Fig. 1, region B). This clone, which has been sequenced to completion and analyzed, contains two fragments which have 88 and 94 % identity to EsV-1 DNA (Fig. 1 and Table 2). The fragments possess pseudo- and homologues of ORF 20, 22 and 23 from EsV-1 (Fig. 1). They are arranged as in the EsV-1 genome but diverge at the DNA level. Moreover, region B has a pseudogene whose translated product contains FNIP repeats. The pseudogene is situated between ORF 20 and ORF 22 instead of ORF 21 (Fig. 1). Interestingly, the EsV-1 genome does not contain FNIP repeats; in contrast, the mimivirus genome codes for at least 8 proteins containing this class of repeat 929. Remarkably, the region contains another pseudogene coding for an origin binding protein (OBP); this OBP has its best identity with the MIMIR1 protein, one of the two putative OBPs of the giant mimivirus (Fig. 3 and Table 2). Although the identity is low, the UL-9 helicase domain with its ATP-binding site and DEAD site is still recognizable (Fig. 3). The OBP is also coded by the genome of the Herpes virus and Asfarvirus but is absent in the EsV-1 genome. OBP protein plays a major role in the replication of viral DNA 30. In addition, region B contains relics of transposase as does region A (Fig. 1B).

The presence of an OBP-like gene immediately raises the question whether other mimiviral-like genes are present elsewhere in the NZVicV14 Ectocarpus genome. Other DNA inserts were sequenced and analyzed. A genomic region of 70 kb was assembled with fragments of positive clones (region C, Fig. 1C). Region C contains a small viral DNA insert having a few repeats that are almost identical to those present in region D of the EsV-1 genome (Fig. 1 and Table 1). Surprisingly, it also has a complete ORF, which is capable of encoding a large OBP. This putative OBP shares identity with MIMIR8, the second and largest OBP coded by the genome of the mimivirus (Fig. 3). Like its MIMIR8 and herpes viral counterparts, this OBP is made of the N-terminal domain of the archaeo-eukaryotic primase (AEP) fused to the UL-9-like helicase domain containing an ATP-binding site and DEAD box 31. However, the Zinc domain present in MIMIR8 is absent in Ectocarpus OBP. No additional ORF of viral origin was found except vestiges of genes coding for tetratricopeptide-repeats (TPR) and ankyrins (Fig. 1). Both kinds of repeats are the most frequent motif detected in putative ORFs of the mimivirus, while ankyrin-repeats were detected only in the EsV-1 genome 923. Current sequencing of other clones indicates that the Ectocarpus genome contains several other genes of EsV-1 that could be assembled to form an apparently complete viral genome.

The variation in GC content in chromosomes is a good indicator of recent transposition and integration occurrences. The GC content, which is close to 52 % for the three genomic regions and the integrated viral fragments, shows no traces of a recent integration event (Table 1). Moreover, the GC content of the genomic regions is identical to that of the EsV-1 genome, which is much more than the 28 % GC content of the mimivirus 239. In addition, the GC content of the OBP-like gene of the region C is 47 %, whereas it is only 28 % for the mimiviral gene (Table 2) 9. Other reported ORFs have a similar GC content close to 52 % (Table 2). Taken together, these data suggest that the integration event of an EsV-like virus must have occurred a long time ago.

The huge genome of certain NCLDVs is believed to be mainly due to HGT from the different hosts they might have infected during their evolution and to gene duplication. However, such an explanation is not sufficient to account, for example, for the enormous genome of the mimivirus. Here, we show that the genome of a multicellular eukaryote has acquired genes from a giant DNA virus or a unicellular organism probably as big as a mimivirus but nevertheless more similar to EsV-1 at a DNA level. A comparison with the EsV-1 and other NCLDV genomes indicates that the ancestor probably lost genes during evolution, thus corroborating the genome accretion theory. Consequently, the virus should be seen not only as a gene robber but also as a major actor in shaping the genome of brown algae; it may act to shuffle genes between species. We know that EsV-1 can infect other, related brown algae 17. Moreover, building on former results, we propose that phaeoviruses have probably developed an original replication system. Further studies will be necessary to confirm this model.

ND conceived of the study, carried out the molecular genetic studies, provided the biological interpretation of the results, and finalized the manuscript. WB drafted the manuscript. All authors read and approved the final manuscript.