study, this is not required to take advantage of the fractionation approach. Our motivation for using a large volume was to ensure that we hadsufficient material to document separation at each stage using PFGE. We also anticipated that with sufficient starting volume, we might be able to avoid amplification of the material before cloning. Direct cloning would have been possible for some of the fractions, but the one we chose for analysis did 15755315 not have sufficient material. The MDA amplification step we employed has been used in other marine viral metagenomes (e.g., [14]), but can result in biases [27,40] and the formation of chimeras [39]. Such problems may explain some of the odd forward and reverse assemblies noted in the materials and methods and the repetition of genes within a contig. The increased assembly we achieved through fractionation and the long reads from Sanger sequencing make these problems more apparent. The use of improved amplificationAssembly of a Viral Metagenome after FractionationFigure 6. Annotation of ORFs in large contigs (.4 kb) assembled from the sequence library. Length and coverage of each contig are listed at right. doi:10.1371/journal.pone.0060604.gmethods [41] or elimination of the amplification step [38], coupled with increases in sequencing power [20], should further improve our ability to accurately reassemble the genomes of uncultivated viruses isolated by physical fractionation. This is a.Ports the hypothesis that, by physically 1516647 fractionating viral assemblages, there will be significantly greater reassembly of sequences from libraries constructed with the resulting fractions [21,22]. The longer contigs assembled from this fractionated viral assemblage allowed for an assessment of genes within the context of genomic fragments from uncultivated viruses. ORFs with high similarity to 2OG-Fe(II) oxygenase were found in five out of theFigure 4. Phylogenetic evaluation of DNA polymerase sequences in the sequence library. The unrooted phylogenetic tree was based on a 96 amino acid residue region of viral DNA polymerase sequences obtained from GenBank and putitive DNA polymerase sequences from this study. The letter designations P, S, and U correspond to Podoviridae, Siphoviridae, and unclassified viruses, respectively. All sequences from the Kane`ohe Bay ?library are designated with KB. Bootstrap values based on 100 resamplings are shown at the nodes if they were .50. doi:10.1371/journal.pone.0060604.gAssembly of a Viral Metagenome after FractionationFigure 5. Contig spectrum and length distribution of contigs assembled from the sequence library. (A) Histogram of the number of sequences in each contig assembled with Sequencher using conditions of 98 minimum match and .20 bp overlap. (B) Histogram of the lengths of those contigs. doi:10.1371/journal.pone.0060604.gnine analyzed contigs. This gene has so far been found exclusively in T4-like cyanophages [35], suggesting that these five contigs came from the genome of the myovirus identified in the fraction. The fact that these genes occurred in multiple contigs, but in different locations relative to other genes, indicates that there could be several types of morphologically similar myoviruses with different genome arrangements in our sequenced fraction. Alternatively, these similar contigs could be chimeric assemblies resulting from low sequence coverage (2.0?.3x), chimeras generated from MDA [39], or both. Although we used a large volume concentrate for this study, this is not required to take advantage of the fractionation approach. Our motivation for using a large volume was to ensure that we hadsufficient material to document separation at each stage using PFGE. We also anticipated that with sufficient starting volume, we might be able to avoid amplification of the material before cloning. Direct cloning would have been possible for some of the fractions, but the 
one we chose for analysis did 15755315 not have sufficient material. The MDA amplification step we employed has been used in other marine viral metagenomes (e.g., [14]), but can result in biases [27,40] and the formation of chimeras [39]. Such problems may explain some of the odd forward and reverse assemblies noted in the materials and methods and the repetition of genes within a contig. The increased assembly we achieved through fractionation and the long reads from Sanger sequencing make these problems more apparent. The use of improved amplificationAssembly of a Viral Metagenome after FractionationFigure 6. Annotation of ORFs in large contigs (.4 kb) assembled from the sequence library. Length and coverage of each contig are listed at right. doi:10.1371/journal.pone.0060604.gmethods [41] or elimination of the amplification step [38], coupled with increases in sequencing power [20], should further improve our ability to accurately reassemble the genomes of uncultivated viruses isolated by physical fractionation. This is a.