Viruses infecting O. tauri
could be found frequently in water samples collected from marine lagoons and open sea in the N. W. Mediterranean. They were found about five times more frequently in lagoons than in the sea (51% of samples vs. 11% of samples, respectively), perhaps reflecting the preferred habitat of its clade C 
host. Detailed ecological observations and analysis of their phylogenies will be published elsewhere. Here we focus our attention on one virus, OtV5, isolated from the Bages lagoon on 24th
January 2006, which lysed host cultures reproducibly and rapidly (within 2 days, and ). We tested the strain specificity of OtV5 on diverse picoeukaryotic strains of algae that were isolated and cloned mostly from local seawater samples between January 2006 and March 2007. Firstly, using 5 different genera of Prasinophyceae or 2 different clades of Ostreococcus spp
., no strains were susceptible. Secondly, none of a set of 9 independent and genetically polymorphic clade C strains was attacked by OtV5, which thus shows very narrow strain specificity. While further studies with larger sets of hosts and strains are required but are beyond the scope of this study, we note that this situation may resemble the gene-for-gene resistance arms race observed in higher plant-pathogen interactions 
, where natural selection drives diversity in plant disease resistance genes. Relatively little is known about the selective forces driving the diversity and speciation of microbes in the marine environment, but viruses appear to play a key role in these processes that furnish the base of the global food web 
The extremely small O. tauri
cell is crowded with organelles, which are often observed to deform the apposed plasmalemma, and the cytoplasmic compartment represents a maximum of about 30% of the O. tauri
cell volume 
. If all of this were packed with viruses, about 100 particles might be accommodated. However this volume also accommodates vesicles and hundreds of ribosomes, virus particles being localized to a limited part of the cytoplasm. It is thus not surprising that the mean burst size is only 25 particles per cell. EM pictures showed a maximum of 15 particles in a 50nm thick section of a single cell. These data are comparable to the burst size of 70–100 particles per cell observed for the related host/virus system Micromonas pusilla/MpV 
, considering the host cell volume of Micromonas
is about 4 times more voluminous than Ostreococcus
(the diameter of O. tauri
in our culture conditions is about 1 μm  and Micromonas
is 1.6 μm 
Using a high moi, EM also showed that many virus particles could adsorb to a single cell () but that high numbers of viruses apparently adsorbed to only a proportion of the cells (about 20%). This observation was confirmed by flow cytometry at lower moi (at a moi of one, about 35% of viruses were adsorbed). The remaining majority of particles then remained in suspension following inoculation despite the presence of an actively growing population of host cells (), suggesting that such particles lacked a host cell attachment function, and/or that the remaining cells were resistant to infection. The latter hypothesis, however, seems unlikely given the clonal nature of the host strain. The above information, together with the actual number of infected host cells estimated by pfu (see the “Results” section), suggests that viral attachment may be a limiting step in the infection. At high moi 5 min after inoculation, a clear layer about 50nm thick separated the particles from the cytoplasm (). Since apparently few of the viruses have formed a “bridge” with the cytoplasm at this stage (the EM slice is 50nm thick), we favor the hypothesis that the cell is protected by a cellular envelope, never previously observed in this algal genus, that is invisible in our EM preparations. The biochemical nature of this putative envelope remains to be determined.
We could not however use EM images from high moi to determine when the particles “injected” their electron-dense contents, because all of the cells with a large numbers of viruses attached deteriorated rapidly (<20min, not shown) and disappeared from the culture, perhaps due to multiple perforations of the membrane or to degradation caused by injection of unknown virion proteins, involved in virulence, as apparently may occur in related NCLDV 
. Host cells with “empty” virus particles attached () were rarely seen at any stage of the infection, suggesting that they might usually detach from the cells after injecting their contents.
The average size of predicted CDSs in the linear double-stranded 186,234 bp DNA genome is similar to other phycodnaviruses. One hundred and fourteen CDSs (42.5%) encoded predicted proteins with similarities to other hypothetical proteins in databases, but only 60 of these (22.4%) showed similarity with proteins to which a function has been attributed. This last category may, however, be overestimated as some of these may not be bona fide CDSs. Fifty four (20.1%) CDSs matched other hypothetical CDSs of unknown function, of which 28 (10.5%) came from other viruses, and 26 (9.7%) from elsewhere, and 154 (57.5%) of genes were unknown, Finally, predicted products of 154 genes (57.5%) correspond to no known protein, emphasizing the extremely diverse range of uncharacterized biological functions potentially represented in the phycodnaviruses.
In common with other algal viruses (Table S2
), numerous enzymes involved in DNA replication and gene expression are predicted. This is an ancient common feature of the super-group of NCLDV, including phycodnaviruses, which retain a set of predicted functions that presumably assure their life-cycles in the host cytoplasm. Like PBVC-1, the OtV5 genome shows terminal inverted repeat sequences, and its ends are thus likely to form hairpin loops 
. Our phylogenetic reconstruction shows that OtV5 clearly clusters with phycodnaviruses infecting the prasinophyte Micromonas pusilla 
. These latter viruses form the Prasinovirus clade within Phycodnaviridae, and Prasinovirus should then include OtV5, from a prasinophyte host. Our tree is different from that of Schroeder et al. (2002), because we use different methods for the analysis, giving higher node support values (posterior probabilities). However, in both trees, Phycodnaviridae (Phaeovirus+Coccolithovirus+Prasinovirus+Chlorovirus+Prymnesiovirus) do not form a monophyletic group, with Herpesviridae and Heterosigma akashiwo
virus nested inside (). NCLDV, which attack hosts as diverse as man, reptiles, fishes, invertebrates and algae, are thought to have evolved from a common ancestor, with a core set of common genes 
. Like some of its poxviral cousins 
, OtV5 probably accumulates embedded in a cytoplasmic region rich in a viral-A type inclusion protein, since a CDS whose product shows similarity to this protein was identified ( and Genbank accession number EU304328).
We found no evidence of site-specific endonucleases which are present or predicted in some other previously characterized phycodnaviruses
, although similarities to 3 different methylation/modification enzymes are present. DNA degradation in host Chlorella
cells by PBCV-1 (Paramecium bursaria Chlorella
virus 1)is an early viral function, occurring within 5 minutes of host cell penetration 
, in stark contrast to OtV5, where host chromosomes remain intact throughout the viral life-cycle (). In this respect, OtV5 resembles MT325, which also lacks predicted endonucleases 
. However, in contrast to MT325, where complete viral genomes are present immediately after inoculation, OtV5 genomic DNA could not be visualized by PFGE within the first 2h after inoculation. Since flow cytometrical observations and pfu measurements show that about 35% of the cells have adhered viruses at this stage, we conclude that this level is below the level of detection by PFGE, and that viral genomes become visible after their replication inside host cells.
OtV5 encodes 3 predicted adenine methyltransferases that are possibly involved in protection of the host and/or the viral DNA, for example from the kind of programmed cell death response observed in other green algae 
, or from attack by a third party invader. Like its larger mimivirus relative 
, OtV5 encodes an asparagine synthase (AS) gene, and several viral tRNAs are predicted, including Asn-tRNA. In green plants, AS is a cornerstone of light-regulated protein synthesis, controlling the ratio of carbon and nitrogen anabolites available for protein or sugar syntheses diurnally 
. OtV5 carries a CDS with similarity to FtsH metalloendopeptidase that is absent in other characterized eukaryotic algal genomes. Perhaps this enzyme performs a similar function to its homologue found in photosynthetic bacteria and plants, where it is involved in the repair maintenance of photosystem II 
, since the chloroplast remains intact throughout the viral life-cycle, and most likely functional as the starch grain continues to enlarge. However, all of the above hypotheses remain speculative and must await experimental investigation.
Interestingly, we found some indications for the capture of host DNA within the viral genome, since 6 predicted host-like sequences (HLS) present similarity to host genes (GenBank accession number EU304328). These are unlikely to be very recent transfers, because their GC content is clearly virus-like, (45%) rather than host-like (58%). Nevertheless they may represent an important evolutionary possibility for transfer of genetic information between host cells, since the amino acid sequences gave very high BLASTP scores with Ostreococcus
spp. proteins, and have thus not had time to degenerate extensively by mutations, deletions, or other rearrangements. Remarkably, one of these CDSs encodes an apparently complete proline dehydrogenase gene. This enzyme is important for proline catabolism in all cellular organisms, is known to play a role in stress response in plants and in some other algae 
, and is used as a source of nutrients by numerous bacterial symbionts and pathogens 
; it could thus play an important role in some aspect of the host-virus interaction. Phylogenetic analysis suggests that the OtV5 gene could originate from bacteria (). The AT content of this CDS is 55% in OtV5 vs
. 41% in O. tauri,
a highly significantly difference (p<10−16
[exact binomial test]). This difference is particularly striking on synonymous positions, suggesting that the functionality of this predicted protein was maintained by natural selection following its transfer to the virus, but a difference nevertheless remains on non-synonymous positions once the synonymous changes are removed (4-fold degenerate codons, p<10−5
). The latter may arise from selective pressures for reducing the cost of viral genome replication and/or requirements for viral transcription and translation 
. We have not found this predicted gene in other viruses in databanks, but in plants proline levels increase in stress conditions such as pathogen attack, and are controlled by proline degradation, inviting speculation that this enzyme plays a role in regulating the host stress response 
. The CDS of the other HLS are smaller, but they nevertheless support the notion that exchange of sequences between host and viral genomes may be an evolutionarily frequent event. However, weak similarities between genes make it difficult to prove this with phylogenetic analyses, because of the absence of information from sufficiently closely related organisms.
The very small size of the cells and genomes of the Mamiellaceae,
together with their worldwide distribution and their ecological importance, promise that they will provide key models for interdisciplinary approaches to global ecology. Several different complete Prasinophyte genomes have already been sequenced or are in the process of being analyzed, incuding 3 Ostreococcus
, 1 Bathycoccus
sp. (unpublished data) and 2 Micromonas
spp. (A. Worden, personal communication). Viruses play a crucial role in the regulation of phytoplankton populations 
, and genomic analyses should provide insight about the biological phenomena driving selection and diversity at the base of the food web.